CMG实验复现

匝抽 · 4 天前

代码

代码库：https://github.com/haihuangcode/CMG

部署测试

安装依赖

git clone https://github.com/haihuangcode/CMGcd CMG#您实际上不必在txt文件中安装所有库，您可以根据需要选择安装它们。#建议使用Python 3.7，因为使用的一些库不支持更高版本的Python。conda create -n your_env_name python=3.7pip install -r requirements.txt

复制代码

下载数据集
下载新模型

AVE_ce 测试（跨模态事件分类）

关于在Audio-Visual Event Classification（音频-视觉事件分类）任务上进行实验的说明。

cd CMG/code/src./ave.sh

复制代码

修改数据地址
修改模型地址

(cmg0218) trimps@trimps-System-Product-Name:~/llm_model/shaohang/CMG/code/src$ ./ave.sh/home/trimps/anaconda3/envs/cmg0218/lib/python3.7/site-packages/gensim/similarities/__init__.py:15: UserWarning: The gensim.similarities.levenshtein submodule is disabled, because the optional Levenshtein package is unavailable. Install Levenhstein (e.g. `pip install python-Levenshtein`) to suppress this warning. warnings.warn(msg)Updating: n_epoch: 100 (default) ----> 30Updating: batch_size: 64 (default) ----> 80Updating: test_batch_size: 16 (default) ----> 128Updating: lr: 0.001 (default) ----> 0.0004Updating: snapshot_pref: ./exp/debug/123 (default) ----> ./Exps/ave/Updating: clip_gradient: 0.8 (default) ----> 0.5Updating: print_freq: 20 (default) ----> 12025-02-25 22:49:03,961 INFO Creating folder: ./Exps/ave/2025-02-25 22:49:03,961 INFO Runtime args{ "dataset_name": "ave_av", "n_epoch": 30, "batch_size": 80, "test_batch_size": 128, "lr": 0.0004, "gpu": "0", "snapshot_pref": "./Exps/ave/", "resume": "", "evaluate": false, "clip_gradient": 0.5, "loss_weights": 0.5, "start_epoch": 0, "model_save_path": "../../../checkpoint", "weight_decay": 0.0005, "print_freq": 1, "save_freq": null, "eval_freq": 1, "mi_lr": 0.001}2025-02-25 22:49:06,164 INFO Resume from number 3-th model.======> Result will be saved at: ./Exps/ave/2025-02-25 22:49:06,961 INFO Epoch 4, Batch 0, lr = 0.00040, audio_event_loss = 3.3787, audio_acc = 1.2500, 2025-02-25 22:49:07,807 INFO Epoch 4, Batch 20, lr = 0.00040, audio_event_loss = 2.8592, audio_acc = 28.8750, 2025-02-25 22:49:08,644 INFO Epoch 4, Batch 40, lr = 0.00040, audio_event_loss = 2.1006, audio_acc = 48.5000, 2025-02-25 22:49:08,719 INFO epoch: *******************************************42025-02-25 22:49:12,398 INFO ************************************************************************** Audio Evaluation results (acc): 51.0632%. Video Evaluation results (acc): 47.7688%.2025-02-25 22:49:12,405 INFO -----------------------------2025-02-25 22:49:12,405 INFO evaluate loss:1.77907929272084722025-02-25 22:49:12,405 INFO -----------------------------2025-02-25 22:49:12,833 INFO Epoch 5, Batch 0, lr = 0.00040, audio_event_loss = 1.8894, audio_acc = 47.4788, 2025-02-25 22:49:13,633 INFO Epoch 5, Batch 20, lr = 0.00040, audio_event_loss = 1.6710, audio_acc = 51.0000, 2025-02-25 22:49:14,426 INFO Epoch 5, Batch 40, lr = 0.00040, audio_event_loss = 1.3975, audio_acc = 59.5000, 2025-02-25 22:49:14,507 INFO epoch: *******************************************52025-02-25 22:49:18,087 INFO ************************************************************************** Audio Evaluation results (acc): 59.1494%. Video Evaluation results (acc): 52.8601%.2025-02-25 22:49:18,093 INFO -----------------------------2025-02-25 22:49:18,093 INFO evaluate loss:1.346460876433272025-02-25 22:49:18,093 INFO -----------------------------2025-02-25 22:49:18,471 INFO Epoch 6, Batch 0, lr = 0.00040, audio_event_loss = 1.2508, audio_acc = 64.8835, 2025-02-25 22:49:19,298 INFO Epoch 6, Batch 20, lr = 0.00040, audio_event_loss = 1.2759, audio_acc = 60.7500, 2025-02-25 22:49:20,079 INFO Epoch 6, Batch 40, lr = 0.00040, audio_event_loss = 1.3259, audio_acc = 58.0625, 2025-02-25 22:49:20,159 INFO epoch: *******************************************62025-02-25 22:49:23,762 INFO ************************************************************************** Audio Evaluation results (acc): 61.3357%. Video Evaluation results (acc): 53.3693%.2025-02-25 22:49:23,769 INFO -----------------------------2025-02-25 22:49:23,769 INFO evaluate loss:1.2458550917961912025-02-25 22:49:23,769 INFO -----------------------------2025-02-25 22:49:24,116 INFO Epoch 7, Batch 0, lr = 0.00040, audio_event_loss = 1.1836, audio_acc = 61.3983, 2025-02-25 22:49:24,953 INFO Epoch 7, Batch 20, lr = 0.00040, audio_event_loss = 1.2190, audio_acc = 62.4375, 2025-02-25 22:49:25,735 INFO Epoch 7, Batch 40, lr = 0.00040, audio_event_loss = 1.2143, audio_acc = 61.3750, 2025-02-25 22:49:25,815 INFO epoch: *******************************************72025-02-25 22:49:29,433 INFO ************************************************************************** Audio Evaluation results (acc): 61.0362%. Video Evaluation results (acc): 53.5190%.2025-02-25 22:49:29,440 INFO -----------------------------2025-02-25 22:49:29,440 INFO evaluate loss:1.19849626592896262025-02-25 22:49:29,440 INFO -----------------------------2025-02-25 22:49:29,777 INFO Epoch 8, Batch 0, lr = 0.00040, audio_event_loss = 1.0752, audio_acc = 66.3559, 2025-02-25 22:49:30,644 INFO Epoch 8, Batch 20, lr = 0.00040, audio_event_loss = 1.1692, audio_acc = 62.3750, 2025-02-25 22:49:31,440 INFO Epoch 8, Batch 40, lr = 0.00040, audio_event_loss = 1.2051, audio_acc = 61.3125, 2025-02-25 22:49:31,525 INFO epoch: *******************************************82025-02-25 22:49:35,147 INFO ************************************************************************** Audio Evaluation results (acc): 62.8332%. Video Evaluation results (acc): 53.8185%.2025-02-25 22:49:35,155 INFO -----------------------------2025-02-25 22:49:35,156 INFO evaluate loss:1.17293285652482272025-02-25 22:49:35,156 INFO -----------------------------2025-02-25 22:49:35,510 INFO Epoch 9, Batch 0, lr = 0.00040, audio_event_loss = 1.1222, audio_acc = 60.5085, 2025-02-25 22:49:36,346 INFO Epoch 9, Batch 20, lr = 0.00040, audio_event_loss = 1.1646, audio_acc = 61.8750, 2025-02-25 22:49:37,119 INFO Epoch 9, Batch 40, lr = 0.00040, audio_event_loss = 1.1671, audio_acc = 63.1250, 2025-02-25 22:49:37,199 INFO epoch: *******************************************92025-02-25 22:49:40,790 INFO ************************************************************************** Audio Evaluation results (acc): 62.6535%. Video Evaluation results (acc): 54.7769%.2025-02-25 22:49:40,798 INFO -----------------------------2025-02-25 22:49:40,798 INFO evaluate loss:1.15909351433899022025-02-25 22:49:40,798 INFO -----------------------------2025-02-25 22:49:41,171 INFO Epoch 10, Batch 0, lr = 0.00040, audio_event_loss = 1.1905, audio_acc = 64.1208, 2025-02-25 22:49:42,012 INFO Epoch 10, Batch 20, lr = 0.00040, audio_event_loss = 1.1277, audio_acc = 62.3750, 2025-02-25 22:49:42,818 INFO Epoch 10, Batch 40, lr = 0.00040, audio_event_loss = 1.1530, audio_acc = 62.9375, 2025-02-25 22:49:42,900 INFO epoch: *******************************************102025-02-25 22:49:46,521 INFO ************************************************************************** Audio Evaluation results (acc): 62.2642%. Video Evaluation results (acc): 54.1779%.2025-02-25 22:49:46,531 INFO -----------------------------2025-02-25 22:49:46,531 INFO evaluate loss:1.14741025999176082025-02-25 22:49:46,531 INFO -----------------------------2025-02-25 22:49:46,894 INFO Epoch 11, Batch 0, lr = 0.00040, audio_event_loss = 1.1051, audio_acc = 62.6059, 2025-02-25 22:49:47,750 INFO Epoch 11, Batch 20, lr = 0.00040, audio_event_loss = 1.1278, audio_acc = 63.1250, 2025-02-25 22:49:48,546 INFO Epoch 11, Batch 40, lr = 0.00040, audio_event_loss = 1.1270, audio_acc = 63.1250, 2025-02-25 22:49:48,629 INFO epoch: *******************************************112025-02-25 22:49:52,254 INFO ************************************************************************** Audio Evaluation results (acc): 63.5220%. Video Evaluation results (acc): 53.1596%.2025-02-25 22:49:52,261 INFO -----------------------------2025-02-25 22:49:52,261 INFO evaluate loss:1.13255764876723952025-02-25 22:49:52,261 INFO -----------------------------2025-02-25 22:49:52,599 INFO Epoch 12, Batch 0, lr = 0.00040, audio_event_loss = 1.1105, audio_acc = 70.1907, 2025-02-25 22:49:53,451 INFO Epoch 12, Batch 20, lr = 0.00040, audio_event_loss = 1.1312, audio_acc = 64.3750, 2025-02-25 22:49:54,245 INFO Epoch 12, Batch 40, lr = 0.00040, audio_event_loss = 1.1238, audio_acc = 63.2500, 2025-02-25 22:49:54,324 INFO epoch: *******************************************122025-02-25 22:49:57,936 INFO ************************************************************************** Audio Evaluation results (acc): 63.9114%. Video Evaluation results (acc): 54.5373%.2025-02-25 22:49:57,941 INFO -----------------------------2025-02-25 22:49:57,941 INFO evaluate loss:1.12494069427498272025-02-25 22:49:57,941 INFO -----------------------------2025-02-25 22:49:58,278 INFO Epoch 13, Batch 0, lr = 0.00040, audio_event_loss = 1.1862, audio_acc = 62.7860, 2025-02-25 22:49:59,124 INFO Epoch 13, Batch 20, lr = 0.00040, audio_event_loss = 1.1049, audio_acc = 64.6875, 2025-02-25 22:49:59,920 INFO Epoch 13, Batch 40, lr = 0.00040, audio_event_loss = 1.1351, audio_acc = 63.8125, 2025-02-25 22:50:00,003 INFO epoch: *******************************************132025-02-25 22:50:03,639 INFO ************************************************************************** Audio Evaluation results (acc): 63.7916%. Video Evaluation results (acc): 54.1180%.2025-02-25 22:50:03,647 INFO -----------------------------2025-02-25 22:50:03,647 INFO evaluate loss:1.12622890621590342025-02-25 22:50:03,647 INFO -----------------------------2025-02-25 22:50:04,006 INFO Epoch 14, Batch 0, lr = 0.00020, audio_event_loss = 1.0881, audio_acc = 64.9258, 2025-02-25 22:50:04,865 INFO Epoch 14, Batch 20, lr = 0.00020, audio_event_loss = 1.1260, audio_acc = 63.5625, 2025-02-25 22:50:05,661 INFO Epoch 14, Batch 40, lr = 0.00020, audio_event_loss = 1.0775, audio_acc = 65.3750, 2025-02-25 22:50:05,739 INFO epoch: *******************************************142025-02-25 22:50:09,351 INFO ************************************************************************** Audio Evaluation results (acc): 64.0611%. Video Evaluation results (acc): 54.6870%.2025-02-25 22:50:09,359 INFO -----------------------------2025-02-25 22:50:09,359 INFO evaluate loss:1.11240193952144882025-02-25 22:50:09,359 INFO -----------------------------2025-02-25 22:50:09,693 INFO Epoch 15, Batch 0, lr = 0.00020, audio_event_loss = 1.1069, audio_acc = 63.4534, 2025-02-25 22:50:10,544 INFO Epoch 15, Batch 20, lr = 0.00020, audio_event_loss = 1.0827, audio_acc = 64.3125, 2025-02-25 22:50:11,327 INFO Epoch 15, Batch 40, lr = 0.00020, audio_event_loss = 1.1290, audio_acc = 63.7500, 2025-02-25 22:50:11,407 INFO epoch: *******************************************152025-02-25 22:50:15,027 INFO ************************************************************************** Audio Evaluation results (acc): 64.3306%. Video Evaluation results (acc): 54.1479%.2025-02-25 22:50:15,033 INFO -----------------------------2025-02-25 22:50:15,033 INFO evaluate loss:1.10912608381916572025-02-25 22:50:15,033 INFO -----------------------------2025-02-25 22:50:15,350 INFO Epoch 16, Batch 0, lr = 0.00020, audio_event_loss = 1.0434, audio_acc = 68.5381, 2025-02-25 22:50:16,224 INFO Epoch 16, Batch 20, lr = 0.00020, audio_event_loss = 1.0926, audio_acc = 64.4375, 2025-02-25 22:50:17,010 INFO Epoch 16, Batch 40, lr = 0.00020, audio_event_loss = 1.0731, audio_acc = 65.3750, 2025-02-25 22:50:17,091 INFO epoch: *******************************************162025-02-25 22:50:20,690 INFO ************************************************************************** Audio Evaluation results (acc): 64.6900%. Video Evaluation results (acc): 54.4774%.2025-02-25 22:50:20,699 INFO -----------------------------2025-02-25 22:50:20,699 INFO evaluate loss:1.10510174604054212025-02-25 22:50:20,699 INFO -----------------------------2025-02-25 22:50:21,051 INFO Epoch 17, Batch 0, lr = 0.00020, audio_event_loss = 1.0580, audio_acc = 68.8559, 2025-02-25 22:50:21,882 INFO Epoch 17, Batch 20, lr = 0.00020, audio_event_loss = 1.1053, audio_acc = 65.0625, 2025-02-25 22:50:22,676 INFO Epoch 17, Batch 40, lr = 0.00020, audio_event_loss = 1.0881, audio_acc = 65.3750, 2025-02-25 22:50:22,753 INFO epoch: *******************************************172025-02-25 22:50:26,383 INFO ************************************************************************** Audio Evaluation results (acc): 64.3306%. Video Evaluation results (acc): 54.3276%.2025-02-25 22:50:26,390 INFO -----------------------------2025-02-25 22:50:26,390 INFO evaluate loss:1.10446162296125472025-02-25 22:50:26,390 INFO -----------------------------2025-02-25 22:50:26,726 INFO Epoch 18, Batch 0, lr = 0.00020, audio_event_loss = 1.2390, audio_acc = 57.1610, 2025-02-25 22:50:27,598 INFO Epoch 18, Batch 20, lr = 0.00020, audio_event_loss = 1.0995, audio_acc = 63.7500, 2025-02-25 22:50:28,386 INFO Epoch 18, Batch 40, lr = 0.00020, audio_event_loss = 1.0806, audio_acc = 65.5000, 2025-02-25 22:50:28,467 INFO epoch: *******************************************182025-02-25 22:50:32,087 INFO ************************************************************************** Audio Evaluation results (acc): 64.5403%. Video Evaluation results (acc): 53.7287%.2025-02-25 22:50:32,092 INFO -----------------------------2025-02-25 22:50:32,092 INFO evaluate loss:1.09899442697847732025-02-25 22:50:32,092 INFO -----------------------------2025-02-25 22:50:32,455 INFO Epoch 19, Batch 0, lr = 0.00020, audio_event_loss = 1.1770, audio_acc = 55.6886, 2025-02-25 22:50:33,282 INFO Epoch 19, Batch 20, lr = 0.00020, audio_event_loss = 1.0712, audio_acc = 65.6250, 2025-02-25 22:50:34,074 INFO Epoch 19, Batch 40, lr = 0.00020, audio_event_loss = 1.1119, audio_acc = 63.6250, 2025-02-25 22:50:34,158 INFO epoch: *******************************************192025-02-25 22:50:37,768 INFO ************************************************************************** Audio Evaluation results (acc): 64.5403%. Video Evaluation results (acc): 53.9683%.2025-02-25 22:50:37,776 INFO -----------------------------2025-02-25 22:50:37,776 INFO evaluate loss:1.09748691698295972025-02-25 22:50:37,776 INFO -----------------------------2025-02-25 22:50:38,116 INFO Epoch 20, Batch 0, lr = 0.00020, audio_event_loss = 0.9353, audio_acc = 70.5932, 2025-02-25 22:50:38,964 INFO Epoch 20, Batch 20, lr = 0.00020, audio_event_loss = 1.0799, audio_acc = 64.7500, 2025-02-25 22:50:39,753 INFO Epoch 20, Batch 40, lr = 0.00020, audio_event_loss = 1.1023, audio_acc = 64.4375, 2025-02-25 22:50:39,836 INFO epoch: *******************************************202025-02-25 22:50:43,458 INFO ************************************************************************** Audio Evaluation results (acc): 64.8398%. Video Evaluation results (acc): 53.7886%.2025-02-25 22:50:43,465 INFO -----------------------------2025-02-25 22:50:43,465 INFO evaluate loss:1.0938891372398282025-02-25 22:50:43,465 INFO -----------------------------2025-02-25 22:50:43,833 INFO Epoch 21, Batch 0, lr = 0.00020, audio_event_loss = 1.0697, audio_acc = 68.0932, 2025-02-25 22:50:44,660 INFO Epoch 21, Batch 20, lr = 0.00020, audio_event_loss = 1.0936, audio_acc = 63.6875, 2025-02-25 22:50:45,461 INFO Epoch 21, Batch 40, lr = 0.00020, audio_event_loss = 1.0874, audio_acc = 64.4375, 2025-02-25 22:50:45,541 INFO epoch: *******************************************212025-02-25 22:50:49,167 INFO ************************************************************************** Audio Evaluation results (acc): 64.7499%. Video Evaluation results (acc): 53.9982%.2025-02-25 22:50:49,172 INFO -----------------------------2025-02-25 22:50:49,172 INFO evaluate loss:1.09082316973100252025-02-25 22:50:49,172 INFO -----------------------------2025-02-25 22:50:49,529 INFO Epoch 22, Batch 0, lr = 0.00020, audio_event_loss = 1.0849, audio_acc = 65.5508, 2025-02-25 22:50:50,373 INFO Epoch 22, Batch 20, lr = 0.00020, audio_event_loss = 1.0798, audio_acc = 64.7500, 2025-02-25 22:50:51,180 INFO Epoch 22, Batch 40, lr = 0.00020, audio_event_loss = 1.0754, audio_acc = 64.2500, 2025-02-25 22:50:51,260 INFO epoch: *******************************************222025-02-25 22:50:54,908 INFO ************************************************************************** Audio Evaluation results (acc): 64.9296%. Video Evaluation results (acc): 53.6089%.2025-02-25 22:50:54,917 INFO -----------------------------2025-02-25 22:50:54,917 INFO evaluate loss:1.0891852723341772025-02-25 22:50:54,917 INFO -----------------------------2025-02-25 22:50:55,272 INFO Epoch 23, Batch 0, lr = 0.00020, audio_event_loss = 1.0584, audio_acc = 64.0784, 2025-02-25 22:50:56,118 INFO Epoch 23, Batch 20, lr = 0.00020, audio_event_loss = 1.0616, audio_acc = 65.6250, 2025-02-25 22:50:56,916 INFO Epoch 23, Batch 40, lr = 0.00020, audio_event_loss = 1.0963, audio_acc = 65.1250, 2025-02-25 22:50:56,994 INFO epoch: *******************************************232025-02-25 22:51:00,618 INFO ************************************************************************** Audio Evaluation results (acc): 64.7499%. Video Evaluation results (acc): 54.4774%.2025-02-25 22:51:00,625 INFO -----------------------------2025-02-25 22:51:00,625 INFO evaluate loss:1.08903499428683432025-02-25 22:51:00,625 INFO -----------------------------2025-02-25 22:51:00,953 INFO Epoch 24, Batch 0, lr = 0.00010, audio_event_loss = 1.1219, audio_acc = 61.9809, 2025-02-25 22:51:01,805 INFO Epoch 24, Batch 20, lr = 0.00010, audio_event_loss = 1.1022, audio_acc = 65.3750, 2025-02-25 22:51:02,604 INFO Epoch 24, Batch 40, lr = 0.00010, audio_event_loss = 1.0363, audio_acc = 66.0625, 2025-02-25 22:51:02,686 INFO epoch: *******************************************242025-02-25 22:51:06,320 INFO ************************************************************************** Audio Evaluation results (acc): 64.9296%. Video Evaluation results (acc): 54.1479%.2025-02-25 22:51:06,332 INFO -----------------------------2025-02-25 22:51:06,332 INFO evaluate loss:1.0844910427453582025-02-25 22:51:06,332 INFO -----------------------------2025-02-25 22:51:06,688 INFO Epoch 25, Batch 0, lr = 0.00010, audio_event_loss = 1.1838, audio_acc = 60.0636, 2025-02-25 22:51:07,549 INFO Epoch 25, Batch 20, lr = 0.00010, audio_event_loss = 1.0651, audio_acc = 66.3125, 2025-02-25 22:51:08,358 INFO Epoch 25, Batch 40, lr = 0.00010, audio_event_loss = 1.0806, audio_acc = 65.1250, 2025-02-25 22:51:08,435 INFO epoch: *******************************************252025-02-25 22:51:12,010 INFO ************************************************************************** Audio Evaluation results (acc): 65.1692%. Video Evaluation results (acc): 53.8784%.2025-02-25 22:51:12,017 INFO -----------------------------2025-02-25 22:51:12,017 INFO evaluate loss:1.08343957047652232025-02-25 22:51:12,017 INFO -----------------------------2025-02-25 22:51:12,375 INFO Epoch 26, Batch 0, lr = 0.00010, audio_event_loss = 1.1671, audio_acc = 60.5085, 2025-02-25 22:51:13,219 INFO Epoch 26, Batch 20, lr = 0.00010, audio_event_loss = 1.0860, audio_acc = 65.1875, 2025-02-25 22:51:14,011 INFO Epoch 26, Batch 40, lr = 0.00010, audio_event_loss = 1.0314, audio_acc = 65.3750, 2025-02-25 22:51:14,093 INFO epoch: *******************************************262025-02-25 22:51:17,728 INFO ************************************************************************** Audio Evaluation results (acc): 65.0494%. Video Evaluation results (acc): 54.0282%.2025-02-25 22:51:17,736 INFO -----------------------------2025-02-25 22:51:17,736 INFO evaluate loss:1.08086563652297122025-02-25 22:51:17,736 INFO -----------------------------2025-02-25 22:51:18,079 INFO Epoch 27, Batch 0, lr = 0.00010, audio_event_loss = 1.0689, audio_acc = 62.7436, 2025-02-25 22:51:18,935 INFO Epoch 27, Batch 20, lr = 0.00010, audio_event_loss = 1.0897, audio_acc = 64.4375, 2025-02-25 22:51:19,740 INFO Epoch 27, Batch 40, lr = 0.00010, audio_event_loss = 1.0360, audio_acc = 66.1250, 2025-02-25 22:51:19,824 INFO epoch: *******************************************272025-02-25 22:51:23,459 INFO ************************************************************************** Audio Evaluation results (acc): 65.0195%. Video Evaluation results (acc): 54.0581%.2025-02-25 22:51:23,466 INFO -----------------------------2025-02-25 22:51:23,466 INFO evaluate loss:1.08015872257692822025-02-25 22:51:23,466 INFO -----------------------------2025-02-25 22:51:23,845 INFO Epoch 28, Batch 0, lr = 0.00010, audio_event_loss = 1.0505, audio_acc = 67.6059, 2025-02-25 22:51:24,677 INFO Epoch 28, Batch 20, lr = 0.00010, audio_event_loss = 1.0574, audio_acc = 66.3750, 2025-02-25 22:51:25,467 INFO Epoch 28, Batch 40, lr = 0.00010, audio_event_loss = 1.0767, audio_acc = 64.4375, 2025-02-25 22:51:25,547 INFO epoch: *******************************************282025-02-25 22:51:29,162 INFO ************************************************************************** Audio Evaluation results (acc): 64.9895%. Video Evaluation results (acc): 53.6987%.2025-02-25 22:51:29,169 INFO -----------------------------2025-02-25 22:51:29,169 INFO evaluate loss:1.07948279209303262025-02-25 22:51:29,169 INFO -----------------------------2025-02-25 22:51:29,539 INFO Epoch 29, Batch 0, lr = 0.00010, audio_event_loss = 1.0822, audio_acc = 64.0784, 2025-02-25 22:51:30,358 INFO Epoch 29, Batch 20, lr = 0.00010, audio_event_loss = 1.0333, audio_acc = 65.7500, 2025-02-25 22:51:31,157 INFO Epoch 29, Batch 40, lr = 0.00010, audio_event_loss = 1.0874, audio_acc = 65.6250, 2025-02-25 22:51:31,239 INFO epoch: *******************************************292025-02-25 22:51:34,886 INFO ************************************************************************** Audio Evaluation results (acc): 64.8398%. Video Evaluation results (acc): 54.2078%.2025-02-25 22:51:34,892 INFO -----------------------------2025-02-25 22:51:34,892 INFO evaluate loss:1.07895582939038982025-02-25 22:51:34,892 INFO -----------------------------

复制代码

准确率变化
loss值变化

AVVP 测试（跨模态事件定位）

关于在Audio-Video-Text Visual Prompt（音频-视频-文本视觉提示）任务上进行实验的说明。

cd CMG/code/src./avvp.sh

复制代码

修改数据位置和模型位置
测试日志

(cmg0218) trimps@trimps-System-Product-Name:~/llm_model/shaohang/CMG/code/src$ ./avvp.sh/home/trimps/anaconda3/envs/cmg0218/lib/python3.7/site-packages/gensim/similarities/__init__.py:15: UserWarning: The gensim.similarities.levenshtein submodule is disabled, because the optional Levenshtein package is unavailable. Install Levenhstein (e.g. `pip install python-Levenshtein`) to suppress this warning. warnings.warn(msg)Updating: n_epoch: 100 (default) ----> 50Updating: batch_size: 64 (default) ----> 80Updating: test_batch_size: 16 (default) ----> 64Updating: lr: 0.001 (default) ----> 0.0004Updating: snapshot_pref: ./exp/debug/123 (default) ----> ./Exps/avvp/Updating: clip_gradient: 0.8 (default) ----> 0.5Updating: print_freq: 20 (default) ----> 12025-02-24 23:09:49,743 INFO Creating folder: ./Exps/avvp/2025-02-24 23:09:49,743 INFO Runtime args{ "dataset_name": "avvp_av", "n_epoch": 50, "batch_size": 80, "test_batch_size": 64, "lr": 0.0004, "gpu": "0", "snapshot_pref": "./Exps/avvp/", "resume": "", "evaluate": false, "clip_gradient": 0.5, "loss_weights": 0.5, "start_epoch": 0, "model_save_path": "../../../checkpoint", "weight_decay": 0.0005, "print_freq": 1, "save_freq": null, "eval_freq": 1, "mi_lr": 0.001}total 25 positive classes in AVVP, 1 negative classes in AVVP1846 samples are used for traintotal 25 positive classes in AVVP, 1 negative classes in AVVP1661 samples are used for val2025-02-24 23:09:52,704 INFO Resume from number 3-th model.======> Result will be saved at: ./Exps/avvp/2025-02-24 23:09:53,461 INFO Epoch 4, Batch 0, lr = 0.00040, audio_event_loss = 4.7932, BCELoss = 0.6823, ExpLogLoss = 4.1109, audio_precision = 0.0350, audio_recall = 0.2680, 2025-02-24 23:09:53,553 INFO Epoch 4, Batch 10, lr = 0.00040, audio_event_loss = 4.0355, BCELoss = 0.5949, ExpLogLoss = 3.4406, audio_precision = 0.1442, audio_recall = 0.6028, 2025-02-24 23:09:53,632 INFO Epoch 4, Batch 20, lr = 0.00040, audio_event_loss = 3.3955, BCELoss = 0.4651, ExpLogLoss = 2.9304, audio_precision = 0.2211, audio_recall = 0.7164, 2025-02-24 23:09:53,679 INFO ********************************************** Train results (accuracy and recall): 0.1806 0.6495.2025-02-24 23:09:53,679 INFO epoch: *******************************************42025-02-24 23:09:55,455 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.18420.6217 0.2842.********************************************** Best results (accuracy and recall F1-score): 0.1842 0.6217 0.2842.2025-02-24 23:09:55,462 INFO -----------------------------2025-02-24 23:09:55,462 INFO evaluate loss:3.08250684568933232025-02-24 23:09:55,462 INFO -----------------------------2025-02-24 23:09:55,700 INFO Epoch 5, Batch 0, lr = 0.00040, audio_event_loss = 3.1642, BCELoss = 0.3997, ExpLogLoss = 2.7645, audio_precision = 0.2243, audio_recall = 0.7171, 2025-02-24 23:09:55,788 INFO Epoch 5, Batch 10, lr = 0.00040, audio_event_loss = 2.8190, BCELoss = 0.3277, ExpLogLoss = 2.4913, audio_precision = 0.2534, audio_recall = 0.7409, 2025-02-24 23:09:55,886 INFO Epoch 5, Batch 20, lr = 0.00040, audio_event_loss = 2.5919, BCELoss = 0.2531, ExpLogLoss = 2.3388, audio_precision = 0.3299, audio_recall = 0.7275, 2025-02-24 23:09:55,947 INFO ********************************************** Train results (accuracy and recall): 0.2921 0.7241.2025-02-24 23:09:55,948 INFO epoch: *******************************************52025-02-24 23:09:57,864 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.26850.5206 0.3543.********************************************** Best results (accuracy and recall F1-score): 0.2685 0.5206 0.3543.2025-02-24 23:09:57,874 INFO -----------------------------2025-02-24 23:09:57,874 INFO evaluate loss:2.55429252384211572025-02-24 23:09:57,874 INFO -----------------------------2025-02-24 23:09:58,119 INFO Epoch 6, Batch 0, lr = 0.00040, audio_event_loss = 2.3741, BCELoss = 0.2130, ExpLogLoss = 2.1611, audio_precision = 0.3493, audio_recall = 0.6677, 2025-02-24 23:09:58,207 INFO Epoch 6, Batch 10, lr = 0.00040, audio_event_loss = 2.4270, BCELoss = 0.1893, ExpLogLoss = 2.2377, audio_precision = 0.3988, audio_recall = 0.6038, 2025-02-24 23:09:58,285 INFO Epoch 6, Batch 20, lr = 0.00040, audio_event_loss = 2.2415, BCELoss = 0.1613, ExpLogLoss = 2.0802, audio_precision = 0.4689, audio_recall = 0.5500, 2025-02-24 23:09:58,344 INFO ********************************************** Train results (accuracy and recall): 0.4425 0.5723.2025-02-24 23:09:58,344 INFO epoch: *******************************************62025-02-24 23:10:00,202 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.45870.3674 0.4080.********************************************** Best results (accuracy and recall F1-score): 0.4587 0.3674 0.4080.2025-02-24 23:10:00,210 INFO -----------------------------2025-02-24 23:10:00,210 INFO evaluate loss:2.32677171840050432025-02-24 23:10:00,210 INFO -----------------------------2025-02-24 23:10:00,430 INFO Epoch 7, Batch 0, lr = 0.00040, audio_event_loss = 2.2682, BCELoss = 0.1505, ExpLogLoss = 2.1176, audio_precision = 0.5663, audio_recall = 0.4815, 2025-02-24 23:10:00,513 INFO Epoch 7, Batch 10, lr = 0.00040, audio_event_loss = 2.2050, BCELoss = 0.1396, ExpLogLoss = 2.0655, audio_precision = 0.5717, audio_recall = 0.3900, 2025-02-24 23:10:00,587 INFO Epoch 7, Batch 20, lr = 0.00040, audio_event_loss = 2.1945, BCELoss = 0.1313, ExpLogLoss = 2.0633, audio_precision = 0.5940, audio_recall = 0.3284, 2025-02-24 23:10:00,643 INFO ********************************************** Train results (accuracy and recall): 0.5821 0.3570.2025-02-24 23:10:00,644 INFO epoch: *******************************************72025-02-24 23:10:02,379 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.57400.2900 0.3854.********************************************** Best results (accuracy and recall F1-score): 0.5740 0.2900 0.3854.2025-02-24 23:10:02,387 INFO -----------------------------2025-02-24 23:10:02,387 INFO evaluate loss:2.2638215912077532025-02-24 23:10:02,387 INFO -----------------------------2025-02-24 23:10:02,597 INFO Epoch 8, Batch 0, lr = 0.00040, audio_event_loss = 2.2994, BCELoss = 0.1323, ExpLogLoss = 2.1671, audio_precision = 0.6037, audio_recall = 0.2675, 2025-02-24 23:10:02,686 INFO Epoch 8, Batch 10, lr = 0.00040, audio_event_loss = 2.1017, BCELoss = 0.1231, ExpLogLoss = 1.9786, audio_precision = 0.6645, audio_recall = 0.2935, 2025-02-24 23:10:02,760 INFO Epoch 8, Batch 20, lr = 0.00040, audio_event_loss = 2.2092, BCELoss = 0.1257, ExpLogLoss = 2.0835, audio_precision = 0.6525, audio_recall = 0.2741, 2025-02-24 23:10:02,818 INFO ********************************************** Train results (accuracy and recall): 0.6651 0.2886.2025-02-24 23:10:02,818 INFO epoch: *******************************************82025-02-24 23:10:04,626 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.65320.2620 0.3740.********************************************** Best results (accuracy and recall F1-score): 0.6532 0.2620 0.3740.2025-02-24 23:10:04,635 INFO -----------------------------2025-02-24 23:10:04,635 INFO evaluate loss:2.21306634133485372025-02-24 23:10:04,635 INFO -----------------------------2025-02-24 23:10:04,856 INFO Epoch 9, Batch 0, lr = 0.00040, audio_event_loss = 2.0096, BCELoss = 0.1166, ExpLogLoss = 1.8930, audio_precision = 0.7260, audio_recall = 0.3093, 2025-02-24 23:10:04,943 INFO Epoch 9, Batch 10, lr = 0.00040, audio_event_loss = 2.0876, BCELoss = 0.1203, ExpLogLoss = 1.9672, audio_precision = 0.6966, audio_recall = 0.2604, 2025-02-24 23:10:05,016 INFO Epoch 9, Batch 20, lr = 0.00040, audio_event_loss = 2.0921, BCELoss = 0.1193, ExpLogLoss = 1.9728, audio_precision = 0.7243, audio_recall = 0.2865, 2025-02-24 23:10:05,077 INFO ********************************************** Train results (accuracy and recall): 0.7123 0.2723.2025-02-24 23:10:05,078 INFO epoch: *******************************************92025-02-24 23:10:06,873 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.65980.2531 0.3658.********************************************** Best results (accuracy and recall F1-score): 0.6598 0.2531 0.3658.2025-02-24 23:10:06,880 INFO -----------------------------2025-02-24 23:10:06,880 INFO evaluate loss:2.1796725594193492025-02-24 23:10:06,880 INFO -----------------------------2025-02-24 23:10:07,094 INFO Epoch 10, Batch 0, lr = 0.00040, audio_event_loss = 2.1362, BCELoss = 0.1210, ExpLogLoss = 2.0153, audio_precision = 0.7773, audio_recall = 0.2641, 2025-02-24 23:10:07,172 INFO Epoch 10, Batch 10, lr = 0.00040, audio_event_loss = 2.1015, BCELoss = 0.1199, ExpLogLoss = 1.9816, audio_precision = 0.6976, audio_recall = 0.2650, 2025-02-24 23:10:07,248 INFO Epoch 10, Batch 20, lr = 0.00040, audio_event_loss = 2.0244, BCELoss = 0.1168, ExpLogLoss = 1.9077, audio_precision = 0.7071, audio_recall = 0.2890, 2025-02-24 23:10:07,304 INFO ********************************************** Train results (accuracy and recall): 0.7002 0.2723.2025-02-24 23:10:07,304 INFO epoch: *******************************************102025-02-24 23:10:09,090 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.65210.2570 0.3687.********************************************** Best results (accuracy and recall F1-score): 0.6598 0.2531 0.3658.2025-02-24 23:10:09,098 INFO -----------------------------2025-02-24 23:10:09,098 INFO evaluate loss:2.1884148720099062025-02-24 23:10:09,098 INFO -----------------------------2025-02-24 23:10:09,315 INFO Epoch 11, Batch 0, lr = 0.00040, audio_event_loss = 2.2082, BCELoss = 0.1234, ExpLogLoss = 2.0848, audio_precision = 0.6253, audio_recall = 0.2072, 2025-02-24 23:10:09,394 INFO Epoch 11, Batch 10, lr = 0.00040, audio_event_loss = 2.0326, BCELoss = 0.1178, ExpLogLoss = 1.9148, audio_precision = 0.6721, audio_recall = 0.2547, 2025-02-24 23:10:09,469 INFO Epoch 11, Batch 20, lr = 0.00040, audio_event_loss = 2.0887, BCELoss = 0.1190, ExpLogLoss = 1.9697, audio_precision = 0.7122, audio_recall = 0.2678, 2025-02-24 23:10:09,526 INFO ********************************************** Train results (accuracy and recall): 0.6976 0.2618.2025-02-24 23:10:09,527 INFO epoch: *******************************************112025-02-24 23:10:11,354 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.68840.2577 0.3751.********************************************** Best results (accuracy and recall F1-score): 0.6884 0.2577 0.3751.2025-02-24 23:10:11,361 INFO -----------------------------2025-02-24 23:10:11,361 INFO evaluate loss:2.1898242253293652025-02-24 23:10:11,361 INFO -----------------------------2025-02-24 23:10:11,572 INFO Epoch 12, Batch 0, lr = 0.00040, audio_event_loss = 1.9223, BCELoss = 0.1153, ExpLogLoss = 1.8071, audio_precision = 0.6545, audio_recall = 0.2565, 2025-02-24 23:10:11,663 INFO Epoch 12, Batch 10, lr = 0.00040, audio_event_loss = 2.0429, BCELoss = 0.1180, ExpLogLoss = 1.9249, audio_precision = 0.7051, audio_recall = 0.2713, 2025-02-24 23:10:11,738 INFO Epoch 12, Batch 20, lr = 0.00040, audio_event_loss = 2.0326, BCELoss = 0.1171, ExpLogLoss = 1.9155, audio_precision = 0.7192, audio_recall = 0.2516, 2025-02-24 23:10:11,796 INFO ********************************************** Train results (accuracy and recall): 0.7185 0.2619.2025-02-24 23:10:11,796 INFO epoch: *******************************************122025-02-24 23:10:13,669 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6547 0.2547 0.3667.********************************************** Best results (accuracy and recall F1-score): 0.6884 0.2577 0.3751.2025-02-24 23:10:13,677 INFO -----------------------------2025-02-24 23:10:13,677 INFO evaluate loss:2.1621538775963022025-02-24 23:10:13,677 INFO -----------------------------2025-02-24 23:10:13,892 INFO Epoch 13, Batch 0, lr = 0.00040, audio_event_loss = 1.9701, BCELoss = 0.1138, ExpLogLoss = 1.8563, audio_precision = 0.8148, audio_recall = 0.2673, 2025-02-24 23:10:13,986 INFO Epoch 13, Batch 10, lr = 0.00040, audio_event_loss = 2.0368, BCELoss = 0.1178, ExpLogLoss = 1.9190, audio_precision = 0.6876, audio_recall = 0.2638, 2025-02-24 23:10:14,053 INFO Epoch 13, Batch 20, lr = 0.00040, audio_event_loss = 1.9808, BCELoss = 0.1158, ExpLogLoss = 1.8650, audio_precision = 0.7409, audio_recall = 0.2586, 2025-02-24 23:10:14,114 INFO ********************************************** Train results (accuracy and recall): 0.7164 0.2666.2025-02-24 23:10:14,114 INFO epoch: *******************************************132025-02-24 23:10:15,958 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6912 0.2439 0.3606.********************************************** Best results (accuracy and recall F1-score): 0.6912 0.2439 0.3606.2025-02-24 23:10:15,965 INFO -----------------------------2025-02-24 23:10:15,965 INFO evaluate loss:2.1911976494817772025-02-24 23:10:15,965 INFO -----------------------------2025-02-24 23:10:16,184 INFO Epoch 14, Batch 0, lr = 0.00020, audio_event_loss = 1.7924, BCELoss = 0.1085, ExpLogLoss = 1.6839, audio_precision = 0.6866, audio_recall = 0.2856, 2025-02-24 23:10:16,268 INFO Epoch 14, Batch 10, lr = 0.00020, audio_event_loss = 2.0137, BCELoss = 0.1174, ExpLogLoss = 1.8963, audio_precision = 0.6996, audio_recall = 0.2782, 2025-02-24 23:10:16,339 INFO Epoch 14, Batch 20, lr = 0.00020, audio_event_loss = 1.9676, BCELoss = 0.1145, ExpLogLoss = 1.8531, audio_precision = 0.7171, audio_recall = 0.2747, 2025-02-24 23:10:16,399 INFO ********************************************** Train results (accuracy and recall): 0.7035 0.2709.2025-02-24 23:10:16,400 INFO epoch: *******************************************142025-02-24 23:10:18,234 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6742 0.2474 0.3619.********************************************** Best results (accuracy and recall F1-score): 0.6912 0.2439 0.3606.2025-02-24 23:10:18,241 INFO -----------------------------2025-02-24 23:10:18,241 INFO evaluate loss:2.15974237111927852025-02-24 23:10:18,241 INFO -----------------------------2025-02-24 23:10:18,454 INFO Epoch 15, Batch 0, lr = 0.00020, audio_event_loss = 2.0228, BCELoss = 0.1165, ExpLogLoss = 1.9064, audio_precision = 0.6905, audio_recall = 0.2387, 2025-02-24 23:10:18,533 INFO Epoch 15, Batch 10, lr = 0.00020, audio_event_loss = 1.9058, BCELoss = 0.1129, ExpLogLoss = 1.7930, audio_precision = 0.7138, audio_recall = 0.2963, 2025-02-24 23:10:18,608 INFO Epoch 15, Batch 20, lr = 0.00020, audio_event_loss = 1.9677, BCELoss = 0.1160, ExpLogLoss = 1.8517, audio_precision = 0.7007, audio_recall = 0.2680, 2025-02-24 23:10:18,664 INFO ********************************************** Train results (accuracy and recall): 0.7031 0.2728.2025-02-24 23:10:18,664 INFO epoch: *******************************************152025-02-24 23:10:20,489 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7142 0.2381 0.3571.********************************************** Best results (accuracy and recall F1-score): 0.7142 0.2381 0.3571.2025-02-24 23:10:20,498 INFO -----------------------------2025-02-24 23:10:20,498 INFO evaluate loss:2.1172603377992842025-02-24 23:10:20,498 INFO -----------------------------2025-02-24 23:10:20,725 INFO Epoch 16, Batch 0, lr = 0.00020, audio_event_loss = 2.1110, BCELoss = 0.1209, ExpLogLoss = 1.9901, audio_precision = 0.6293, audio_recall = 0.2323, 2025-02-24 23:10:20,813 INFO Epoch 16, Batch 10, lr = 0.00020, audio_event_loss = 1.9815, BCELoss = 0.1166, ExpLogLoss = 1.8649, audio_precision = 0.7029, audio_recall = 0.2591, 2025-02-24 23:10:20,888 INFO Epoch 16, Batch 20, lr = 0.00020, audio_event_loss = 1.9202, BCELoss = 0.1136, ExpLogLoss = 1.8065, audio_precision = 0.7046, audio_recall = 0.2636, 2025-02-24 23:10:20,952 INFO ********************************************** Train results (accuracy and recall): 0.7108 0.2674.2025-02-24 23:10:20,952 INFO epoch: *******************************************162025-02-24 23:10:22,842 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6948 0.2590 0.3774.********************************************** Best results (accuracy and recall F1-score): 0.7142 0.2381 0.3571.2025-02-24 23:10:22,850 INFO -----------------------------2025-02-24 23:10:22,850 INFO evaluate loss:2.1262691784548942025-02-24 23:10:22,850 INFO -----------------------------2025-02-24 23:10:23,079 INFO Epoch 17, Batch 0, lr = 0.00020, audio_event_loss = 2.0544, BCELoss = 0.1177, ExpLogLoss = 1.9367, audio_precision = 0.5775, audio_recall = 0.2015, 2025-02-24 23:10:23,156 INFO Epoch 17, Batch 10, lr = 0.00020, audio_event_loss = 1.9366, BCELoss = 0.1148, ExpLogLoss = 1.8218, audio_precision = 0.7085, audio_recall = 0.2673, 2025-02-24 23:10:23,230 INFO Epoch 17, Batch 20, lr = 0.00020, audio_event_loss = 1.9681, BCELoss = 0.1149, ExpLogLoss = 1.8531, audio_precision = 0.6972, audio_recall = 0.2676, 2025-02-24 23:10:23,290 INFO ********************************************** Train results (accuracy and recall): 0.7034 0.2686.2025-02-24 23:10:23,290 INFO epoch: *******************************************172025-02-24 23:10:25,119 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:25,127 INFO -----------------------------2025-02-24 23:10:25,127 INFO evaluate loss:2.11901677243453972025-02-24 23:10:25,127 INFO -----------------------------2025-02-24 23:10:25,355 INFO Epoch 18, Batch 0, lr = 0.00020, audio_event_loss = 1.8343, BCELoss = 0.1116, ExpLogLoss = 1.7227, audio_precision = 0.6424, audio_recall = 0.2454, 2025-02-24 23:10:25,449 INFO Epoch 18, Batch 10, lr = 0.00020, audio_event_loss = 1.9572, BCELoss = 0.1154, ExpLogLoss = 1.8418, audio_precision = 0.7050, audio_recall = 0.2601, 2025-02-24 23:10:25,525 INFO Epoch 18, Batch 20, lr = 0.00020, audio_event_loss = 1.9284, BCELoss = 0.1145, ExpLogLoss = 1.8138, audio_precision = 0.7414, audio_recall = 0.2554, 2025-02-24 23:10:25,583 INFO ********************************************** Train results (accuracy and recall): 0.7206 0.2621.2025-02-24 23:10:25,584 INFO epoch: *******************************************182025-02-24 23:10:27,430 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6913 0.2511 0.3684.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:27,439 INFO -----------------------------2025-02-24 23:10:27,439 INFO evaluate loss:2.1219954176516962025-02-24 23:10:27,439 INFO -----------------------------2025-02-24 23:10:27,672 INFO Epoch 19, Batch 0, lr = 0.00020, audio_event_loss = 1.8900, BCELoss = 0.1130, ExpLogLoss = 1.7770, audio_precision = 0.7745, audio_recall = 0.2795, 2025-02-24 23:10:27,762 INFO Epoch 19, Batch 10, lr = 0.00020, audio_event_loss = 1.9334, BCELoss = 0.1148, ExpLogLoss = 1.8186, audio_precision = 0.6957, audio_recall = 0.2635, 2025-02-24 23:10:27,833 INFO Epoch 19, Batch 20, lr = 0.00020, audio_event_loss = 1.9452, BCELoss = 0.1151, ExpLogLoss = 1.8302, audio_precision = 0.7089, audio_recall = 0.2701, 2025-02-24 23:10:27,894 INFO ********************************************** Train results (accuracy and recall): 0.7052 0.2715.2025-02-24 23:10:27,894 INFO epoch: *******************************************192025-02-24 23:10:29,749 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7081 0.2419 0.3606.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:29,759 INFO -----------------------------2025-02-24 23:10:29,759 INFO evaluate loss:2.11517984713934442025-02-24 23:10:29,759 INFO -----------------------------2025-02-24 23:10:29,975 INFO Epoch 20, Batch 0, lr = 0.00020, audio_event_loss = 1.6041, BCELoss = 0.1033, ExpLogLoss = 1.5007, audio_precision = 0.7884, audio_recall = 0.3587, 2025-02-24 23:10:30,074 INFO Epoch 20, Batch 10, lr = 0.00020, audio_event_loss = 1.9217, BCELoss = 0.1142, ExpLogLoss = 1.8075, audio_precision = 0.7041, audio_recall = 0.2502, 2025-02-24 23:10:30,148 INFO Epoch 20, Batch 20, lr = 0.00020, audio_event_loss = 1.8924, BCELoss = 0.1133, ExpLogLoss = 1.7790, audio_precision = 0.7428, audio_recall = 0.2728, 2025-02-24 23:10:30,209 INFO ********************************************** Train results (accuracy and recall): 0.7198 0.2609.2025-02-24 23:10:30,210 INFO epoch: *******************************************202025-02-24 23:10:32,043 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7133 0.2528 0.3733.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:32,050 INFO -----------------------------2025-02-24 23:10:32,050 INFO evaluate loss:2.1012781339883462025-02-24 23:10:32,050 INFO -----------------------------2025-02-24 23:10:32,275 INFO Epoch 21, Batch 0, lr = 0.00020, audio_event_loss = 1.8187, BCELoss = 0.1117, ExpLogLoss = 1.7069, audio_precision = 0.5740, audio_recall = 0.2317, 2025-02-24 23:10:32,371 INFO Epoch 21, Batch 10, lr = 0.00020, audio_event_loss = 1.9517, BCELoss = 0.1162, ExpLogLoss = 1.8355, audio_precision = 0.6989, audio_recall = 0.2788, 2025-02-24 23:10:32,448 INFO Epoch 21, Batch 20, lr = 0.00020, audio_event_loss = 1.8406, BCELoss = 0.1110, ExpLogLoss = 1.7296, audio_precision = 0.7316, audio_recall = 0.2686, 2025-02-24 23:10:32,511 INFO ********************************************** Train results (accuracy and recall): 0.7066 0.2709.2025-02-24 23:10:32,511 INFO epoch: *******************************************212025-02-24 23:10:34,361 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6830 0.2470 0.3628.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:34,371 INFO -----------------------------2025-02-24 23:10:34,371 INFO evaluate loss:2.0768740503948772025-02-24 23:10:34,371 INFO -----------------------------2025-02-24 23:10:34,587 INFO Epoch 22, Batch 0, lr = 0.00020, audio_event_loss = 1.9792, BCELoss = 0.1155, ExpLogLoss = 1.8637, audio_precision = 0.7016, audio_recall = 0.2986, 2025-02-24 23:10:34,670 INFO Epoch 22, Batch 10, lr = 0.00020, audio_event_loss = 1.8986, BCELoss = 0.1148, ExpLogLoss = 1.7838, audio_precision = 0.6816, audio_recall = 0.2618, 2025-02-24 23:10:34,745 INFO Epoch 22, Batch 20, lr = 0.00020, audio_event_loss = 1.8926, BCELoss = 0.1129, ExpLogLoss = 1.7797, audio_precision = 0.7171, audio_recall = 0.2796, 2025-02-24 23:10:34,802 INFO ********************************************** Train results (accuracy and recall): 0.7014 0.2729.2025-02-24 23:10:34,803 INFO epoch: *******************************************222025-02-24 23:10:36,603 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7146 0.2412 0.3607.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:36,612 INFO -----------------------------2025-02-24 23:10:36,612 INFO evaluate loss:2.0691793718330862025-02-24 23:10:36,612 INFO -----------------------------2025-02-24 23:10:36,824 INFO Epoch 23, Batch 0, lr = 0.00020, audio_event_loss = 1.9020, BCELoss = 0.1131, ExpLogLoss = 1.7889, audio_precision = 0.7520, audio_recall = 0.3014, 2025-02-24 23:10:36,908 INFO Epoch 23, Batch 10, lr = 0.00020, audio_event_loss = 1.9072, BCELoss = 0.1137, ExpLogLoss = 1.7934, audio_precision = 0.7196, audio_recall = 0.2911, 2025-02-24 23:10:36,983 INFO Epoch 23, Batch 20, lr = 0.00020, audio_event_loss = 1.9056, BCELoss = 0.1148, ExpLogLoss = 1.7908, audio_precision = 0.7074, audio_recall = 0.2706, 2025-02-24 23:10:37,044 INFO ********************************************** Train results (accuracy and recall): 0.7127 0.2783.2025-02-24 23:10:37,044 INFO epoch: *******************************************232025-02-24 23:10:38,874 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7030 0.2575 0.3769.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:38,882 INFO -----------------------------2025-02-24 23:10:38,882 INFO evaluate loss:2.0858724733522862025-02-24 23:10:38,882 INFO -----------------------------2025-02-24 23:10:39,094 INFO Epoch 24, Batch 0, lr = 0.00010, audio_event_loss = 1.7531, BCELoss = 0.1094, ExpLogLoss = 1.6437, audio_precision = 0.7634, audio_recall = 0.2627, 2025-02-24 23:10:39,183 INFO Epoch 24, Batch 10, lr = 0.00010, audio_event_loss = 1.9263, BCELoss = 0.1153, ExpLogLoss = 1.8109, audio_precision = 0.7020, audio_recall = 0.2797, 2025-02-24 23:10:39,258 INFO Epoch 24, Batch 20, lr = 0.00010, audio_event_loss = 1.8074, BCELoss = 0.1107, ExpLogLoss = 1.6967, audio_precision = 0.7061, audio_recall = 0.2634, 2025-02-24 23:10:39,321 INFO ********************************************** Train results (accuracy and recall): 0.7049 0.2734.2025-02-24 23:10:39,322 INFO epoch: *******************************************242025-02-24 23:10:41,171 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7027 0.2465 0.3649.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:41,177 INFO -----------------------------2025-02-24 23:10:41,177 INFO evaluate loss:2.0890464154276182025-02-24 23:10:41,177 INFO -----------------------------2025-02-24 23:10:41,391 INFO Epoch 25, Batch 0, lr = 0.00010, audio_event_loss = 1.8848, BCELoss = 0.1124, ExpLogLoss = 1.7724, audio_precision = 0.7423, audio_recall = 0.3007, 2025-02-24 23:10:41,476 INFO Epoch 25, Batch 10, lr = 0.00010, audio_event_loss = 1.8704, BCELoss = 0.1126, ExpLogLoss = 1.7578, audio_precision = 0.7168, audio_recall = 0.2659, 2025-02-24 23:10:41,552 INFO Epoch 25, Batch 20, lr = 0.00010, audio_event_loss = 1.9106, BCELoss = 0.1152, ExpLogLoss = 1.7954, audio_precision = 0.7027, audio_recall = 0.2791, 2025-02-24 23:10:41,615 INFO ********************************************** Train results (accuracy and recall): 0.7051 0.2742.2025-02-24 23:10:41,615 INFO epoch: *******************************************252025-02-24 23:10:43,450 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7022 0.2461 0.3645.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:43,460 INFO -----------------------------2025-02-24 23:10:43,460 INFO evaluate loss:2.0872008662302952025-02-24 23:10:43,460 INFO -----------------------------2025-02-24 23:10:43,675 INFO Epoch 26, Batch 0, lr = 0.00010, audio_event_loss = 1.7383, BCELoss = 0.1108, ExpLogLoss = 1.6275, audio_precision = 0.5863, audio_recall = 0.2319, 2025-02-24 23:10:43,758 INFO Epoch 26, Batch 10, lr = 0.00010, audio_event_loss = 1.8747, BCELoss = 0.1129, ExpLogLoss = 1.7618, audio_precision = 0.7397, audio_recall = 0.2623, 2025-02-24 23:10:43,834 INFO Epoch 26, Batch 20, lr = 0.00010, audio_event_loss = 1.8764, BCELoss = 0.1140, ExpLogLoss = 1.7624, audio_precision = 0.6964, audio_recall = 0.2659, 2025-02-24 23:10:43,896 INFO ********************************************** Train results (accuracy and recall): 0.7145 0.2642.2025-02-24 23:10:43,897 INFO epoch: *******************************************262025-02-24 23:10:45,707 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7100 0.2458 0.3652.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:45,716 INFO -----------------------------2025-02-24 23:10:45,716 INFO evaluate loss:2.06576726781772062025-02-24 23:10:45,716 INFO -----------------------------2025-02-24 23:10:45,936 INFO Epoch 27, Batch 0, lr = 0.00010, audio_event_loss = 1.8331, BCELoss = 0.1142, ExpLogLoss = 1.7190, audio_precision = 0.6145, audio_recall = 0.2701, 2025-02-24 23:10:46,023 INFO Epoch 27, Batch 10, lr = 0.00010, audio_event_loss = 1.7938, BCELoss = 0.1102, ExpLogLoss = 1.6836, audio_precision = 0.7216, audio_recall = 0.2857, 2025-02-24 23:10:46,097 INFO Epoch 27, Batch 20, lr = 0.00010, audio_event_loss = 1.9478, BCELoss = 0.1163, ExpLogLoss = 1.8315, audio_precision = 0.7129, audio_recall = 0.2459, 2025-02-24 23:10:46,157 INFO ********************************************** Train results (accuracy and recall): 0.7130 0.2694.2025-02-24 23:10:46,157 INFO epoch: *******************************************272025-02-24 23:10:48,005 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7029 0.2492 0.3679.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:48,014 INFO -----------------------------2025-02-24 23:10:48,014 INFO evaluate loss:2.08227583518019752025-02-24 23:10:48,014 INFO -----------------------------2025-02-24 23:10:48,229 INFO Epoch 28, Batch 0, lr = 0.00010, audio_event_loss = 2.1328, BCELoss = 0.1206, ExpLogLoss = 2.0122, audio_precision = 0.7194, audio_recall = 0.3090, 2025-02-24 23:10:48,320 INFO Epoch 28, Batch 10, lr = 0.00010, audio_event_loss = 1.8628, BCELoss = 0.1131, ExpLogLoss = 1.7497, audio_precision = 0.7109, audio_recall = 0.2841, 2025-02-24 23:10:48,394 INFO Epoch 28, Batch 20, lr = 0.00010, audio_event_loss = 1.8744, BCELoss = 0.1135, ExpLogLoss = 1.7609, audio_precision = 0.6857, audio_recall = 0.2738, 2025-02-24 23:10:48,458 INFO ********************************************** Train results (accuracy and recall): 0.7047 0.2750.2025-02-24 23:10:48,459 INFO epoch: *******************************************282025-02-24 23:10:50,326 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6895 0.2480 0.3648.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:50,333 INFO -----------------------------2025-02-24 23:10:50,334 INFO evaluate loss:2.07120333142666852025-02-24 23:10:50,334 INFO -----------------------------2025-02-24 23:10:50,563 INFO Epoch 29, Batch 0, lr = 0.00010, audio_event_loss = 1.7241, BCELoss = 0.1076, ExpLogLoss = 1.6164, audio_precision = 0.7715, audio_recall = 0.3245, 2025-02-24 23:10:50,648 INFO Epoch 29, Batch 10, lr = 0.00010, audio_event_loss = 1.8426, BCELoss = 0.1124, ExpLogLoss = 1.7302, audio_precision = 0.7151, audio_recall = 0.2718, 2025-02-24 23:10:50,724 INFO Epoch 29, Batch 20, lr = 0.00010, audio_event_loss = 1.8762, BCELoss = 0.1140, ExpLogLoss = 1.7622, audio_precision = 0.7132, audio_recall = 0.2715, 2025-02-24 23:10:50,789 INFO ********************************************** Train results (accuracy and recall): 0.7166 0.2729.2025-02-24 23:10:50,789 INFO epoch: *******************************************292025-02-24 23:10:52,575 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7086 0.2443 0.3633.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:52,585 INFO -----------------------------2025-02-24 23:10:52,585 INFO evaluate loss:2.07001840744042152025-02-24 23:10:52,585 INFO -----------------------------2025-02-24 23:10:52,796 INFO Epoch 30, Batch 0, lr = 0.00010, audio_event_loss = 1.7444, BCELoss = 0.1096, ExpLogLoss = 1.6348, audio_precision = 0.6577, audio_recall = 0.2546, 2025-02-24 23:10:52,869 INFO Epoch 30, Batch 10, lr = 0.00010, audio_event_loss = 1.8792, BCELoss = 0.1144, ExpLogLoss = 1.7649, audio_precision = 0.7161, audio_recall = 0.2629, 2025-02-24 23:10:52,944 INFO Epoch 30, Batch 20, lr = 0.00010, audio_event_loss = 1.8294, BCELoss = 0.1120, ExpLogLoss = 1.7174, audio_precision = 0.7096, audio_recall = 0.2866, 2025-02-24 23:10:53,003 INFO ********************************************** Train results (accuracy and recall): 0.7099 0.2734.2025-02-24 23:10:53,003 INFO epoch: *******************************************302025-02-24 23:10:54,852 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6958 0.2453 0.3627.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:54,862 INFO -----------------------------2025-02-24 23:10:54,862 INFO evaluate loss:2.0739177645622562025-02-24 23:10:54,862 INFO -----------------------------2025-02-24 23:10:55,086 INFO Epoch 31, Batch 0, lr = 0.00010, audio_event_loss = 2.1193, BCELoss = 0.1204, ExpLogLoss = 1.9989, audio_precision = 0.7186, audio_recall = 0.2403, 2025-02-24 23:10:55,180 INFO Epoch 31, Batch 10, lr = 0.00010, audio_event_loss = 1.8448, BCELoss = 0.1122, ExpLogLoss = 1.7325, audio_precision = 0.7181, audio_recall = 0.2822, 2025-02-24 23:10:55,254 INFO Epoch 31, Batch 20, lr = 0.00010, audio_event_loss = 1.8520, BCELoss = 0.1139, ExpLogLoss = 1.7381, audio_precision = 0.7013, audio_recall = 0.2463, 2025-02-24 23:10:55,312 INFO ********************************************** Train results (accuracy and recall): 0.7086 0.2649.2025-02-24 23:10:55,313 INFO epoch: *******************************************312025-02-24 23:10:57,090 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6972 0.2460 0.3637.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:57,100 INFO -----------------------------2025-02-24 23:10:57,100 INFO evaluate loss:2.05511727167641342025-02-24 23:10:57,100 INFO -----------------------------2025-02-24 23:10:57,327 INFO Epoch 32, Batch 0, lr = 0.00010, audio_event_loss = 1.8917, BCELoss = 0.1126, ExpLogLoss = 1.7791, audio_precision = 0.7088, audio_recall = 0.2527, 2025-02-24 23:10:57,421 INFO Epoch 32, Batch 10, lr = 0.00010, audio_event_loss = 1.8387, BCELoss = 0.1131, ExpLogLoss = 1.7257, audio_precision = 0.7027, audio_recall = 0.2762, 2025-02-24 23:10:57,495 INFO Epoch 32, Batch 20, lr = 0.00010, audio_event_loss = 1.8281, BCELoss = 0.1116, ExpLogLoss = 1.7166, audio_precision = 0.7145, audio_recall = 0.2808, 2025-02-24 23:10:57,557 INFO ********************************************** Train results (accuracy and recall): 0.7076 0.2792.2025-02-24 23:10:57,557 INFO epoch: *******************************************322025-02-24 23:10:59,407 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6957 0.2465 0.3640.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:10:59,415 INFO -----------------------------2025-02-24 23:10:59,415 INFO evaluate loss:2.05517091645300942025-02-24 23:10:59,415 INFO -----------------------------2025-02-24 23:10:59,636 INFO Epoch 33, Batch 0, lr = 0.00010, audio_event_loss = 1.7572, BCELoss = 0.1094, ExpLogLoss = 1.6478, audio_precision = 0.7320, audio_recall = 0.3296, 2025-02-24 23:10:59,721 INFO Epoch 33, Batch 10, lr = 0.00010, audio_event_loss = 1.8846, BCELoss = 0.1144, ExpLogLoss = 1.7702, audio_precision = 0.7190, audio_recall = 0.2744, 2025-02-24 23:10:59,795 INFO Epoch 33, Batch 20, lr = 0.00010, audio_event_loss = 1.8372, BCELoss = 0.1130, ExpLogLoss = 1.7242, audio_precision = 0.6941, audio_recall = 0.2502, 2025-02-24 23:10:59,859 INFO ********************************************** Train results (accuracy and recall): 0.7140 0.2690.2025-02-24 23:10:59,859 INFO epoch: *******************************************332025-02-24 23:11:01,682 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6942 0.2433 0.3603.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:01,692 INFO -----------------------------2025-02-24 23:11:01,692 INFO evaluate loss:2.0660383527136852025-02-24 23:11:01,692 INFO -----------------------------2025-02-24 23:11:01,912 INFO Epoch 34, Batch 0, lr = 0.00005, audio_event_loss = 1.6853, BCELoss = 0.1042, ExpLogLoss = 1.5811, audio_precision = 0.7931, audio_recall = 0.2907, 2025-02-24 23:11:01,991 INFO Epoch 34, Batch 10, lr = 0.00005, audio_event_loss = 1.8749, BCELoss = 0.1140, ExpLogLoss = 1.7609, audio_precision = 0.7105, audio_recall = 0.2740, 2025-02-24 23:11:02,066 INFO Epoch 34, Batch 20, lr = 0.00005, audio_event_loss = 1.8493, BCELoss = 0.1135, ExpLogLoss = 1.7359, audio_precision = 0.7000, audio_recall = 0.2645, 2025-02-24 23:11:02,127 INFO ********************************************** Train results (accuracy and recall): 0.7132 0.2684.2025-02-24 23:11:02,128 INFO epoch: *******************************************342025-02-24 23:11:03,965 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6896 0.2473 0.3640.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:03,974 INFO -----------------------------2025-02-24 23:11:03,974 INFO evaluate loss:2.0587680045615562025-02-24 23:11:03,974 INFO -----------------------------2025-02-24 23:11:04,188 INFO Epoch 35, Batch 0, lr = 0.00005, audio_event_loss = 1.6495, BCELoss = 0.1030, ExpLogLoss = 1.5465, audio_precision = 0.8264, audio_recall = 0.2590, 2025-02-24 23:11:04,278 INFO Epoch 35, Batch 10, lr = 0.00005, audio_event_loss = 1.8630, BCELoss = 0.1144, ExpLogLoss = 1.7486, audio_precision = 0.6835, audio_recall = 0.2650, 2025-02-24 23:11:04,353 INFO Epoch 35, Batch 20, lr = 0.00005, audio_event_loss = 1.7709, BCELoss = 0.1095, ExpLogLoss = 1.6614, audio_precision = 0.7212, audio_recall = 0.3012, 2025-02-24 23:11:04,419 INFO ********************************************** Train results (accuracy and recall): 0.7040 0.2775.2025-02-24 23:11:04,419 INFO epoch: *******************************************352025-02-24 23:11:06,248 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6946 0.2448 0.3620.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:06,257 INFO -----------------------------2025-02-24 23:11:06,257 INFO evaluate loss:2.0565578680133792025-02-24 23:11:06,258 INFO -----------------------------2025-02-24 23:11:06,478 INFO Epoch 36, Batch 0, lr = 0.00005, audio_event_loss = 2.2623, BCELoss = 0.1264, ExpLogLoss = 2.1359, audio_precision = 0.7475, audio_recall = 0.2632, 2025-02-24 23:11:06,558 INFO Epoch 36, Batch 10, lr = 0.00005, audio_event_loss = 1.8946, BCELoss = 0.1149, ExpLogLoss = 1.7797, audio_precision = 0.6903, audio_recall = 0.2741, 2025-02-24 23:11:06,633 INFO Epoch 36, Batch 20, lr = 0.00005, audio_event_loss = 1.8046, BCELoss = 0.1119, ExpLogLoss = 1.6927, audio_precision = 0.7066, audio_recall = 0.2763, 2025-02-24 23:11:06,694 INFO ********************************************** Train results (accuracy and recall): 0.7001 0.2804.2025-02-24 23:11:06,694 INFO epoch: *******************************************362025-02-24 23:11:08,592 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6902 0.2495 0.3665.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:08,601 INFO -----------------------------2025-02-24 23:11:08,601 INFO evaluate loss:2.0519131691955612025-02-24 23:11:08,601 INFO -----------------------------2025-02-24 23:11:08,822 INFO Epoch 37, Batch 0, lr = 0.00005, audio_event_loss = 1.6755, BCELoss = 0.1056, ExpLogLoss = 1.5699, audio_precision = 0.6716, audio_recall = 0.2972, 2025-02-24 23:11:08,911 INFO Epoch 37, Batch 10, lr = 0.00005, audio_event_loss = 1.8298, BCELoss = 0.1122, ExpLogLoss = 1.7176, audio_precision = 0.7306, audio_recall = 0.2678, 2025-02-24 23:11:08,987 INFO Epoch 37, Batch 20, lr = 0.00005, audio_event_loss = 1.8349, BCELoss = 0.1129, ExpLogLoss = 1.7220, audio_precision = 0.6846, audio_recall = 0.2814, 2025-02-24 23:11:09,048 INFO ********************************************** Train results (accuracy and recall): 0.7099 0.2734.2025-02-24 23:11:09,048 INFO epoch: *******************************************372025-02-24 23:11:10,917 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6957 0.2465 0.3640.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:10,926 INFO -----------------------------2025-02-24 23:11:10,926 INFO evaluate loss:2.05239362829944572025-02-24 23:11:10,926 INFO -----------------------------2025-02-24 23:11:11,142 INFO Epoch 38, Batch 0, lr = 0.00005, audio_event_loss = 1.7441, BCELoss = 0.1100, ExpLogLoss = 1.6341, audio_precision = 0.7282, audio_recall = 0.2929, 2025-02-24 23:11:11,231 INFO Epoch 38, Batch 10, lr = 0.00005, audio_event_loss = 1.8289, BCELoss = 0.1131, ExpLogLoss = 1.7158, audio_precision = 0.6907, audio_recall = 0.2779, 2025-02-24 23:11:11,305 INFO Epoch 38, Batch 20, lr = 0.00005, audio_event_loss = 1.8137, BCELoss = 0.1118, ExpLogLoss = 1.7019, audio_precision = 0.7288, audio_recall = 0.2655, 2025-02-24 23:11:11,366 INFO ********************************************** Train results (accuracy and recall): 0.7115 0.2729.2025-02-24 23:11:11,367 INFO epoch: *******************************************382025-02-24 23:11:13,212 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6940 0.2449 0.3620.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:13,221 INFO -----------------------------2025-02-24 23:11:13,221 INFO evaluate loss:2.05097390770293762025-02-24 23:11:13,221 INFO -----------------------------2025-02-24 23:11:13,430 INFO Epoch 39, Batch 0, lr = 0.00005, audio_event_loss = 1.9156, BCELoss = 0.1151, ExpLogLoss = 1.8005, audio_precision = 0.6730, audio_recall = 0.2481, 2025-02-24 23:11:13,517 INFO Epoch 39, Batch 10, lr = 0.00005, audio_event_loss = 1.7953, BCELoss = 0.1108, ExpLogLoss = 1.6845, audio_precision = 0.7097, audio_recall = 0.2851, 2025-02-24 23:11:13,593 INFO Epoch 39, Batch 20, lr = 0.00005, audio_event_loss = 1.8257, BCELoss = 0.1123, ExpLogLoss = 1.7134, audio_precision = 0.7202, audio_recall = 0.2705, 2025-02-24 23:11:13,649 INFO ********************************************** Train results (accuracy and recall): 0.7095 0.2709.2025-02-24 23:11:13,649 INFO epoch: *******************************************392025-02-24 23:11:15,479 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6957 0.2465 0.3640.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:15,486 INFO -----------------------------2025-02-24 23:11:15,486 INFO evaluate loss:2.04957481071634762025-02-24 23:11:15,486 INFO -----------------------------2025-02-24 23:11:15,700 INFO Epoch 40, Batch 0, lr = 0.00005, audio_event_loss = 1.8231, BCELoss = 0.1146, ExpLogLoss = 1.7085, audio_precision = 0.6840, audio_recall = 0.2683, 2025-02-24 23:11:15,785 INFO Epoch 40, Batch 10, lr = 0.00005, audio_event_loss = 1.8569, BCELoss = 0.1140, ExpLogLoss = 1.7429, audio_precision = 0.7122, audio_recall = 0.2803, 2025-02-24 23:11:15,862 INFO Epoch 40, Batch 20, lr = 0.00005, audio_event_loss = 1.8051, BCELoss = 0.1115, ExpLogLoss = 1.6935, audio_precision = 0.7128, audio_recall = 0.2692, 2025-02-24 23:11:15,923 INFO ********************************************** Train results (accuracy and recall): 0.7155 0.2733.2025-02-24 23:11:15,923 INFO epoch: *******************************************402025-02-24 23:11:17,759 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6957 0.2465 0.3640.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:17,766 INFO -----------------------------2025-02-24 23:11:17,767 INFO evaluate loss:2.0423852076610922025-02-24 23:11:17,767 INFO -----------------------------2025-02-24 23:11:17,981 INFO Epoch 41, Batch 0, lr = 0.00005, audio_event_loss = 1.8908, BCELoss = 0.1163, ExpLogLoss = 1.7746, audio_precision = 0.7262, audio_recall = 0.2511, 2025-02-24 23:11:18,065 INFO Epoch 41, Batch 10, lr = 0.00005, audio_event_loss = 1.8075, BCELoss = 0.1117, ExpLogLoss = 1.6958, audio_precision = 0.7175, audio_recall = 0.2767, 2025-02-24 23:11:18,140 INFO Epoch 41, Batch 20, lr = 0.00005, audio_event_loss = 1.8647, BCELoss = 0.1144, ExpLogLoss = 1.7503, audio_precision = 0.7126, audio_recall = 0.2521, 2025-02-24 23:11:18,200 INFO ********************************************** Train results (accuracy and recall): 0.7162 0.2644.2025-02-24 23:11:18,200 INFO epoch: *******************************************412025-02-24 23:11:19,973 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6957 0.2465 0.3640.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:19,980 INFO -----------------------------2025-02-24 23:11:19,980 INFO evaluate loss:2.04580797392750042025-02-24 23:11:19,980 INFO -----------------------------2025-02-24 23:11:20,191 INFO Epoch 42, Batch 0, lr = 0.00005, audio_event_loss = 1.8193, BCELoss = 0.1074, ExpLogLoss = 1.7118, audio_precision = 0.7318, audio_recall = 0.3399, 2025-02-24 23:11:20,278 INFO Epoch 42, Batch 10, lr = 0.00005, audio_event_loss = 1.7908, BCELoss = 0.1106, ExpLogLoss = 1.6802, audio_precision = 0.7287, audio_recall = 0.2754, 2025-02-24 23:11:20,343 INFO Epoch 42, Batch 20, lr = 0.00005, audio_event_loss = 1.8231, BCELoss = 0.1128, ExpLogLoss = 1.7103, audio_precision = 0.7071, audio_recall = 0.2654, 2025-02-24 23:11:20,404 INFO ********************************************** Train results (accuracy and recall): 0.7134 0.2701.2025-02-24 23:11:20,404 INFO epoch: *******************************************422025-02-24 23:11:22,255 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6960 0.2474 0.3650.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:22,264 INFO -----------------------------2025-02-24 23:11:22,264 INFO evaluate loss:2.04105481651683542025-02-24 23:11:22,264 INFO -----------------------------2025-02-24 23:11:22,492 INFO Epoch 43, Batch 0, lr = 0.00005, audio_event_loss = 1.9421, BCELoss = 0.1155, ExpLogLoss = 1.8265, audio_precision = 0.7456, audio_recall = 0.2509, 2025-02-24 23:11:22,573 INFO Epoch 43, Batch 10, lr = 0.00005, audio_event_loss = 1.7838, BCELoss = 0.1112, ExpLogLoss = 1.6725, audio_precision = 0.6937, audio_recall = 0.2949, 2025-02-24 23:11:22,631 INFO Epoch 43, Batch 20, lr = 0.00005, audio_event_loss = 1.8564, BCELoss = 0.1142, ExpLogLoss = 1.7421, audio_precision = 0.7053, audio_recall = 0.2724, 2025-02-24 23:11:22,691 INFO ********************************************** Train results (accuracy and recall): 0.6996 0.2826.2025-02-24 23:11:22,691 INFO epoch: *******************************************432025-02-24 23:11:24,520 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6890 0.2467 0.3633.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:24,527 INFO -----------------------------2025-02-24 23:11:24,527 INFO evaluate loss:2.0418201469982522025-02-24 23:11:24,527 INFO -----------------------------2025-02-24 23:11:24,743 INFO Epoch 44, Batch 0, lr = 0.00005, audio_event_loss = 1.6721, BCELoss = 0.1085, ExpLogLoss = 1.5636, audio_precision = 0.6435, audio_recall = 0.3095, 2025-02-24 23:11:24,827 INFO Epoch 44, Batch 10, lr = 0.00005, audio_event_loss = 1.7758, BCELoss = 0.1109, ExpLogLoss = 1.6649, audio_precision = 0.7404, audio_recall = 0.2826, 2025-02-24 23:11:24,902 INFO Epoch 44, Batch 20, lr = 0.00005, audio_event_loss = 1.8453, BCELoss = 0.1133, ExpLogLoss = 1.7320, audio_precision = 0.7022, audio_recall = 0.2635, 2025-02-24 23:11:24,964 INFO ********************************************** Train results (accuracy and recall): 0.7175 0.2728.2025-02-24 23:11:24,964 INFO epoch: *******************************************442025-02-24 23:11:26,790 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6871 0.2434 0.3595.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:26,800 INFO -----------------------------2025-02-24 23:11:26,800 INFO evaluate loss:2.0512118848134312025-02-24 23:11:26,800 INFO -----------------------------2025-02-24 23:11:27,017 INFO Epoch 45, Batch 0, lr = 0.00005, audio_event_loss = 1.9382, BCELoss = 0.1157, ExpLogLoss = 1.8225, audio_precision = 0.6433, audio_recall = 0.2156, 2025-02-24 23:11:27,103 INFO Epoch 45, Batch 10, lr = 0.00005, audio_event_loss = 1.7842, BCELoss = 0.1109, ExpLogLoss = 1.6732, audio_precision = 0.7151, audio_recall = 0.2705, 2025-02-24 23:11:27,179 INFO Epoch 45, Batch 20, lr = 0.00005, audio_event_loss = 1.8341, BCELoss = 0.1127, ExpLogLoss = 1.7214, audio_precision = 0.7052, audio_recall = 0.2691, 2025-02-24 23:11:27,239 INFO ********************************************** Train results (accuracy and recall): 0.7082 0.2708.2025-02-24 23:11:27,239 INFO epoch: *******************************************452025-02-24 23:11:29,087 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6953 0.2465 0.3639.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:29,096 INFO -----------------------------2025-02-24 23:11:29,096 INFO evaluate loss:2.0407774808896172025-02-24 23:11:29,096 INFO -----------------------------2025-02-24 23:11:29,307 INFO Epoch 46, Batch 0, lr = 0.00005, audio_event_loss = 1.7513, BCELoss = 0.1101, ExpLogLoss = 1.6412, audio_precision = 0.6916, audio_recall = 0.2702, 2025-02-24 23:11:29,400 INFO Epoch 46, Batch 10, lr = 0.00005, audio_event_loss = 1.7843, BCELoss = 0.1110, ExpLogLoss = 1.6733, audio_precision = 0.7095, audio_recall = 0.2877, 2025-02-24 23:11:29,475 INFO Epoch 46, Batch 20, lr = 0.00005, audio_event_loss = 1.8251, BCELoss = 0.1135, ExpLogLoss = 1.7117, audio_precision = 0.6734, audio_recall = 0.2768, 2025-02-24 23:11:29,534 INFO ********************************************** Train results (accuracy and recall): 0.6880 0.2822.2025-02-24 23:11:29,534 INFO epoch: *******************************************462025-02-24 23:11:31,343 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6960 0.2474 0.3650.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:31,353 INFO -----------------------------2025-02-24 23:11:31,353 INFO evaluate loss:2.0384766662636072025-02-24 23:11:31,353 INFO -----------------------------2025-02-24 23:11:31,569 INFO Epoch 47, Batch 0, lr = 0.00005, audio_event_loss = 1.7327, BCELoss = 0.1063, ExpLogLoss = 1.6264, audio_precision = 0.7443, audio_recall = 0.3258, 2025-02-24 23:11:31,646 INFO Epoch 47, Batch 10, lr = 0.00005, audio_event_loss = 1.8223, BCELoss = 0.1134, ExpLogLoss = 1.7089, audio_precision = 0.6939, audio_recall = 0.2565, 2025-02-24 23:11:31,722 INFO Epoch 47, Batch 20, lr = 0.00005, audio_event_loss = 1.8053, BCELoss = 0.1112, ExpLogLoss = 1.6941, audio_precision = 0.7234, audio_recall = 0.2744, 2025-02-24 23:11:31,786 INFO ********************************************** Train results (accuracy and recall): 0.7154 0.2719.2025-02-24 23:11:31,787 INFO epoch: *******************************************472025-02-24 23:11:33,621 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6880 0.2451 0.3614.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:33,630 INFO -----------------------------2025-02-24 23:11:33,630 INFO evaluate loss:2.03805067692948552025-02-24 23:11:33,630 INFO -----------------------------2025-02-24 23:11:33,851 INFO Epoch 48, Batch 0, lr = 0.00005, audio_event_loss = 2.0729, BCELoss = 0.1247, ExpLogLoss = 1.9482, audio_precision = 0.7137, audio_recall = 0.2892, 2025-02-24 23:11:33,925 INFO Epoch 48, Batch 10, lr = 0.00005, audio_event_loss = 1.7914, BCELoss = 0.1111, ExpLogLoss = 1.6803, audio_precision = 0.6968, audio_recall = 0.2650, 2025-02-24 23:11:33,999 INFO Epoch 48, Batch 20, lr = 0.00005, audio_event_loss = 1.8174, BCELoss = 0.1129, ExpLogLoss = 1.7045, audio_precision = 0.7118, audio_recall = 0.2716, 2025-02-24 23:11:34,061 INFO ********************************************** Train results (accuracy and recall): 0.7044 0.2673.2025-02-24 23:11:34,061 INFO epoch: *******************************************482025-02-24 23:11:35,900 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6929 0.2421 0.3588.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:35,908 INFO -----------------------------2025-02-24 23:11:35,908 INFO evaluate loss:2.03806623204426932025-02-24 23:11:35,908 INFO -----------------------------2025-02-24 23:11:36,120 INFO Epoch 49, Batch 0, lr = 0.00005, audio_event_loss = 1.7591, BCELoss = 0.1134, ExpLogLoss = 1.6457, audio_precision = 0.5992, audio_recall = 0.2121, 2025-02-24 23:11:36,204 INFO Epoch 49, Batch 10, lr = 0.00005, audio_event_loss = 1.8135, BCELoss = 0.1125, ExpLogLoss = 1.7010, audio_precision = 0.7147, audio_recall = 0.2746, 2025-02-24 23:11:36,278 INFO Epoch 49, Batch 20, lr = 0.00005, audio_event_loss = 1.7821, BCELoss = 0.1110, ExpLogLoss = 1.6711, audio_precision = 0.7161, audio_recall = 0.2868, 2025-02-24 23:11:36,337 INFO ********************************************** Train results (accuracy and recall): 0.7050 0.2711.2025-02-24 23:11:36,337 INFO epoch: *******************************************492025-02-24 23:11:38,179 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6957 0.2465 0.3640.********************************************** Best results (accuracy and recall F1-score): 0.7163 0.2548 0.3759.2025-02-24 23:11:38,190 INFO -----------------------------2025-02-24 23:11:38,190 INFO evaluate loss:2.0328498010687382025-02-24 23:11:38,190 INFO -----------------------------

复制代码

准确率、召回率和F1值变化
loss值变化

AVE_AVVP 测试（跨模态检索）

关于在Audio-Video-Text Visual Prompt任务上同时使用音频-视觉和音频-视觉-文本数据进行实验的说明。

cd CMG/code/src./ave_avvp.sh

复制代码

修改数据地址
修改模型地址

(cmg0218) trimps@trimps-System-Product-Name:~/llm_model/shaohang/CMG/code/src$ ./ave_avvp.sh/home/trimps/anaconda3/envs/cmg0218/lib/python3.7/site-packages/gensim/similarities/__init__.py:15: UserWarning: The gensim.similarities.levenshtein submodule is disabled, because the optional Levenshtein package is unavailable. Install Levenhstein (e.g. `pip install python-Levenshtein`) to suppress this warning. warnings.warn(msg)Updating: n_epoch: 100 (default) ----> 50Updating: batch_size: 64 (default) ----> 80Updating: test_batch_size: 16 (default) ----> 64Updating: lr: 0.001 (default) ----> 0.0004Updating: snapshot_pref: ./exp/debug/123 (default) ----> ./Exps/ave_avvp/Updating: clip_gradient: 0.8 (default) ----> 0.5Updating: print_freq: 20 (default) ----> 12025-02-25 22:56:41,858 INFO Creating folder: ./Exps/ave_avvp/2025-02-25 22:56:41,858 INFO Runtime args{ "dataset_name": "ave_avvp_av", "n_epoch": 50, "batch_size": 80, "test_batch_size": 64, "lr": 0.0004, "gpu": "0", "snapshot_pref": "./Exps/ave_avvp/", "resume": "", "evaluate": false, "clip_gradient": 0.5, "loss_weights": 0.5, "start_epoch": 0, "model_save_path": "../../../checkpoint", "weight_decay": 0.0005, "print_freq": 1, "save_freq": null, "eval_freq": 1, "mi_lr": 0.001}total 12 positive classes in AVVP, 1 negative classes in AVVP1032 samples are used for val2025-02-25 22:56:44,054 INFO Resume from number 3-th model.======> Result will be saved at: ./Exps/ave_avvp/2025-02-25 22:56:44,706 INFO Epoch 4, Batch 0, lr = 0.00040, audio_event_loss = 3.0712, BCELoss = 0.6861, ExpLogLoss = 2.3851, audio_precision = 0.0936, audio_recall = 0.5013, 2025-02-25 22:56:45,472 INFO Epoch 4, Batch 20, lr = 0.00040, audio_event_loss = 2.2275, BCELoss = 0.4730, ExpLogLoss = 1.7545, audio_precision = 0.3066, audio_recall = 0.6616, 2025-02-25 22:56:45,580 INFO ********************************************** Train results (accuracy and recall): 0.3103 0.6535.2025-02-25 22:56:45,586 INFO epoch: *******************************************42025-02-25 22:56:47,050 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.3931 0.5146 0.4457.********************************************** Best results (accuracy and recall F1-score): 0.3931 0.5146 0.4457.2025-02-25 22:56:47,057 INFO -----------------------------2025-02-25 22:56:47,057 INFO evaluate loss:1.99487407270156332025-02-25 22:56:47,057 INFO -----------------------------2025-02-25 22:56:47,439 INFO Epoch 5, Batch 0, lr = 0.00040, audio_event_loss = 1.6375, BCELoss = 0.3070, ExpLogLoss = 1.3305, audio_precision = 0.4664, audio_recall = 0.6378, 2025-02-25 22:56:48,310 INFO Epoch 5, Batch 20, lr = 0.00040, audio_event_loss = 1.3368, BCELoss = 0.2219, ExpLogLoss = 1.1149, audio_precision = 0.5981, audio_recall = 0.6048, 2025-02-25 22:56:48,427 INFO ********************************************** Train results (accuracy and recall): 0.6006 0.5998.2025-02-25 22:56:48,438 INFO epoch: *******************************************52025-02-25 22:56:49,912 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.5638 0.4264 0.4856.********************************************** Best results (accuracy and recall F1-score): 0.5638 0.4264 0.4856.2025-02-25 22:56:49,919 INFO -----------------------------2025-02-25 22:56:49,919 INFO evaluate loss:1.6401680183056682025-02-25 22:56:49,919 INFO -----------------------------2025-02-25 22:56:50,293 INFO Epoch 6, Batch 0, lr = 0.00040, audio_event_loss = 1.1927, BCELoss = 0.1723, ExpLogLoss = 1.0204, audio_precision = 0.6896, audio_recall = 0.5181, 2025-02-25 22:56:51,097 INFO Epoch 6, Batch 20, lr = 0.00040, audio_event_loss = 1.1117, BCELoss = 0.1539, ExpLogLoss = 0.9578, audio_precision = 0.7120, audio_recall = 0.4884, 2025-02-25 22:56:51,212 INFO ********************************************** Train results (accuracy and recall): 0.7158 0.4907.2025-02-25 22:56:51,218 INFO epoch: *******************************************62025-02-25 22:56:52,649 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6529 0.4151 0.5075.********************************************** Best results (accuracy and recall F1-score): 0.6529 0.4151 0.5075.2025-02-25 22:56:52,656 INFO -----------------------------2025-02-25 22:56:52,656 INFO evaluate loss:1.58020014045227082025-02-25 22:56:52,656 INFO -----------------------------2025-02-25 22:56:52,991 INFO Epoch 7, Batch 0, lr = 0.00040, audio_event_loss = 1.0686, BCELoss = 0.1436, ExpLogLoss = 0.9251, audio_precision = 0.7526, audio_recall = 0.5302, 2025-02-25 22:56:53,798 INFO Epoch 7, Batch 20, lr = 0.00040, audio_event_loss = 1.0586, BCELoss = 0.1424, ExpLogLoss = 0.9162, audio_precision = 0.7403, audio_recall = 0.4728, 2025-02-25 22:56:53,910 INFO ********************************************** Train results (accuracy and recall): 0.7357 0.4760.2025-02-25 22:56:53,916 INFO epoch: *******************************************72025-02-25 22:56:55,347 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6880 0.4089 0.5130.********************************************** Best results (accuracy and recall F1-score): 0.6880 0.4089 0.5130.2025-02-25 22:56:55,353 INFO -----------------------------2025-02-25 22:56:55,353 INFO evaluate loss:1.55942684402813232025-02-25 22:56:55,353 INFO -----------------------------2025-02-25 22:56:55,716 INFO Epoch 8, Batch 0, lr = 0.00040, audio_event_loss = 1.0485, BCELoss = 0.1405, ExpLogLoss = 0.9079, audio_precision = 0.7309, audio_recall = 0.5045, 2025-02-25 22:56:56,528 INFO Epoch 8, Batch 20, lr = 0.00040, audio_event_loss = 1.0300, BCELoss = 0.1390, ExpLogLoss = 0.8909, audio_precision = 0.7500, audio_recall = 0.4576, 2025-02-25 22:56:56,646 INFO ********************************************** Train results (accuracy and recall): 0.7494 0.4584.2025-02-25 22:56:56,653 INFO epoch: *******************************************82025-02-25 22:56:58,109 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.8146 0.4085 0.5441.********************************************** Best results (accuracy and recall F1-score): 0.8146 0.4085 0.5441.2025-02-25 22:56:58,116 INFO -----------------------------2025-02-25 22:56:58,116 INFO evaluate loss:1.53489073094955322025-02-25 22:56:58,116 INFO -----------------------------2025-02-25 22:56:58,483 INFO Epoch 9, Batch 0, lr = 0.00040, audio_event_loss = 1.0463, BCELoss = 0.1406, ExpLogLoss = 0.9057, audio_precision = 0.7592, audio_recall = 0.4353, 2025-02-25 22:56:59,295 INFO Epoch 9, Batch 20, lr = 0.00040, audio_event_loss = 1.0075, BCELoss = 0.1370, ExpLogLoss = 0.8705, audio_precision = 0.7657, audio_recall = 0.4732, 2025-02-25 22:56:59,412 INFO ********************************************** Train results (accuracy and recall): 0.7711 0.4729.2025-02-25 22:56:59,419 INFO epoch: *******************************************92025-02-25 22:57:00,854 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:00,860 INFO -----------------------------2025-02-25 22:57:00,860 INFO evaluate loss:1.52200084488665042025-02-25 22:57:00,860 INFO -----------------------------2025-02-25 22:57:01,251 INFO Epoch 10, Batch 0, lr = 0.00040, audio_event_loss = 1.0620, BCELoss = 0.1397, ExpLogLoss = 0.9223, audio_precision = 0.8030, audio_recall = 0.4597, 2025-02-25 22:57:02,043 INFO Epoch 10, Batch 20, lr = 0.00040, audio_event_loss = 1.0027, BCELoss = 0.1371, ExpLogLoss = 0.8656, audio_precision = 0.7514, audio_recall = 0.4707, 2025-02-25 22:57:02,160 INFO ********************************************** Train results (accuracy and recall): 0.7554 0.4725.2025-02-25 22:57:02,170 INFO epoch: *******************************************102025-02-25 22:57:03,612 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6992 0.4422 0.5418.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:03,618 INFO -----------------------------2025-02-25 22:57:03,618 INFO evaluate loss:1.55123134049691452025-02-25 22:57:03,618 INFO -----------------------------2025-02-25 22:57:03,984 INFO Epoch 11, Batch 0, lr = 0.00040, audio_event_loss = 0.9654, BCELoss = 0.1336, ExpLogLoss = 0.8318, audio_precision = 0.7575, audio_recall = 0.4942, 2025-02-25 22:57:04,786 INFO Epoch 11, Batch 20, lr = 0.00040, audio_event_loss = 0.9876, BCELoss = 0.1356, ExpLogLoss = 0.8520, audio_precision = 0.7643, audio_recall = 0.4895, 2025-02-25 22:57:04,901 INFO ********************************************** Train results (accuracy and recall): 0.7602 0.4851.2025-02-25 22:57:04,908 INFO epoch: *******************************************112025-02-25 22:57:06,360 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6915 0.3895 0.4983.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:06,366 INFO -----------------------------2025-02-25 22:57:06,366 INFO evaluate loss:1.5865624405189882025-02-25 22:57:06,366 INFO -----------------------------2025-02-25 22:57:06,731 INFO Epoch 12, Batch 0, lr = 0.00040, audio_event_loss = 0.9863, BCELoss = 0.1347, ExpLogLoss = 0.8516, audio_precision = 0.7420, audio_recall = 0.4627, 2025-02-25 22:57:07,556 INFO Epoch 12, Batch 20, lr = 0.00040, audio_event_loss = 0.9691, BCELoss = 0.1340, ExpLogLoss = 0.8351, audio_precision = 0.7789, audio_recall = 0.4778, 2025-02-25 22:57:07,673 INFO ********************************************** Train results (accuracy and recall): 0.7753 0.4741.2025-02-25 22:57:07,679 INFO epoch: *******************************************122025-02-25 22:57:09,113 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6852 0.4093 0.5125.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:09,122 INFO -----------------------------2025-02-25 22:57:09,122 INFO evaluate loss:1.62258205193317222025-02-25 22:57:09,122 INFO -----------------------------2025-02-25 22:57:09,468 INFO Epoch 13, Batch 0, lr = 0.00040, audio_event_loss = 1.0953, BCELoss = 0.1473, ExpLogLoss = 0.9480, audio_precision = 0.7278, audio_recall = 0.3978, 2025-02-25 22:57:10,305 INFO Epoch 13, Batch 20, lr = 0.00040, audio_event_loss = 0.9583, BCELoss = 0.1332, ExpLogLoss = 0.8251, audio_precision = 0.7755, audio_recall = 0.4742, 2025-02-25 22:57:10,421 INFO ********************************************** Train results (accuracy and recall): 0.7705 0.4685.2025-02-25 22:57:10,428 INFO epoch: *******************************************132025-02-25 22:57:11,872 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6843 0.3857 0.4934.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:11,879 INFO -----------------------------2025-02-25 22:57:11,879 INFO evaluate loss:1.59607566351722042025-02-25 22:57:11,879 INFO -----------------------------2025-02-25 22:57:12,250 INFO Epoch 14, Batch 0, lr = 0.00020, audio_event_loss = 0.9056, BCELoss = 0.1275, ExpLogLoss = 0.7781, audio_precision = 0.7760, audio_recall = 0.5295, 2025-02-25 22:57:13,065 INFO Epoch 14, Batch 20, lr = 0.00020, audio_event_loss = 0.9612, BCELoss = 0.1339, ExpLogLoss = 0.8273, audio_precision = 0.7707, audio_recall = 0.4700, 2025-02-25 22:57:13,178 INFO ********************************************** Train results (accuracy and recall): 0.7689 0.4726.2025-02-25 22:57:13,185 INFO epoch: *******************************************142025-02-25 22:57:14,623 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6855 0.3951 0.5013.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:14,631 INFO -----------------------------2025-02-25 22:57:14,631 INFO evaluate loss:1.58279073306339042025-02-25 22:57:14,631 INFO -----------------------------2025-02-25 22:57:15,025 INFO Epoch 15, Batch 0, lr = 0.00020, audio_event_loss = 1.0387, BCELoss = 0.1408, ExpLogLoss = 0.8979, audio_precision = 0.7316, audio_recall = 0.4204, 2025-02-25 22:57:15,842 INFO Epoch 15, Batch 20, lr = 0.00020, audio_event_loss = 0.9403, BCELoss = 0.1323, ExpLogLoss = 0.8080, audio_precision = 0.7780, audio_recall = 0.4668, 2025-02-25 22:57:15,959 INFO ********************************************** Train results (accuracy and recall): 0.7752 0.4638.2025-02-25 22:57:15,965 INFO epoch: *******************************************152025-02-25 22:57:17,421 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7054 0.4173 0.5244.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:17,428 INFO -----------------------------2025-02-25 22:57:17,428 INFO evaluate loss:1.56540096496491742025-02-25 22:57:17,428 INFO -----------------------------2025-02-25 22:57:17,792 INFO Epoch 16, Batch 0, lr = 0.00020, audio_event_loss = 1.0116, BCELoss = 0.1396, ExpLogLoss = 0.8720, audio_precision = 0.7785, audio_recall = 0.4302, 2025-02-25 22:57:18,609 INFO Epoch 16, Batch 20, lr = 0.00020, audio_event_loss = 0.9369, BCELoss = 0.1317, ExpLogLoss = 0.8053, audio_precision = 0.7754, audio_recall = 0.4816, 2025-02-25 22:57:18,726 INFO ********************************************** Train results (accuracy and recall): 0.7768 0.4779.2025-02-25 22:57:18,735 INFO epoch: *******************************************162025-02-25 22:57:20,181 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6969 0.4092 0.5156.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:20,188 INFO -----------------------------2025-02-25 22:57:20,188 INFO evaluate loss:1.5535128274786952025-02-25 22:57:20,188 INFO -----------------------------2025-02-25 22:57:20,549 INFO Epoch 17, Batch 0, lr = 0.00020, audio_event_loss = 0.9441, BCELoss = 0.1327, ExpLogLoss = 0.8114, audio_precision = 0.7883, audio_recall = 0.5006, 2025-02-25 22:57:21,380 INFO Epoch 17, Batch 20, lr = 0.00020, audio_event_loss = 0.9477, BCELoss = 0.1333, ExpLogLoss = 0.8145, audio_precision = 0.7803, audio_recall = 0.4676, 2025-02-25 22:57:21,498 INFO ********************************************** Train results (accuracy and recall): 0.7781 0.4707.2025-02-25 22:57:21,506 INFO epoch: *******************************************172025-02-25 22:57:22,956 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6819 0.4034 0.5069.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:22,963 INFO -----------------------------2025-02-25 22:57:22,963 INFO evaluate loss:1.57362623158570132025-02-25 22:57:22,963 INFO -----------------------------2025-02-25 22:57:23,313 INFO Epoch 18, Batch 0, lr = 0.00020, audio_event_loss = 0.9405, BCELoss = 0.1339, ExpLogLoss = 0.8066, audio_precision = 0.7535, audio_recall = 0.4697, 2025-02-25 22:57:24,139 INFO Epoch 18, Batch 20, lr = 0.00020, audio_event_loss = 0.9411, BCELoss = 0.1331, ExpLogLoss = 0.8080, audio_precision = 0.7775, audio_recall = 0.4654, 2025-02-25 22:57:24,259 INFO ********************************************** Train results (accuracy and recall): 0.7763 0.4684.2025-02-25 22:57:24,266 INFO epoch: *******************************************182025-02-25 22:57:25,715 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6854 0.4002 0.5053.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:25,723 INFO -----------------------------2025-02-25 22:57:25,723 INFO evaluate loss:1.55916403218264682025-02-25 22:57:25,723 INFO -----------------------------2025-02-25 22:57:26,090 INFO Epoch 19, Batch 0, lr = 0.00020, audio_event_loss = 0.8089, BCELoss = 0.1193, ExpLogLoss = 0.6896, audio_precision = 0.7984, audio_recall = 0.5308, 2025-02-25 22:57:26,914 INFO Epoch 19, Batch 20, lr = 0.00020, audio_event_loss = 0.9368, BCELoss = 0.1324, ExpLogLoss = 0.8044, audio_precision = 0.7639, audio_recall = 0.4768, 2025-02-25 22:57:27,030 INFO ********************************************** Train results (accuracy and recall): 0.7689 0.4803.2025-02-25 22:57:27,036 INFO epoch: *******************************************192025-02-25 22:57:28,491 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7041 0.3986 0.5091.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:28,498 INFO -----------------------------2025-02-25 22:57:28,498 INFO evaluate loss:1.54587497788707972025-02-25 22:57:28,498 INFO -----------------------------2025-02-25 22:57:28,859 INFO Epoch 20, Batch 0, lr = 0.00020, audio_event_loss = 0.9199, BCELoss = 0.1305, ExpLogLoss = 0.7894, audio_precision = 0.7998, audio_recall = 0.4828, 2025-02-25 22:57:29,685 INFO Epoch 20, Batch 20, lr = 0.00020, audio_event_loss = 0.9317, BCELoss = 0.1320, ExpLogLoss = 0.7997, audio_precision = 0.7733, audio_recall = 0.4713, 2025-02-25 22:57:29,800 INFO ********************************************** Train results (accuracy and recall): 0.7766 0.4685.2025-02-25 22:57:29,806 INFO epoch: *******************************************202025-02-25 22:57:31,247 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7016 0.3989 0.5086.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:31,254 INFO -----------------------------2025-02-25 22:57:31,254 INFO evaluate loss:1.56048957368198732025-02-25 22:57:31,254 INFO -----------------------------2025-02-25 22:57:31,613 INFO Epoch 21, Batch 0, lr = 0.00020, audio_event_loss = 0.9140, BCELoss = 0.1323, ExpLogLoss = 0.7818, audio_precision = 0.8025, audio_recall = 0.4508, 2025-02-25 22:57:32,423 INFO Epoch 21, Batch 20, lr = 0.00020, audio_event_loss = 0.9305, BCELoss = 0.1326, ExpLogLoss = 0.7979, audio_precision = 0.7849, audio_recall = 0.4576, 2025-02-25 22:57:32,536 INFO ********************************************** Train results (accuracy and recall): 0.7840 0.4640.2025-02-25 22:57:32,543 INFO epoch: *******************************************212025-02-25 22:57:33,998 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6845 0.3976 0.5030.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:34,005 INFO -----------------------------2025-02-25 22:57:34,005 INFO evaluate loss:1.57705492481068952025-02-25 22:57:34,005 INFO -----------------------------2025-02-25 22:57:34,377 INFO Epoch 22, Batch 0, lr = 0.00020, audio_event_loss = 0.9288, BCELoss = 0.1331, ExpLogLoss = 0.7957, audio_precision = 0.7269, audio_recall = 0.4808, 2025-02-25 22:57:35,189 INFO Epoch 22, Batch 20, lr = 0.00020, audio_event_loss = 0.9223, BCELoss = 0.1319, ExpLogLoss = 0.7904, audio_precision = 0.7745, audio_recall = 0.4691, 2025-02-25 22:57:35,305 INFO ********************************************** Train results (accuracy and recall): 0.7754 0.4753.2025-02-25 22:57:35,314 INFO epoch: *******************************************222025-02-25 22:57:36,762 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6947 0.4009 0.5084.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:36,769 INFO -----------------------------2025-02-25 22:57:36,769 INFO evaluate loss:1.53688310018627042025-02-25 22:57:36,769 INFO -----------------------------2025-02-25 22:57:37,137 INFO Epoch 23, Batch 0, lr = 0.00020, audio_event_loss = 0.8507, BCELoss = 0.1231, ExpLogLoss = 0.7276, audio_precision = 0.7929, audio_recall = 0.5286, 2025-02-25 22:57:37,952 INFO Epoch 23, Batch 20, lr = 0.00020, audio_event_loss = 0.9074, BCELoss = 0.1303, ExpLogLoss = 0.7771, audio_precision = 0.7760, audio_recall = 0.4806, 2025-02-25 22:57:38,070 INFO ********************************************** Train results (accuracy and recall): 0.7727 0.4760.2025-02-25 22:57:38,078 INFO epoch: *******************************************232025-02-25 22:57:39,538 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6965 0.4005 0.5086.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:39,545 INFO -----------------------------2025-02-25 22:57:39,545 INFO evaluate loss:1.53058165282658392025-02-25 22:57:39,545 INFO -----------------------------2025-02-25 22:57:39,920 INFO Epoch 24, Batch 0, lr = 0.00010, audio_event_loss = 0.9591, BCELoss = 0.1365, ExpLogLoss = 0.8226, audio_precision = 0.7845, audio_recall = 0.4454, 2025-02-25 22:57:40,741 INFO Epoch 24, Batch 20, lr = 0.00010, audio_event_loss = 0.9048, BCELoss = 0.1303, ExpLogLoss = 0.7745, audio_precision = 0.7875, audio_recall = 0.4704, 2025-02-25 22:57:40,857 INFO ********************************************** Train results (accuracy and recall): 0.7832 0.4717.2025-02-25 22:57:40,865 INFO epoch: *******************************************242025-02-25 22:57:42,335 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6887 0.4088 0.5131.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:42,342 INFO -----------------------------2025-02-25 22:57:42,342 INFO evaluate loss:1.54689252204600522025-02-25 22:57:42,342 INFO -----------------------------2025-02-25 22:57:42,710 INFO Epoch 25, Batch 0, lr = 0.00010, audio_event_loss = 0.8813, BCELoss = 0.1272, ExpLogLoss = 0.7541, audio_precision = 0.7285, audio_recall = 0.5011, 2025-02-25 22:57:43,517 INFO Epoch 25, Batch 20, lr = 0.00010, audio_event_loss = 0.9068, BCELoss = 0.1308, ExpLogLoss = 0.7760, audio_precision = 0.7705, audio_recall = 0.4741, 2025-02-25 22:57:43,631 INFO ********************************************** Train results (accuracy and recall): 0.7712 0.4720.2025-02-25 22:57:43,638 INFO epoch: *******************************************252025-02-25 22:57:45,112 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7041 0.4039 0.5134.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:45,118 INFO -----------------------------2025-02-25 22:57:45,118 INFO evaluate loss:1.54247722325676762025-02-25 22:57:45,118 INFO -----------------------------2025-02-25 22:57:45,495 INFO Epoch 26, Batch 0, lr = 0.00010, audio_event_loss = 0.9656, BCELoss = 0.1361, ExpLogLoss = 0.8294, audio_precision = 0.7932, audio_recall = 0.4332, 2025-02-25 22:57:46,310 INFO Epoch 26, Batch 20, lr = 0.00010, audio_event_loss = 0.9035, BCELoss = 0.1303, ExpLogLoss = 0.7731, audio_precision = 0.7710, audio_recall = 0.4797, 2025-02-25 22:57:46,425 INFO ********************************************** Train results (accuracy and recall): 0.7741 0.4794.2025-02-25 22:57:46,433 INFO epoch: *******************************************262025-02-25 22:57:47,886 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7051 0.4039 0.5136.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:47,895 INFO -----------------------------2025-02-25 22:57:47,895 INFO evaluate loss:1.52798828812604962025-02-25 22:57:47,896 INFO -----------------------------2025-02-25 22:57:48,250 INFO Epoch 27, Batch 0, lr = 0.00010, audio_event_loss = 0.9033, BCELoss = 0.1290, ExpLogLoss = 0.7743, audio_precision = 0.8047, audio_recall = 0.5107, 2025-02-25 22:57:49,071 INFO Epoch 27, Batch 20, lr = 0.00010, audio_event_loss = 0.8975, BCELoss = 0.1302, ExpLogLoss = 0.7673, audio_precision = 0.7836, audio_recall = 0.4629, 2025-02-25 22:57:49,187 INFO ********************************************** Train results (accuracy and recall): 0.7814 0.4644.2025-02-25 22:57:49,194 INFO epoch: *******************************************272025-02-25 22:57:50,635 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7035 0.4039 0.5132.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:50,641 INFO -----------------------------2025-02-25 22:57:50,641 INFO evaluate loss:1.53817536236696432025-02-25 22:57:50,641 INFO -----------------------------2025-02-25 22:57:51,006 INFO Epoch 28, Batch 0, lr = 0.00010, audio_event_loss = 0.9754, BCELoss = 0.1377, ExpLogLoss = 0.8377, audio_precision = 0.7457, audio_recall = 0.4288, 2025-02-25 22:57:51,833 INFO Epoch 28, Batch 20, lr = 0.00010, audio_event_loss = 0.8988, BCELoss = 0.1303, ExpLogLoss = 0.7685, audio_precision = 0.7784, audio_recall = 0.4741, 2025-02-25 22:57:51,950 INFO ********************************************** Train results (accuracy and recall): 0.7760 0.4680.2025-02-25 22:57:51,957 INFO epoch: *******************************************282025-02-25 22:57:53,425 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6969 0.4058 0.5130.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:53,431 INFO -----------------------------2025-02-25 22:57:53,431 INFO evaluate loss:1.53174519133445462025-02-25 22:57:53,431 INFO -----------------------------2025-02-25 22:57:53,797 INFO Epoch 29, Batch 0, lr = 0.00010, audio_event_loss = 0.9486, BCELoss = 0.1357, ExpLogLoss = 0.8129, audio_precision = 0.7729, audio_recall = 0.4165, 2025-02-25 22:57:54,627 INFO Epoch 29, Batch 20, lr = 0.00010, audio_event_loss = 0.8922, BCELoss = 0.1299, ExpLogLoss = 0.7623, audio_precision = 0.7765, audio_recall = 0.4729, 2025-02-25 22:57:54,746 INFO ********************************************** Train results (accuracy and recall): 0.7767 0.4737.2025-02-25 22:57:54,752 INFO epoch: *******************************************292025-02-25 22:57:56,196 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6932 0.3993 0.5067.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:56,202 INFO -----------------------------2025-02-25 22:57:56,202 INFO evaluate loss:1.53160064526663622025-02-25 22:57:56,202 INFO -----------------------------2025-02-25 22:57:56,551 INFO Epoch 30, Batch 0, lr = 0.00010, audio_event_loss = 0.8984, BCELoss = 0.1292, ExpLogLoss = 0.7692, audio_precision = 0.7653, audio_recall = 0.5213, 2025-02-25 22:57:57,370 INFO Epoch 30, Batch 20, lr = 0.00010, audio_event_loss = 0.8898, BCELoss = 0.1294, ExpLogLoss = 0.7604, audio_precision = 0.7807, audio_recall = 0.4721, 2025-02-25 22:57:57,488 INFO ********************************************** Train results (accuracy and recall): 0.7772 0.4707.2025-02-25 22:57:57,495 INFO epoch: *******************************************302025-02-25 22:57:58,933 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6995 0.4055 0.5134.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:57:58,941 INFO -----------------------------2025-02-25 22:57:58,941 INFO evaluate loss:1.52942490782278062025-02-25 22:57:58,941 INFO -----------------------------2025-02-25 22:57:59,296 INFO Epoch 31, Batch 0, lr = 0.00010, audio_event_loss = 0.8901, BCELoss = 0.1305, ExpLogLoss = 0.7596, audio_precision = 0.7612, audio_recall = 0.4719, 2025-02-25 22:58:00,113 INFO Epoch 31, Batch 20, lr = 0.00010, audio_event_loss = 0.9095, BCELoss = 0.1318, ExpLogLoss = 0.7777, audio_precision = 0.7746, audio_recall = 0.4583, 2025-02-25 22:58:00,231 INFO ********************************************** Train results (accuracy and recall): 0.7758 0.4636.2025-02-25 22:58:00,237 INFO epoch: *******************************************312025-02-25 22:58:01,701 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6939 0.3993 0.5069.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:01,708 INFO -----------------------------2025-02-25 22:58:01,708 INFO evaluate loss:1.52560917476796122025-02-25 22:58:01,708 INFO -----------------------------2025-02-25 22:58:02,074 INFO Epoch 32, Batch 0, lr = 0.00010, audio_event_loss = 0.9049, BCELoss = 0.1309, ExpLogLoss = 0.7740, audio_precision = 0.7748, audio_recall = 0.4656, 2025-02-25 22:58:02,911 INFO Epoch 32, Batch 20, lr = 0.00010, audio_event_loss = 0.8946, BCELoss = 0.1302, ExpLogLoss = 0.7644, audio_precision = 0.7806, audio_recall = 0.4716, 2025-02-25 22:58:03,030 INFO ********************************************** Train results (accuracy and recall): 0.7841 0.4717.2025-02-25 22:58:03,037 INFO epoch: *******************************************322025-02-25 22:58:04,498 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6886 0.3989 0.5052.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:04,507 INFO -----------------------------2025-02-25 22:58:04,508 INFO evaluate loss:1.52337438819176672025-02-25 22:58:04,508 INFO -----------------------------2025-02-25 22:58:04,855 INFO Epoch 33, Batch 0, lr = 0.00010, audio_event_loss = 0.8795, BCELoss = 0.1292, ExpLogLoss = 0.7503, audio_precision = 0.8401, audio_recall = 0.4798, 2025-02-25 22:58:05,691 INFO Epoch 33, Batch 20, lr = 0.00010, audio_event_loss = 0.8945, BCELoss = 0.1305, ExpLogLoss = 0.7640, audio_precision = 0.7745, audio_recall = 0.4762, 2025-02-25 22:58:05,811 INFO ********************************************** Train results (accuracy and recall): 0.7765 0.4776.2025-02-25 22:58:05,819 INFO epoch: *******************************************332025-02-25 22:58:07,258 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7055 0.4138 0.5216.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:07,265 INFO -----------------------------2025-02-25 22:58:07,265 INFO evaluate loss:1.52186664387596872025-02-25 22:58:07,265 INFO -----------------------------2025-02-25 22:58:07,625 INFO Epoch 34, Batch 0, lr = 0.00005, audio_event_loss = 0.8544, BCELoss = 0.1257, ExpLogLoss = 0.7287, audio_precision = 0.7844, audio_recall = 0.5068, 2025-02-25 22:58:08,446 INFO Epoch 34, Batch 20, lr = 0.00005, audio_event_loss = 0.8958, BCELoss = 0.1303, ExpLogLoss = 0.7655, audio_precision = 0.7710, audio_recall = 0.4768, 2025-02-25 22:58:08,561 INFO ********************************************** Train results (accuracy and recall): 0.7737 0.4776.2025-02-25 22:58:08,571 INFO epoch: *******************************************342025-02-25 22:58:10,036 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7022 0.4043 0.5132.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:10,043 INFO -----------------------------2025-02-25 22:58:10,043 INFO evaluate loss:1.52582085739519442025-02-25 22:58:10,043 INFO -----------------------------2025-02-25 22:58:10,381 INFO Epoch 35, Batch 0, lr = 0.00005, audio_event_loss = 0.9069, BCELoss = 0.1326, ExpLogLoss = 0.7743, audio_precision = 0.7734, audio_recall = 0.4697, 2025-02-25 22:58:11,211 INFO Epoch 35, Batch 20, lr = 0.00005, audio_event_loss = 0.8935, BCELoss = 0.1304, ExpLogLoss = 0.7631, audio_precision = 0.7756, audio_recall = 0.4713, 2025-02-25 22:58:11,328 INFO ********************************************** Train results (accuracy and recall): 0.7743 0.4723.2025-02-25 22:58:11,334 INFO epoch: *******************************************352025-02-25 22:58:12,783 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6953 0.4138 0.5188.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:12,789 INFO -----------------------------2025-02-25 22:58:12,789 INFO evaluate loss:1.51546478997732572025-02-25 22:58:12,789 INFO -----------------------------2025-02-25 22:58:13,157 INFO Epoch 36, Batch 0, lr = 0.00005, audio_event_loss = 0.8963, BCELoss = 0.1302, ExpLogLoss = 0.7661, audio_precision = 0.7858, audio_recall = 0.4914, 2025-02-25 22:58:13,978 INFO Epoch 36, Batch 20, lr = 0.00005, audio_event_loss = 0.8868, BCELoss = 0.1299, ExpLogLoss = 0.7570, audio_precision = 0.7759, audio_recall = 0.4625, 2025-02-25 22:58:14,095 INFO ********************************************** Train results (accuracy and recall): 0.7770 0.4685.2025-02-25 22:58:14,103 INFO epoch: *******************************************362025-02-25 22:58:15,559 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6981 0.4039 0.5118.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:15,566 INFO -----------------------------2025-02-25 22:58:15,566 INFO evaluate loss:1.51697530383790122025-02-25 22:58:15,566 INFO -----------------------------2025-02-25 22:58:15,935 INFO Epoch 37, Batch 0, lr = 0.00005, audio_event_loss = 0.8908, BCELoss = 0.1297, ExpLogLoss = 0.7611, audio_precision = 0.7643, audio_recall = 0.4827, 2025-02-25 22:58:16,738 INFO Epoch 37, Batch 20, lr = 0.00005, audio_event_loss = 0.8784, BCELoss = 0.1292, ExpLogLoss = 0.7492, audio_precision = 0.7794, audio_recall = 0.4738, 2025-02-25 22:58:16,854 INFO ********************************************** Train results (accuracy and recall): 0.7785 0.4733.2025-02-25 22:58:16,859 INFO epoch: *******************************************372025-02-25 22:58:18,298 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6897 0.4058 0.5110.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:18,305 INFO -----------------------------2025-02-25 22:58:18,305 INFO evaluate loss:1.51299884200676532025-02-25 22:58:18,305 INFO -----------------------------2025-02-25 22:58:18,684 INFO Epoch 38, Batch 0, lr = 0.00005, audio_event_loss = 0.9484, BCELoss = 0.1354, ExpLogLoss = 0.8129, audio_precision = 0.7820, audio_recall = 0.4701, 2025-02-25 22:58:19,505 INFO Epoch 38, Batch 20, lr = 0.00005, audio_event_loss = 0.8830, BCELoss = 0.1292, ExpLogLoss = 0.7538, audio_precision = 0.7787, audio_recall = 0.4756, 2025-02-25 22:58:19,626 INFO ********************************************** Train results (accuracy and recall): 0.7774 0.4720.2025-02-25 22:58:19,632 INFO epoch: *******************************************382025-02-25 22:58:21,086 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6915 0.4055 0.5112.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:21,095 INFO -----------------------------2025-02-25 22:58:21,095 INFO evaluate loss:1.51732484283961092025-02-25 22:58:21,095 INFO -----------------------------2025-02-25 22:58:21,444 INFO Epoch 39, Batch 0, lr = 0.00005, audio_event_loss = 0.8993, BCELoss = 0.1323, ExpLogLoss = 0.7670, audio_precision = 0.7380, audio_recall = 0.4445, 2025-02-25 22:58:22,292 INFO Epoch 39, Batch 20, lr = 0.00005, audio_event_loss = 0.8901, BCELoss = 0.1302, ExpLogLoss = 0.7599, audio_precision = 0.7712, audio_recall = 0.4819, 2025-02-25 22:58:22,409 INFO ********************************************** Train results (accuracy and recall): 0.7676 0.4803.2025-02-25 22:58:22,416 INFO epoch: *******************************************392025-02-25 22:58:23,851 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7044 0.4043 0.5137.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:23,858 INFO -----------------------------2025-02-25 22:58:23,858 INFO evaluate loss:1.51660700988379522025-02-25 22:58:23,858 INFO -----------------------------2025-02-25 22:58:24,206 INFO Epoch 40, Batch 0, lr = 0.00005, audio_event_loss = 0.8384, BCELoss = 0.1252, ExpLogLoss = 0.7132, audio_precision = 0.7921, audio_recall = 0.5139, 2025-02-25 22:58:25,031 INFO Epoch 40, Batch 20, lr = 0.00005, audio_event_loss = 0.8738, BCELoss = 0.1288, ExpLogLoss = 0.7450, audio_precision = 0.7761, audio_recall = 0.4732, 2025-02-25 22:58:25,147 INFO ********************************************** Train results (accuracy and recall): 0.7755 0.4722.2025-02-25 22:58:25,154 INFO epoch: *******************************************402025-02-25 22:58:26,594 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.7063 0.4039 0.5140.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:26,600 INFO -----------------------------2025-02-25 22:58:26,600 INFO evaluate loss:1.5204035636677442025-02-25 22:58:26,600 INFO -----------------------------2025-02-25 22:58:26,943 INFO Epoch 41, Batch 0, lr = 0.00005, audio_event_loss = 0.9419, BCELoss = 0.1351, ExpLogLoss = 0.8068, audio_precision = 0.7845, audio_recall = 0.4654, 2025-02-25 22:58:27,778 INFO Epoch 41, Batch 20, lr = 0.00005, audio_event_loss = 0.8969, BCELoss = 0.1311, ExpLogLoss = 0.7659, audio_precision = 0.7738, audio_recall = 0.4589, 2025-02-25 22:58:27,899 INFO ********************************************** Train results (accuracy and recall): 0.7782 0.4666.2025-02-25 22:58:27,906 INFO epoch: *******************************************412025-02-25 22:58:29,358 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6907 0.4058 0.5113.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:29,366 INFO -----------------------------2025-02-25 22:58:29,366 INFO evaluate loss:1.51427260307405052025-02-25 22:58:29,366 INFO -----------------------------2025-02-25 22:58:29,707 INFO Epoch 42, Batch 0, lr = 0.00005, audio_event_loss = 0.8779, BCELoss = 0.1288, ExpLogLoss = 0.7491, audio_precision = 0.7466, audio_recall = 0.4851, 2025-02-25 22:58:30,542 INFO Epoch 42, Batch 20, lr = 0.00005, audio_event_loss = 0.8821, BCELoss = 0.1295, ExpLogLoss = 0.7526, audio_precision = 0.7818, audio_recall = 0.4786, 2025-02-25 22:58:30,659 INFO ********************************************** Train results (accuracy and recall): 0.7786 0.4767.2025-02-25 22:58:30,666 INFO epoch: *******************************************422025-02-25 22:58:32,127 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6951 0.4043 0.5113.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:32,134 INFO -----------------------------2025-02-25 22:58:32,134 INFO evaluate loss:1.51459952746143682025-02-25 22:58:32,134 INFO -----------------------------2025-02-25 22:58:32,508 INFO Epoch 43, Batch 0, lr = 0.00005, audio_event_loss = 0.7950, BCELoss = 0.1218, ExpLogLoss = 0.6732, audio_precision = 0.7796, audio_recall = 0.5005, 2025-02-25 22:58:33,321 INFO Epoch 43, Batch 20, lr = 0.00005, audio_event_loss = 0.8819, BCELoss = 0.1297, ExpLogLoss = 0.7522, audio_precision = 0.7749, audio_recall = 0.4722, 2025-02-25 22:58:33,434 INFO ********************************************** Train results (accuracy and recall): 0.7750 0.4712.2025-02-25 22:58:33,443 INFO epoch: *******************************************432025-02-25 22:58:34,888 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6907 0.4058 0.5113.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:34,894 INFO -----------------------------2025-02-25 22:58:34,894 INFO evaluate loss:1.51114376722631732025-02-25 22:58:34,894 INFO -----------------------------2025-02-25 22:58:35,250 INFO Epoch 44, Batch 0, lr = 0.00005, audio_event_loss = 0.9247, BCELoss = 0.1347, ExpLogLoss = 0.7900, audio_precision = 0.7682, audio_recall = 0.4395, 2025-02-25 22:58:36,073 INFO Epoch 44, Batch 20, lr = 0.00005, audio_event_loss = 0.8695, BCELoss = 0.1287, ExpLogLoss = 0.7408, audio_precision = 0.7768, audio_recall = 0.4739, 2025-02-25 22:58:36,191 INFO ********************************************** Train results (accuracy and recall): 0.7774 0.4699.2025-02-25 22:58:36,198 INFO epoch: *******************************************442025-02-25 22:58:37,647 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6962 0.4043 0.5115.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:37,654 INFO -----------------------------2025-02-25 22:58:37,654 INFO evaluate loss:1.50636890717526842025-02-25 22:58:37,654 INFO -----------------------------2025-02-25 22:58:38,018 INFO Epoch 45, Batch 0, lr = 0.00005, audio_event_loss = 0.9061, BCELoss = 0.1311, ExpLogLoss = 0.7750, audio_precision = 0.7790, audio_recall = 0.4421, 2025-02-25 22:58:38,846 INFO Epoch 45, Batch 20, lr = 0.00005, audio_event_loss = 0.8816, BCELoss = 0.1299, ExpLogLoss = 0.7517, audio_precision = 0.7801, audio_recall = 0.4685, 2025-02-25 22:58:38,959 INFO ********************************************** Train results (accuracy and recall): 0.7768 0.4683.2025-02-25 22:58:38,968 INFO epoch: *******************************************452025-02-25 22:58:40,415 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6907 0.4058 0.5113.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:40,421 INFO -----------------------------2025-02-25 22:58:40,421 INFO evaluate loss:1.50813660860307342025-02-25 22:58:40,421 INFO -----------------------------2025-02-25 22:58:40,794 INFO Epoch 46, Batch 0, lr = 0.00005, audio_event_loss = 0.9150, BCELoss = 0.1328, ExpLogLoss = 0.7822, audio_precision = 0.7486, audio_recall = 0.4564, 2025-02-25 22:58:41,603 INFO Epoch 46, Batch 20, lr = 0.00005, audio_event_loss = 0.8691, BCELoss = 0.1288, ExpLogLoss = 0.7403, audio_precision = 0.7812, audio_recall = 0.4700, 2025-02-25 22:58:41,718 INFO ********************************************** Train results (accuracy and recall): 0.7786 0.4662.2025-02-25 22:58:41,725 INFO epoch: *******************************************462025-02-25 22:58:43,176 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6987 0.4058 0.5135.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:43,183 INFO -----------------------------2025-02-25 22:58:43,183 INFO evaluate loss:1.51404853986189862025-02-25 22:58:43,183 INFO -----------------------------2025-02-25 22:58:43,525 INFO Epoch 47, Batch 0, lr = 0.00005, audio_event_loss = 0.9770, BCELoss = 0.1386, ExpLogLoss = 0.8384, audio_precision = 0.8005, audio_recall = 0.4432, 2025-02-25 22:58:44,358 INFO Epoch 47, Batch 20, lr = 0.00005, audio_event_loss = 0.8689, BCELoss = 0.1287, ExpLogLoss = 0.7402, audio_precision = 0.7799, audio_recall = 0.4710, 2025-02-25 22:58:44,472 INFO ********************************************** Train results (accuracy and recall): 0.7782 0.4687.2025-02-25 22:58:44,480 INFO epoch: *******************************************472025-02-25 22:58:45,922 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6934 0.4055 0.5117.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:45,929 INFO -----------------------------2025-02-25 22:58:45,929 INFO evaluate loss:1.51024472706552372025-02-25 22:58:45,929 INFO -----------------------------2025-02-25 22:58:46,285 INFO Epoch 48, Batch 0, lr = 0.00005, audio_event_loss = 0.9010, BCELoss = 0.1327, ExpLogLoss = 0.7683, audio_precision = 0.7265, audio_recall = 0.4618, 2025-02-25 22:58:47,120 INFO Epoch 48, Batch 20, lr = 0.00005, audio_event_loss = 0.8755, BCELoss = 0.1295, ExpLogLoss = 0.7460, audio_precision = 0.7794, audio_recall = 0.4684, 2025-02-25 22:58:47,231 INFO ********************************************** Train results (accuracy and recall): 0.7784 0.4705.2025-02-25 22:58:47,240 INFO epoch: *******************************************482025-02-25 22:58:48,690 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6981 0.4039 0.5118.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:48,697 INFO -----------------------------2025-02-25 22:58:48,697 INFO evaluate loss:1.51177847357402382025-02-25 22:58:48,697 INFO -----------------------------2025-02-25 22:58:49,066 INFO Epoch 49, Batch 0, lr = 0.00005, audio_event_loss = 0.8301, BCELoss = 0.1244, ExpLogLoss = 0.7058, audio_precision = 0.8071, audio_recall = 0.5020, 2025-02-25 22:58:49,884 INFO Epoch 49, Batch 20, lr = 0.00005, audio_event_loss = 0.8874, BCELoss = 0.1306, ExpLogLoss = 0.7568, audio_precision = 0.7825, audio_recall = 0.4618, 2025-02-25 22:58:50,000 INFO ********************************************** Train results (accuracy and recall): 0.7882 0.4659.2025-02-25 22:58:50,007 INFO epoch: *******************************************492025-02-25 22:58:51,456 INFO ********************************************** Evaluation results (accuracy and recall F1-score): 0.6926 0.4055 0.5115.********************************************** Best results (accuracy and recall F1-score): 0.8118 0.4317 0.5636.2025-02-25 22:58:51,463 INFO -----------------------------2025-02-25 22:58:51,463 INFO evaluate loss:1.50518929758060962025-02-25 22:58:51,463 INFO -----------------------------

复制代码

准确率、召回率和F1值变化
loss值变化

UCF_VGGSOUND 测试（跨模态视频分割）

关于在UCF101数据集上进行实验的说明。

cd CMG/code/src./ucf_vggsound.sh

复制代码

修改数据地址
修改模型地址

(cmg0218) trimps@trimps-System-Product-Name:~/llm_model/shaohang/CMG/code/src$ ./ucf_vggsound.sh/home/trimps/anaconda3/envs/cmg0218/lib/python3.7/site-packages/gensim/similarities/__init__.py:15: UserWarning: The gensim.similarities.levenshtein submodule is disabled, because the optional Levenshtein package is unavailable. Install Levenhstein (e.g. `pip install python-Levenshtein`) to suppress this warning. warnings.warn(msg)Updating: n_epoch: 100 (default) ----> 30Updating: batch_size: 64 (default) ----> 80Updating: test_batch_size: 16 (default) ----> 64Updating: lr: 0.001 (default) ----> 0.0004Updating: snapshot_pref: ./exp/debug/123 (default) ----> ./Exps/ucf_vggsound/Updating: clip_gradient: 0.8 (default) ----> 0.5Updating: print_freq: 20 (default) ----> 12025-02-25 22:41:52,260 INFO Creating folder: ./Exps/ucf_vggsound/2025-02-25 22:41:52,261 INFO Runtime args{ "dataset_name": "vgga_ucfv", "n_epoch": 30, "batch_size": 80, "test_batch_size": 64, "lr": 0.0004, "gpu": "1", "snapshot_pref": "./Exps/ucf_vggsound/", "resume": "", "evaluate": false, "clip_gradient": 0.5, "loss_weights": 0.5, "start_epoch": 0, "model_save_path": "../../../checkpoint", "weight_decay": 0.0005, "print_freq": 1, "save_freq": null, "eval_freq": 1, "mi_lr": 0.001}2025-02-25 22:41:55,110 INFO Resume from number 3-th model.======> Result will be saved at: ./Exps/ucf_vggsound/2025-02-25 22:41:55,838 INFO Epoch 4, Batch 0, lr = 0.00040, train_event_loss = 2.8568, train_precision = 2.5000, 2025-02-25 22:41:55,903 INFO Epoch 4, Batch 10, lr = 0.00040, train_event_loss = 2.2586, train_precision = 47.5000, 2025-02-25 22:41:55,956 INFO Epoch 4, Batch 20, lr = 0.00040, train_event_loss = 1.5580, train_precision = 71.0000, 2025-02-25 22:41:56,011 INFO Epoch 4, Batch 30, lr = 0.00040, train_event_loss = 1.1186, train_precision = 74.3750, 2025-02-25 22:41:56,065 INFO Epoch 4, Batch 40, lr = 0.00040, train_event_loss = 0.8392, train_precision = 79.3750, 2025-02-25 22:41:56,117 INFO Epoch 4, Batch 50, lr = 0.00040, train_event_loss = 0.6768, train_precision = 82.3750, 2025-02-25 22:41:56,165 INFO Epoch 4, Batch 60, lr = 0.00040, train_event_loss = 0.6477, train_precision = 82.0865, 2025-02-25 22:41:56,189 INFO Train results (precision): 71.45232025-02-25 22:41:56,189 INFO epoch: *******************************************42025-02-25 22:41:58,103 INFO Eval results (precision and loss): 58.8118 1.63782025-02-25 22:41:58,103 INFO Best results (precision): 58.81182025-02-25 22:41:58,104 INFO -----------------------------2025-02-25 22:41:58,104 INFO evaluate loss:1.63781768746576972025-02-25 22:41:58,104 INFO -----------------------------2025-02-25 22:41:58,342 INFO Epoch 5, Batch 0, lr = 0.00040, val_event_loss = 1.5417, val_precision = 58.1019, train_event_loss = 0.0296, train_precision = 2.6852, 2025-02-25 22:41:58,404 INFO Epoch 5, Batch 10, lr = 0.00040, train_event_loss = 0.5767, train_precision = 82.6250, 2025-02-25 22:41:58,463 INFO Epoch 5, Batch 20, lr = 0.00040, train_event_loss = 0.5288, train_precision = 83.0000, 2025-02-25 22:41:58,521 INFO Epoch 5, Batch 30, lr = 0.00040, train_event_loss = 0.5346, train_precision = 82.3750, 2025-02-25 22:41:58,577 INFO Epoch 5, Batch 40, lr = 0.00040, train_event_loss = 0.5303, train_precision = 82.5000, 2025-02-25 22:41:58,629 INFO Epoch 5, Batch 50, lr = 0.00040, train_event_loss = 0.5015, train_precision = 83.6250, 2025-02-25 22:41:58,677 INFO Epoch 5, Batch 60, lr = 0.00040, train_event_loss = 0.4869, train_precision = 86.2308, 2025-02-25 22:41:58,709 INFO Train results (precision): 83.08752025-02-25 22:41:58,709 INFO epoch: *******************************************52025-02-25 22:42:00,701 INFO Eval results (precision and loss): 58.8617 1.59042025-02-25 22:42:00,701 INFO Best results (precision): 58.86172025-02-25 22:42:00,701 INFO -----------------------------2025-02-25 22:42:00,701 INFO evaluate loss:1.59036972749861222025-02-25 22:42:00,701 INFO -----------------------------2025-02-25 22:42:00,926 INFO Epoch 6, Batch 0, lr = 0.00040, val_event_loss = 1.4931, val_precision = 58.1481, train_event_loss = 0.0168, train_precision = 3.2407, 2025-02-25 22:42:00,992 INFO Epoch 6, Batch 10, lr = 0.00040, train_event_loss = 0.5284, train_precision = 82.6250, 2025-02-25 22:42:01,048 INFO Epoch 6, Batch 20, lr = 0.00040, train_event_loss = 0.4895, train_precision = 83.2500, 2025-02-25 22:42:01,103 INFO Epoch 6, Batch 30, lr = 0.00040, train_event_loss = 0.5068, train_precision = 83.8750, 2025-02-25 22:42:01,159 INFO Epoch 6, Batch 40, lr = 0.00040, train_event_loss = 0.5124, train_precision = 83.7500, 2025-02-25 22:42:01,212 INFO Epoch 6, Batch 50, lr = 0.00040, train_event_loss = 0.4686, train_precision = 84.8750, 2025-02-25 22:42:01,267 INFO Epoch 6, Batch 60, lr = 0.00040, train_event_loss = 0.4549, train_precision = 86.1058, 2025-02-25 22:42:01,302 INFO Train results (precision): 84.02242025-02-25 22:42:01,302 INFO epoch: *******************************************62025-02-25 22:42:03,130 INFO Eval results (precision and loss): 57.4638 1.63322025-02-25 22:42:03,131 INFO Best results (precision): 58.86172025-02-25 22:42:03,131 INFO -----------------------------2025-02-25 22:42:03,131 INFO evaluate loss:1.63318694199451772025-02-25 22:42:03,131 INFO -----------------------------2025-02-25 22:42:03,329 INFO Epoch 7, Batch 0, lr = 0.00040, val_event_loss = 1.5293, val_precision = 56.8519, train_event_loss = 0.0187, train_precision = 3.1019, 2025-02-25 22:42:03,385 INFO Epoch 7, Batch 10, lr = 0.00040, train_event_loss = 0.4992, train_precision = 83.7500, 2025-02-25 22:42:03,434 INFO Epoch 7, Batch 20, lr = 0.00040, train_event_loss = 0.5349, train_precision = 83.1250, 2025-02-25 22:42:03,485 INFO Epoch 7, Batch 30, lr = 0.00040, train_event_loss = 0.4720, train_precision = 84.8750, 2025-02-25 22:42:03,536 INFO Epoch 7, Batch 40, lr = 0.00040, train_event_loss = 0.4623, train_precision = 85.7500, 2025-02-25 22:42:03,589 INFO Epoch 7, Batch 50, lr = 0.00040, train_event_loss = 0.5191, train_precision = 82.8750, 2025-02-25 22:42:03,635 INFO Epoch 7, Batch 60, lr = 0.00040, train_event_loss = 0.5114, train_precision = 84.4615, 2025-02-25 22:42:03,668 INFO Train results (precision): 84.12632025-02-25 22:42:03,668 INFO epoch: *******************************************72025-02-25 22:42:05,501 INFO Eval results (precision and loss): 61.0584 1.62922025-02-25 22:42:05,501 INFO Best results (precision): 61.05842025-02-25 22:42:05,501 INFO -----------------------------2025-02-25 22:42:05,501 INFO evaluate loss:1.62917980588973892025-02-25 22:42:05,501 INFO -----------------------------2025-02-25 22:42:05,705 INFO Epoch 8, Batch 0, lr = 0.00040, val_event_loss = 1.5205, val_precision = 60.1852, train_event_loss = 0.0186, train_precision = 3.0093, 2025-02-25 22:42:05,769 INFO Epoch 8, Batch 10, lr = 0.00040, train_event_loss = 0.4748, train_precision = 84.6250, 2025-02-25 22:42:05,824 INFO Epoch 8, Batch 20, lr = 0.00040, train_event_loss = 0.4299, train_precision = 85.1250, 2025-02-25 22:42:05,878 INFO Epoch 8, Batch 30, lr = 0.00040, train_event_loss = 0.4942, train_precision = 84.0000, 2025-02-25 22:42:05,935 INFO Epoch 8, Batch 40, lr = 0.00040, train_event_loss = 0.4771, train_precision = 84.8750, 2025-02-25 22:42:05,991 INFO Epoch 8, Batch 50, lr = 0.00040, train_event_loss = 0.4828, train_precision = 84.3750, 2025-02-25 22:42:06,039 INFO Epoch 8, Batch 60, lr = 0.00040, train_event_loss = 0.5091, train_precision = 82.7981, 2025-02-25 22:42:06,071 INFO Train results (precision): 84.45882025-02-25 22:42:06,071 INFO epoch: *******************************************82025-02-25 22:42:07,890 INFO Eval results (precision and loss): 57.5637 1.62272025-02-25 22:42:07,890 INFO Best results (precision): 61.05842025-02-25 22:42:07,890 INFO -----------------------------2025-02-25 22:42:07,890 INFO evaluate loss:1.62273895261863822025-02-25 22:42:07,890 INFO -----------------------------2025-02-25 22:42:08,095 INFO Epoch 9, Batch 0, lr = 0.00040, val_event_loss = 1.5127, val_precision = 56.9444, train_event_loss = 0.0132, train_precision = 3.3333, 2025-02-25 22:42:08,157 INFO Epoch 9, Batch 10, lr = 0.00040, train_event_loss = 0.4713, train_precision = 84.7500, 2025-02-25 22:42:08,211 INFO Epoch 9, Batch 20, lr = 0.00040, train_event_loss = 0.4706, train_precision = 84.7500, 2025-02-25 22:42:08,271 INFO Epoch 9, Batch 30, lr = 0.00040, train_event_loss = 0.4865, train_precision = 84.5000, 2025-02-25 22:42:08,326 INFO Epoch 9, Batch 40, lr = 0.00040, train_event_loss = 0.4947, train_precision = 83.5000, 2025-02-25 22:42:08,377 INFO Epoch 9, Batch 50, lr = 0.00040, train_event_loss = 0.4283, train_precision = 85.6250, 2025-02-25 22:42:08,424 INFO Epoch 9, Batch 60, lr = 0.00040, train_event_loss = 0.4741, train_precision = 84.2115, 2025-02-25 22:42:08,458 INFO Train results (precision): 84.64582025-02-25 22:42:08,458 INFO epoch: *******************************************92025-02-25 22:42:10,293 INFO Eval results (precision and loss): 59.1613 1.64612025-02-25 22:42:10,293 INFO Best results (precision): 61.05842025-02-25 22:42:10,294 INFO -----------------------------2025-02-25 22:42:10,294 INFO evaluate loss:1.64605157781690232025-02-25 22:42:10,294 INFO -----------------------------2025-02-25 22:42:10,494 INFO Epoch 10, Batch 0, lr = 0.00040, val_event_loss = 1.5338, val_precision = 58.4259, train_event_loss = 0.0113, train_precision = 3.3333, 2025-02-25 22:42:10,551 INFO Epoch 10, Batch 10, lr = 0.00040, train_event_loss = 0.4640, train_precision = 85.3750, 2025-02-25 22:42:10,600 INFO Epoch 10, Batch 20, lr = 0.00040, train_event_loss = 0.4829, train_precision = 84.7500, 2025-02-25 22:42:10,655 INFO Epoch 10, Batch 30, lr = 0.00040, train_event_loss = 0.5066, train_precision = 82.2500, 2025-02-25 22:42:10,708 INFO Epoch 10, Batch 40, lr = 0.00040, train_event_loss = 0.4805, train_precision = 84.3750, 2025-02-25 22:42:10,761 INFO Epoch 10, Batch 50, lr = 0.00040, train_event_loss = 0.4717, train_precision = 84.6250, 2025-02-25 22:42:10,808 INFO Epoch 10, Batch 60, lr = 0.00040, train_event_loss = 0.4193, train_precision = 86.0000, 2025-02-25 22:42:10,843 INFO Train results (precision): 84.43802025-02-25 22:42:10,844 INFO epoch: *******************************************102025-02-25 22:42:12,659 INFO Eval results (precision and loss): 60.5592 1.64032025-02-25 22:42:12,660 INFO Best results (precision): 61.05842025-02-25 22:42:12,660 INFO -----------------------------2025-02-25 22:42:12,660 INFO evaluate loss:1.64033482668165952025-02-25 22:42:12,660 INFO -----------------------------2025-02-25 22:42:12,865 INFO Epoch 11, Batch 0, lr = 0.00040, val_event_loss = 1.5318, val_precision = 59.7222, train_event_loss = 0.0201, train_precision = 3.0556, 2025-02-25 22:42:12,920 INFO Epoch 11, Batch 10, lr = 0.00040, train_event_loss = 0.4799, train_precision = 84.2500, 2025-02-25 22:42:12,972 INFO Epoch 11, Batch 20, lr = 0.00040, train_event_loss = 0.4577, train_precision = 84.8750, 2025-02-25 22:42:13,023 INFO Epoch 11, Batch 30, lr = 0.00040, train_event_loss = 0.4240, train_precision = 85.8750, 2025-02-25 22:42:13,074 INFO Epoch 11, Batch 40, lr = 0.00040, train_event_loss = 0.4359, train_precision = 86.1250, 2025-02-25 22:42:13,124 INFO Epoch 11, Batch 50, lr = 0.00040, train_event_loss = 0.4748, train_precision = 85.0000, 2025-02-25 22:42:13,171 INFO Epoch 11, Batch 60, lr = 0.00040, train_event_loss = 0.4874, train_precision = 84.0673, 2025-02-25 22:42:13,207 INFO Train results (precision): 85.10282025-02-25 22:42:13,208 INFO epoch: *******************************************112025-02-25 22:42:15,008 INFO Eval results (precision and loss): 59.6106 1.68242025-02-25 22:42:15,008 INFO Best results (precision): 61.05842025-02-25 22:42:15,008 INFO -----------------------------2025-02-25 22:42:15,008 INFO evaluate loss:1.68243587807335332025-02-25 22:42:15,008 INFO -----------------------------2025-02-25 22:42:15,208 INFO Epoch 12, Batch 0, lr = 0.00040, val_event_loss = 1.5685, val_precision = 58.8426, train_event_loss = 0.0126, train_precision = 3.2407, 2025-02-25 22:42:15,263 INFO Epoch 12, Batch 10, lr = 0.00040, train_event_loss = 0.4434, train_precision = 84.8750, 2025-02-25 22:42:15,314 INFO Epoch 12, Batch 20, lr = 0.00040, train_event_loss = 0.4363, train_precision = 84.5000, 2025-02-25 22:42:15,364 INFO Epoch 12, Batch 30, lr = 0.00040, train_event_loss = 0.4706, train_precision = 84.2500, 2025-02-25 22:42:15,422 INFO Epoch 12, Batch 40, lr = 0.00040, train_event_loss = 0.5029, train_precision = 83.0000, 2025-02-25 22:42:15,481 INFO Epoch 12, Batch 50, lr = 0.00040, train_event_loss = 0.4268, train_precision = 86.3750, 2025-02-25 22:42:15,528 INFO Epoch 12, Batch 60, lr = 0.00040, train_event_loss = 0.5466, train_precision = 83.5865, 2025-02-25 22:42:15,563 INFO Train results (precision): 84.47952025-02-25 22:42:15,563 INFO epoch: *******************************************122025-02-25 22:42:17,379 INFO Eval results (precision and loss): 59.9101 1.67972025-02-25 22:42:17,380 INFO Best results (precision): 61.05842025-02-25 22:42:17,380 INFO -----------------------------2025-02-25 22:42:17,380 INFO evaluate loss:1.67970023478423382025-02-25 22:42:17,380 INFO -----------------------------2025-02-25 22:42:17,579 INFO Epoch 13, Batch 0, lr = 0.00040, val_event_loss = 1.5684, val_precision = 59.1204, train_event_loss = 0.0170, train_precision = 3.1481, 2025-02-25 22:42:17,640 INFO Epoch 13, Batch 10, lr = 0.00040, train_event_loss = 0.4286, train_precision = 86.7500, 2025-02-25 22:42:17,688 INFO Epoch 13, Batch 20, lr = 0.00040, train_event_loss = 0.4306, train_precision = 86.6250, 2025-02-25 22:42:17,739 INFO Epoch 13, Batch 30, lr = 0.00040, train_event_loss = 0.4741, train_precision = 83.8750, 2025-02-25 22:42:17,788 INFO Epoch 13, Batch 40, lr = 0.00040, train_event_loss = 0.4371, train_precision = 85.5000, 2025-02-25 22:42:17,840 INFO Epoch 13, Batch 50, lr = 0.00040, train_event_loss = 0.4822, train_precision = 83.8750, 2025-02-25 22:42:17,887 INFO Epoch 13, Batch 60, lr = 0.00040, train_event_loss = 0.4298, train_precision = 84.9808, 2025-02-25 22:42:17,923 INFO Train results (precision): 85.16522025-02-25 22:42:17,923 INFO epoch: *******************************************132025-02-25 22:42:19,765 INFO Eval results (precision and loss): 57.9631 1.68092025-02-25 22:42:19,766 INFO Best results (precision): 61.05842025-02-25 22:42:19,766 INFO -----------------------------2025-02-25 22:42:19,766 INFO evaluate loss:1.68088777018011682025-02-25 22:42:19,766 INFO -----------------------------2025-02-25 22:42:19,966 INFO Epoch 14, Batch 0, lr = 0.00020, val_event_loss = 1.5665, val_precision = 57.3148, train_event_loss = 0.0133, train_precision = 3.1944, 2025-02-25 22:42:20,029 INFO Epoch 14, Batch 10, lr = 0.00020, train_event_loss = 0.4485, train_precision = 84.1250, 2025-02-25 22:42:20,082 INFO Epoch 14, Batch 20, lr = 0.00020, train_event_loss = 0.4474, train_precision = 84.0000, 2025-02-25 22:42:20,133 INFO Epoch 14, Batch 30, lr = 0.00020, train_event_loss = 0.3895, train_precision = 87.3750, 2025-02-25 22:42:20,183 INFO Epoch 14, Batch 40, lr = 0.00020, train_event_loss = 0.4169, train_precision = 87.2500, 2025-02-25 22:42:20,233 INFO Epoch 14, Batch 50, lr = 0.00020, train_event_loss = 0.5203, train_precision = 83.3750, 2025-02-25 22:42:20,280 INFO Epoch 14, Batch 60, lr = 0.00020, train_event_loss = 0.4518, train_precision = 84.5673, 2025-02-25 22:42:20,316 INFO Train results (precision): 85.24832025-02-25 22:42:20,316 INFO epoch: *******************************************142025-02-25 22:42:22,138 INFO Eval results (precision and loss): 58.1628 1.69192025-02-25 22:42:22,139 INFO Best results (precision): 61.05842025-02-25 22:42:22,139 INFO -----------------------------2025-02-25 22:42:22,139 INFO evaluate loss:1.69188838964808962025-02-25 22:42:22,139 INFO -----------------------------2025-02-25 22:42:22,335 INFO Epoch 15, Batch 0, lr = 0.00020, val_event_loss = 1.5774, val_precision = 57.5000, train_event_loss = 0.0157, train_precision = 3.3333, 2025-02-25 22:42:22,389 INFO Epoch 15, Batch 10, lr = 0.00020, train_event_loss = 0.4257, train_precision = 86.1250, 2025-02-25 22:42:22,441 INFO Epoch 15, Batch 20, lr = 0.00020, train_event_loss = 0.4310, train_precision = 86.1250, 2025-02-25 22:42:22,493 INFO Epoch 15, Batch 30, lr = 0.00020, train_event_loss = 0.5049, train_precision = 84.1250, 2025-02-25 22:42:22,544 INFO Epoch 15, Batch 40, lr = 0.00020, train_event_loss = 0.4523, train_precision = 83.7500, 2025-02-25 22:42:22,595 INFO Epoch 15, Batch 50, lr = 0.00020, train_event_loss = 0.4267, train_precision = 85.1250, 2025-02-25 22:42:22,643 INFO Epoch 15, Batch 60, lr = 0.00020, train_event_loss = 0.4588, train_precision = 84.1731, 2025-02-25 22:42:22,680 INFO Train results (precision): 85.20672025-02-25 22:42:22,680 INFO epoch: *******************************************152025-02-25 22:42:24,509 INFO Eval results (precision and loss): 58.8617 1.71322025-02-25 22:42:24,509 INFO Best results (precision): 61.05842025-02-25 22:42:24,510 INFO -----------------------------2025-02-25 22:42:24,510 INFO evaluate loss:1.71321525092230492025-02-25 22:42:24,510 INFO -----------------------------2025-02-25 22:42:24,720 INFO Epoch 16, Batch 0, lr = 0.00020, val_event_loss = 1.5983, val_precision = 58.1481, train_event_loss = 0.0134, train_precision = 3.1944, 2025-02-25 22:42:24,781 INFO Epoch 16, Batch 10, lr = 0.00020, train_event_loss = 0.4752, train_precision = 85.5000, 2025-02-25 22:42:24,830 INFO Epoch 16, Batch 20, lr = 0.00020, train_event_loss = 0.4355, train_precision = 85.0000, 2025-02-25 22:42:24,883 INFO Epoch 16, Batch 30, lr = 0.00020, train_event_loss = 0.4002, train_precision = 87.3750, 2025-02-25 22:42:24,934 INFO Epoch 16, Batch 40, lr = 0.00020, train_event_loss = 0.4490, train_precision = 86.0000, 2025-02-25 22:42:24,984 INFO Epoch 16, Batch 50, lr = 0.00020, train_event_loss = 0.4906, train_precision = 83.1250, 2025-02-25 22:42:25,031 INFO Epoch 16, Batch 60, lr = 0.00020, train_event_loss = 0.3813, train_precision = 87.7308, 2025-02-25 22:42:25,065 INFO Train results (precision): 85.70542025-02-25 22:42:25,065 INFO epoch: *******************************************162025-02-25 22:42:26,884 INFO Eval results (precision and loss): 58.5122 1.70582025-02-25 22:42:26,884 INFO Best results (precision): 61.05842025-02-25 22:42:26,884 INFO -----------------------------2025-02-25 22:42:26,884 INFO evaluate loss:1.7058066173964312025-02-25 22:42:26,884 INFO -----------------------------2025-02-25 22:42:27,083 INFO Epoch 17, Batch 0, lr = 0.00020, val_event_loss = 1.5911, val_precision = 57.8241, train_event_loss = 0.0155, train_precision = 3.1019, 2025-02-25 22:42:27,143 INFO Epoch 17, Batch 10, lr = 0.00020, train_event_loss = 0.4034, train_precision = 87.5000, 2025-02-25 22:42:27,192 INFO Epoch 17, Batch 20, lr = 0.00020, train_event_loss = 0.4376, train_precision = 85.1250, 2025-02-25 22:42:27,242 INFO Epoch 17, Batch 30, lr = 0.00020, train_event_loss = 0.4633, train_precision = 85.1250, 2025-02-25 22:42:27,292 INFO Epoch 17, Batch 40, lr = 0.00020, train_event_loss = 0.4326, train_precision = 86.7500, 2025-02-25 22:42:27,342 INFO Epoch 17, Batch 50, lr = 0.00020, train_event_loss = 0.4214, train_precision = 86.5000, 2025-02-25 22:42:27,388 INFO Epoch 17, Batch 60, lr = 0.00020, train_event_loss = 0.4673, train_precision = 84.1058, 2025-02-25 22:42:27,423 INFO Train results (precision): 85.72622025-02-25 22:42:27,423 INFO epoch: *******************************************172025-02-25 22:42:29,265 INFO Eval results (precision and loss): 60.0100 1.69812025-02-25 22:42:29,265 INFO Best results (precision): 61.05842025-02-25 22:42:29,266 INFO -----------------------------2025-02-25 22:42:29,266 INFO evaluate loss:1.69811033373097952025-02-25 22:42:29,266 INFO -----------------------------2025-02-25 22:42:29,466 INFO Epoch 18, Batch 0, lr = 0.00020, val_event_loss = 1.5858, val_precision = 59.2130, train_event_loss = 0.0161, train_precision = 3.0093, 2025-02-25 22:42:29,527 INFO Epoch 18, Batch 10, lr = 0.00020, train_event_loss = 0.4283, train_precision = 86.6250, 2025-02-25 22:42:29,578 INFO Epoch 18, Batch 20, lr = 0.00020, train_event_loss = 0.4492, train_precision = 84.7500, 2025-02-25 22:42:29,628 INFO Epoch 18, Batch 30, lr = 0.00020, train_event_loss = 0.4858, train_precision = 83.2500, 2025-02-25 22:42:29,679 INFO Epoch 18, Batch 40, lr = 0.00020, train_event_loss = 0.4327, train_precision = 85.1250, 2025-02-25 22:42:29,729 INFO Epoch 18, Batch 50, lr = 0.00020, train_event_loss = 0.4242, train_precision = 85.7500, 2025-02-25 22:42:29,775 INFO Epoch 18, Batch 60, lr = 0.00020, train_event_loss = 0.4679, train_precision = 83.6923, 2025-02-25 22:42:29,808 INFO Train results (precision): 84.91592025-02-25 22:42:29,808 INFO epoch: *******************************************182025-02-25 22:42:31,626 INFO Eval results (precision and loss): 59.9601 1.72852025-02-25 22:42:31,626 INFO Best results (precision): 61.05842025-02-25 22:42:31,627 INFO -----------------------------2025-02-25 22:42:31,627 INFO evaluate loss:1.72852713665860812025-02-25 22:42:31,627 INFO -----------------------------2025-02-25 22:42:31,830 INFO Epoch 19, Batch 0, lr = 0.00020, val_event_loss = 1.6145, val_precision = 59.1667, train_event_loss = 0.0135, train_precision = 3.3333, 2025-02-25 22:42:31,886 INFO Epoch 19, Batch 10, lr = 0.00020, train_event_loss = 0.4490, train_precision = 84.5000, 2025-02-25 22:42:31,936 INFO Epoch 19, Batch 20, lr = 0.00020, train_event_loss = 0.4112, train_precision = 86.8750, 2025-02-25 22:42:31,991 INFO Epoch 19, Batch 30, lr = 0.00020, train_event_loss = 0.4221, train_precision = 85.5000, 2025-02-25 22:42:32,043 INFO Epoch 19, Batch 40, lr = 0.00020, train_event_loss = 0.4768, train_precision = 84.0000, 2025-02-25 22:42:32,093 INFO Epoch 19, Batch 50, lr = 0.00020, train_event_loss = 0.4219, train_precision = 86.6250, 2025-02-25 22:42:32,141 INFO Epoch 19, Batch 60, lr = 0.00020, train_event_loss = 0.4733, train_precision = 84.5865, 2025-02-25 22:42:32,177 INFO Train results (precision): 85.43532025-02-25 22:42:32,177 INFO epoch: *******************************************192025-02-25 22:42:34,023 INFO Eval results (precision and loss): 60.1098 1.70792025-02-25 22:42:34,024 INFO Best results (precision): 61.05842025-02-25 22:42:34,024 INFO -----------------------------2025-02-25 22:42:34,024 INFO evaluate loss:1.7079448522272112025-02-25 22:42:34,024 INFO -----------------------------2025-02-25 22:42:34,225 INFO Epoch 20, Batch 0, lr = 0.00020, val_event_loss = 1.5924, val_precision = 59.3056, train_event_loss = 0.0159, train_precision = 3.1481, 2025-02-25 22:42:34,276 INFO Epoch 20, Batch 10, lr = 0.00020, train_event_loss = 0.4281, train_precision = 85.7500, 2025-02-25 22:42:34,327 INFO Epoch 20, Batch 20, lr = 0.00020, train_event_loss = 0.4440, train_precision = 83.8750, 2025-02-25 22:42:34,378 INFO Epoch 20, Batch 30, lr = 0.00020, train_event_loss = 0.4087, train_precision = 85.3750, 2025-02-25 22:42:34,429 INFO Epoch 20, Batch 40, lr = 0.00020, train_event_loss = 0.4043, train_precision = 87.1250, 2025-02-25 22:42:34,479 INFO Epoch 20, Batch 50, lr = 0.00020, train_event_loss = 0.4815, train_precision = 85.0000, 2025-02-25 22:42:34,531 INFO Epoch 20, Batch 60, lr = 0.00020, train_event_loss = 0.4707, train_precision = 83.8558, 2025-02-25 22:42:34,572 INFO Train results (precision): 85.06132025-02-25 22:42:34,573 INFO epoch: *******************************************202025-02-25 22:42:36,444 INFO Eval results (precision and loss): 60.2097 1.71912025-02-25 22:42:36,444 INFO Best results (precision): 61.05842025-02-25 22:42:36,445 INFO -----------------------------2025-02-25 22:42:36,445 INFO evaluate loss:1.71909243052304642025-02-25 22:42:36,445 INFO -----------------------------2025-02-25 22:42:36,643 INFO Epoch 21, Batch 0, lr = 0.00020, val_event_loss = 1.6037, val_precision = 59.3981, train_event_loss = 0.0136, train_precision = 3.1019, 2025-02-25 22:42:36,697 INFO Epoch 21, Batch 10, lr = 0.00020, train_event_loss = 0.4345, train_precision = 85.7500, 2025-02-25 22:42:36,748 INFO Epoch 21, Batch 20, lr = 0.00020, train_event_loss = 0.4343, train_precision = 85.2500, 2025-02-25 22:42:36,798 INFO Epoch 21, Batch 30, lr = 0.00020, train_event_loss = 0.4924, train_precision = 82.7500, 2025-02-25 22:42:36,850 INFO Epoch 21, Batch 40, lr = 0.00020, train_event_loss = 0.4285, train_precision = 86.3750, 2025-02-25 22:42:36,901 INFO Epoch 21, Batch 50, lr = 0.00020, train_event_loss = 0.4136, train_precision = 87.7500, 2025-02-25 22:42:36,949 INFO Epoch 21, Batch 60, lr = 0.00020, train_event_loss = 0.4185, train_precision = 85.3365, 2025-02-25 22:42:36,982 INFO Train results (precision): 85.51842025-02-25 22:42:36,982 INFO epoch: *******************************************212025-02-25 22:42:38,833 INFO Eval results (precision and loss): 59.7604 1.72082025-02-25 22:42:38,833 INFO Best results (precision): 61.05842025-02-25 22:42:38,834 INFO -----------------------------2025-02-25 22:42:38,834 INFO evaluate loss:1.72079370225443462025-02-25 22:42:38,834 INFO -----------------------------2025-02-25 22:42:39,042 INFO Epoch 22, Batch 0, lr = 0.00020, val_event_loss = 1.6033, val_precision = 58.9815, train_event_loss = 0.0103, train_precision = 3.3333, 2025-02-25 22:42:39,102 INFO Epoch 22, Batch 10, lr = 0.00020, train_event_loss = 0.4229, train_precision = 85.3750, 2025-02-25 22:42:39,152 INFO Epoch 22, Batch 20, lr = 0.00020, train_event_loss = 0.4397, train_precision = 85.2500, 2025-02-25 22:42:39,208 INFO Epoch 22, Batch 30, lr = 0.00020, train_event_loss = 0.4411, train_precision = 85.2500, 2025-02-25 22:42:39,260 INFO Epoch 22, Batch 40, lr = 0.00020, train_event_loss = 0.4374, train_precision = 84.6250, 2025-02-25 22:42:39,310 INFO Epoch 22, Batch 50, lr = 0.00020, train_event_loss = 0.4447, train_precision = 85.7500, 2025-02-25 22:42:39,357 INFO Epoch 22, Batch 60, lr = 0.00020, train_event_loss = 0.5090, train_precision = 83.7981, 2025-02-25 22:42:39,395 INFO Train results (precision): 85.31062025-02-25 22:42:39,395 INFO epoch: *******************************************222025-02-25 22:42:41,255 INFO Eval results (precision and loss): 58.1628 1.71332025-02-25 22:42:41,255 INFO Best results (precision): 61.05842025-02-25 22:42:41,256 INFO -----------------------------2025-02-25 22:42:41,256 INFO evaluate loss:1.71325676639014262025-02-25 22:42:41,256 INFO -----------------------------2025-02-25 22:42:41,454 INFO Epoch 23, Batch 0, lr = 0.00020, val_event_loss = 1.5976, val_precision = 57.5000, train_event_loss = 0.0134, train_precision = 3.2870, 2025-02-25 22:42:41,507 INFO Epoch 23, Batch 10, lr = 0.00020, train_event_loss = 0.4003, train_precision = 86.6250, 2025-02-25 22:42:41,556 INFO Epoch 23, Batch 20, lr = 0.00020, train_event_loss = 0.4814, train_precision = 84.1250, 2025-02-25 22:42:41,606 INFO Epoch 23, Batch 30, lr = 0.00020, train_event_loss = 0.4176, train_precision = 86.7500, 2025-02-25 22:42:41,656 INFO Epoch 23, Batch 40, lr = 0.00020, train_event_loss = 0.4549, train_precision = 85.2500, 2025-02-25 22:42:41,707 INFO Epoch 23, Batch 50, lr = 0.00020, train_event_loss = 0.4540, train_precision = 86.0000, 2025-02-25 22:42:41,754 INFO Epoch 23, Batch 60, lr = 0.00020, train_event_loss = 0.3917, train_precision = 85.7308, 2025-02-25 22:42:41,793 INFO Train results (precision): 85.70542025-02-25 22:42:41,793 INFO epoch: *******************************************232025-02-25 22:42:43,630 INFO Eval results (precision and loss): 60.1098 1.71182025-02-25 22:42:43,630 INFO Best results (precision): 61.05842025-02-25 22:42:43,630 INFO -----------------------------2025-02-25 22:42:43,630 INFO evaluate loss:1.71180309780965412025-02-25 22:42:43,630 INFO -----------------------------2025-02-25 22:42:43,825 INFO Epoch 24, Batch 0, lr = 0.00010, val_event_loss = 1.5955, val_precision = 59.3056, train_event_loss = 0.0125, train_precision = 3.2407, 2025-02-25 22:42:43,882 INFO Epoch 24, Batch 10, lr = 0.00010, train_event_loss = 0.4433, train_precision = 85.6250, 2025-02-25 22:42:43,931 INFO Epoch 24, Batch 20, lr = 0.00010, train_event_loss = 0.3954, train_precision = 87.1250, 2025-02-25 22:42:43,981 INFO Epoch 24, Batch 30, lr = 0.00010, train_event_loss = 0.4506, train_precision = 85.2500, 2025-02-25 22:42:44,033 INFO Epoch 24, Batch 40, lr = 0.00010, train_event_loss = 0.4370, train_precision = 86.2500, 2025-02-25 22:42:44,087 INFO Epoch 24, Batch 50, lr = 0.00010, train_event_loss = 0.4351, train_precision = 85.0000, 2025-02-25 22:42:44,134 INFO Epoch 24, Batch 60, lr = 0.00010, train_event_loss = 0.4506, train_precision = 85.5000, 2025-02-25 22:42:44,169 INFO Train results (precision): 85.62232025-02-25 22:42:44,170 INFO epoch: *******************************************242025-02-25 22:42:46,033 INFO Eval results (precision and loss): 58.8617 1.74012025-02-25 22:42:46,033 INFO Best results (precision): 61.05842025-02-25 22:42:46,034 INFO -----------------------------2025-02-25 22:42:46,034 INFO evaluate loss:1.7400983318730042025-02-25 22:42:46,034 INFO -----------------------------2025-02-25 22:42:46,241 INFO Epoch 25, Batch 0, lr = 0.00010, val_event_loss = 1.6213, val_precision = 58.1481, train_event_loss = 0.0152, train_precision = 3.2870, 2025-02-25 22:42:46,301 INFO Epoch 25, Batch 10, lr = 0.00010, train_event_loss = 0.4017, train_precision = 86.5000, 2025-02-25 22:42:46,350 INFO Epoch 25, Batch 20, lr = 0.00010, train_event_loss = 0.4232, train_precision = 85.2500, 2025-02-25 22:42:46,401 INFO Epoch 25, Batch 30, lr = 0.00010, train_event_loss = 0.4324, train_precision = 86.2500, 2025-02-25 22:42:46,455 INFO Epoch 25, Batch 40, lr = 0.00010, train_event_loss = 0.4453, train_precision = 85.5000, 2025-02-25 22:42:46,505 INFO Epoch 25, Batch 50, lr = 0.00010, train_event_loss = 0.4158, train_precision = 85.8750, 2025-02-25 22:42:46,553 INFO Epoch 25, Batch 60, lr = 0.00010, train_event_loss = 0.4690, train_precision = 84.4808, 2025-02-25 22:42:46,589 INFO Train results (precision): 85.60152025-02-25 22:42:46,589 INFO epoch: *******************************************252025-02-25 22:42:48,462 INFO Eval results (precision and loss): 59.5107 1.72842025-02-25 22:42:48,462 INFO Best results (precision): 61.05842025-02-25 22:42:48,462 INFO -----------------------------2025-02-25 22:42:48,462 INFO evaluate loss:1.72837298645862152025-02-25 22:42:48,462 INFO -----------------------------2025-02-25 22:42:48,659 INFO Epoch 26, Batch 0, lr = 0.00010, val_event_loss = 1.6150, val_precision = 58.7500, train_event_loss = 0.0110, train_precision = 3.2870, 2025-02-25 22:42:48,714 INFO Epoch 26, Batch 10, lr = 0.00010, train_event_loss = 0.3901, train_precision = 87.5000, 2025-02-25 22:42:48,767 INFO Epoch 26, Batch 20, lr = 0.00010, train_event_loss = 0.3977, train_precision = 87.5000, 2025-02-25 22:42:48,817 INFO Epoch 26, Batch 30, lr = 0.00010, train_event_loss = 0.4476, train_precision = 84.3750, 2025-02-25 22:42:48,868 INFO Epoch 26, Batch 40, lr = 0.00010, train_event_loss = 0.4240, train_precision = 85.5000, 2025-02-25 22:42:48,921 INFO Epoch 26, Batch 50, lr = 0.00010, train_event_loss = 0.4538, train_precision = 84.0000, 2025-02-25 22:42:48,969 INFO Epoch 26, Batch 60, lr = 0.00010, train_event_loss = 0.4298, train_precision = 86.2308, 2025-02-25 22:42:49,004 INFO Train results (precision): 85.80932025-02-25 22:42:49,004 INFO epoch: *******************************************262025-02-25 22:42:50,869 INFO Eval results (precision and loss): 61.2581 1.73622025-02-25 22:42:50,869 INFO Best results (precision): 61.25812025-02-25 22:42:50,870 INFO -----------------------------2025-02-25 22:42:50,870 INFO evaluate loss:1.73624917939647122025-02-25 22:42:50,870 INFO -----------------------------2025-02-25 22:42:51,067 INFO Epoch 27, Batch 0, lr = 0.00010, val_event_loss = 1.6194, val_precision = 60.3704, train_event_loss = 0.0172, train_precision = 3.1019, 2025-02-25 22:42:51,126 INFO Epoch 27, Batch 10, lr = 0.00010, train_event_loss = 0.4313, train_precision = 87.1250, 2025-02-25 22:42:51,180 INFO Epoch 27, Batch 20, lr = 0.00010, train_event_loss = 0.4144, train_precision = 86.1250, 2025-02-25 22:42:51,238 INFO Epoch 27, Batch 30, lr = 0.00010, train_event_loss = 0.3923, train_precision = 86.3750, 2025-02-25 22:42:51,290 INFO Epoch 27, Batch 40, lr = 0.00010, train_event_loss = 0.3991, train_precision = 85.2500, 2025-02-25 22:42:51,341 INFO Epoch 27, Batch 50, lr = 0.00010, train_event_loss = 0.5136, train_precision = 83.5000, 2025-02-25 22:42:51,387 INFO Epoch 27, Batch 60, lr = 0.00010, train_event_loss = 0.4328, train_precision = 85.8365, 2025-02-25 22:42:51,427 INFO Train results (precision): 85.68462025-02-25 22:42:51,427 INFO epoch: *******************************************272025-02-25 22:42:53,284 INFO Eval results (precision and loss): 57.8133 1.73662025-02-25 22:42:53,284 INFO Best results (precision): 61.25812025-02-25 22:42:53,285 INFO -----------------------------2025-02-25 22:42:53,285 INFO evaluate loss:1.73659518664282062025-02-25 22:42:53,285 INFO -----------------------------2025-02-25 22:42:53,488 INFO Epoch 28, Batch 0, lr = 0.00010, val_event_loss = 1.6206, val_precision = 57.1759, train_event_loss = 0.0140, train_precision = 3.1944, 2025-02-25 22:42:53,547 INFO Epoch 28, Batch 10, lr = 0.00010, train_event_loss = 0.4243, train_precision = 87.0000, 2025-02-25 22:42:53,596 INFO Epoch 28, Batch 20, lr = 0.00010, train_event_loss = 0.4168, train_precision = 86.6250, 2025-02-25 22:42:53,647 INFO Epoch 28, Batch 30, lr = 0.00010, train_event_loss = 0.3982, train_precision = 86.0000, 2025-02-25 22:42:53,699 INFO Epoch 28, Batch 40, lr = 0.00010, train_event_loss = 0.4565, train_precision = 85.2500, 2025-02-25 22:42:53,752 INFO Epoch 28, Batch 50, lr = 0.00010, train_event_loss = 0.4407, train_precision = 85.6250, 2025-02-25 22:42:53,801 INFO Epoch 28, Batch 60, lr = 0.00010, train_event_loss = 0.4662, train_precision = 84.9808, 2025-02-25 22:42:53,835 INFO Train results (precision): 85.83002025-02-25 22:42:53,836 INFO epoch: *******************************************282025-02-25 22:42:55,694 INFO Eval results (precision and loss): 59.7104 1.75802025-02-25 22:42:55,694 INFO Best results (precision): 61.25812025-02-25 22:42:55,695 INFO -----------------------------2025-02-25 22:42:55,695 INFO evaluate loss:1.75800099148261022025-02-25 22:42:55,695 INFO -----------------------------2025-02-25 22:42:55,894 INFO Epoch 29, Batch 0, lr = 0.00010, val_event_loss = 1.6417, val_precision = 58.9352, train_event_loss = 0.0148, train_precision = 3.2407, 2025-02-25 22:42:55,947 INFO Epoch 29, Batch 10, lr = 0.00010, train_event_loss = 0.4310, train_precision = 85.6250, 2025-02-25 22:42:55,999 INFO Epoch 29, Batch 20, lr = 0.00010, train_event_loss = 0.3987, train_precision = 86.8750, 2025-02-25 22:42:56,053 INFO Epoch 29, Batch 30, lr = 0.00010, train_event_loss = 0.4140, train_precision = 87.3750, 2025-02-25 22:42:56,103 INFO Epoch 29, Batch 40, lr = 0.00010, train_event_loss = 0.4589, train_precision = 84.3750, 2025-02-25 22:42:56,154 INFO Epoch 29, Batch 50, lr = 0.00010, train_event_loss = 0.4306, train_precision = 85.2500, 2025-02-25 22:42:56,201 INFO Epoch 29, Batch 60, lr = 0.00010, train_event_loss = 0.4017, train_precision = 87.1058, 2025-02-25 22:42:56,238 INFO Train results (precision): 86.03782025-02-25 22:42:56,238 INFO epoch: *******************************************292025-02-25 22:42:58,101 INFO Eval results (precision and loss): 59.6106 1.74162025-02-25 22:42:58,101 INFO Best results (precision): 61.25812025-02-25 22:42:58,101 INFO -----------------------------2025-02-25 22:42:58,101 INFO evaluate loss:1.74162832468800692025-02-25 22:42:58,101 INFO -----------------------------

复制代码

精确率变化
loss值变化

论文

论文解读：https://blog.csdn.net/m0_59164520/article/details/142768900

本文介绍了一种新的任务——跨模态泛化（Cross Modal Generalization），旨在解决在预训练过程中如何学习配对多模态数据的统一离散表示的问题。该方法能够在下游任务中实现零样本泛化能力，在只有一类数据被标记的情况下，模型可以在其他类别上进行推理。现有的多模态表征学习方法更注重粗粒度对齐或依赖于不同模态信息完全对齐的假设，这在现实世界场景下是不切实际的。为了解决这个问题，本文提出了Uni-Code，其中包含两个关键贡献：双交叉模态信息解耦模块和多模态指数移动平均值模块。这些方法促进了模态之间的双向监督，并在共享的离散潜空间中对齐语义等效信息，从而实现了多模态序列的精细统一表示。在预训练阶段，我们探究了各种模态组合，包括音频视觉、音频文本以及音频视觉文本三元组。在各种下游任务上的广泛实验表明，我们的方法是有效的。
论文背景与动机

在多模态学习中，标注数据往往成本高昂且难以获取，尤其是对于某些模态（如音频）的标注更为稀缺。因此，如何利用未标注的多模态数据对齐不同模态的语义信息，从而实现跨模态的零样本泛化能力，是一个重要的研究方向。例如，我们希望在只有视频模态标注的情况下，模型能够迁移到音频模态上完成任务。
为了解决这一问题，论文提出了 Cross Modal Generalization (CMG) 任务，目标是通过预训练阶段学习多模态数据对的统一离散表示，使得在下游任务中，模型能够在只有部分模态标注的情况下，迁移到其他未标注模态。
研究方法

方法描述

本文提出了一种跨模态泛化任务（Cross Modal Generalization）的方法，该方法通过在预训练阶段将不同模态的数据映射到一个统一的离散空间中，使得这些不同的模态具有相同的语义信息，并且能够共享离散编码。在下游任务中，当只有某一模态（例如A模式）有标注信息时，模型可以利用预训练过程中获得的共享离散空间来实现零样本泛化能力。
方法改进

与之前的研究相比，本文提出了统一表示学习（Unified Representation Learning）的方法。首先，引入了DCID模块，用于提取细粒度的语义信息并将其从每个模态中的相应特定信息中分离出来。其次，使用VQ-VAE压缩提取的语义特征为离散变量，确保压缩后的离散变量仍能包含原始的语义信息通过重建损失。本文以两个模态为例说明了整个过程。
解决的问题

本文提出的跨模态泛化任务解决了在实际应用中多模态数据之间的语义差异问题，提高了模型的泛化能力和性能表现。同时，统一表示学习的方法进一步提升了模型的表达能力和效果。
实验设计

第五章的实验部分详细介绍了作者提出的**Cross Modal Generalization (CMG)任务的实验设计、实验过程以及实验结果的分析。
数据集

预训练数据集：作者使用了VGGSound-AVEL数据集的音频-视觉对进行预训练，并将其划分为不同规模（24K、40K、81K）以研究数据规模对模型性能的影响。
下游任务数据集：为了验证预训练模型的泛化能力，作者在多个下游任务中使用了不同的数据集，包括：
- AVE（Audio-Visual Event Classification）：用于跨模态事件分类。
- AVVP（Audio-Visual Video Parsing）：用于跨模态事件定位。
- UCF-101 和 VGGSound-AVEL：用于跨数据集和跨模态的分类任务。
- AVS-S4（Audio-Visual Segmentation）：用于跨模态视频分割。
- VGGSound-AVEL：用于跨模态检索任务。

实验任务

跨模态事件分类（AVE）：使用单模态（如视觉）训练事件分类器，并在另一模态（如音频）上直接评估性能。
跨模态事件定位（AVVP）：在单模态上进行事件定位训练，并在另一模态上测试。
跨模态检索：训练视觉-文本统一表示，通过音频作为中介测试跨模态检索能力。
跨模态视频分割（AVS-S4）：使用单模态（如音频）训练查询式视频分割，并在另一模态（如文本）上测试。
跨数据集和跨模态分类：在 AVE 数据集上训练分类器，并在 AVVP 数据集上测试。

基线和对比方法

基线方法：使用第 4.1 节中描述的基线模型。
对比方法：与多种现有的多模态统一表示方法进行对比，包括 MST、CODIS、S-MiT、CMCM 和 TURN。

评估指标

AVE 和 AVVP：使用精度（Precision）和准确率（Accuracy）评估模型性能。
AVE 到 AVVP 的泛化任务：使用 F1 分数。
AVS-S4：使用 mIoU 和 F-Score。
跨模态检索：使用 R@5 和 R@10（Top-5 和 Top-10 的召回率）。

实验结果分析

与现有方法的对比

表 1 显示了在 AVE 和 AVVP 任务上，Uni-Code 方法与其他现有方法的性能对比。实验结果表明：
- Uni-Code 在所有设置下均优于现有的方法，例如在 VGGSound-AVEL 24K 数据集上，Uni-Code 的 AVE 和 AVVP 任务性能分别达到了 44.0% 和 49.7%，显著高于其他方法。
- 随着预训练数据规模的增加，模型性能有所提升，但当数据规模达到 81K 时，性能提升不明显甚至略有下降，这表明过多的数据可能导致模型难以学习到正确的对齐关系。

消融研究

表 2 和表 3 分别展示了在音频-视觉和音频-视觉-文本预训练任务上的消融研究结果：
- 去除 Cross-CPC：性能下降最为显著，表明 Cross-CPC 在跨模态细粒度对齐中的重要性。
- 去除 MM-EMA：模型性能显著下降，说明 MM-EMA 通过跨模态指导有效促进了统一表示。
- 去除“重置代码”模块：导致性能下降，表明该模块有效减少了冗余的离散码。
- 代码本大小的影响：如图 3 所示，适当的代码本大小可以实现最佳的预训练性能。代码本过大或过小都会导致性能下降。
- 三模态统一表示：与双模态相比，三模态预训练的性能更高，且去除模块后的性能下降幅度较小，说明第三模态的引入有助于其他模态的对齐。

下游任务的性能

跨模态视频分割（AVS-S4）：
- 表 4 显示 Uni-Code 在 A2T 和 T2A 任务上的性能分别为 mIoU=78.0 和 mIoU=77.7，显著优于基线模型。
- 图 5 的可视化结果表明，Uni-Code 能够准确识别未知模态中的声音区域。
跨模态检索任务：
- 表 5 显示 Uni-Code 在 V2T 和 T2V 的音频检索任务中，R@5 和 R@10 的性能显著优于基线模型。

实验结果的总结

有效性：Uni-Code 方法在多个下游任务中均表现出色，证明了其在跨模态泛化任务中的有效性。
模块的重要性：消融研究结果表明，DCID 和 MM-EMA 模块对于实现细粒度的多模态统一表示至关重要。
数据规模的影响：适当的预训练数据规模有助于模型学习正确的对齐关系，过多的数据可能导致性能下降。
多模态组合的优势：Uni-Code 不仅支持音频-视觉和音频-文本的对齐，还能扩展到三模态（音频-视觉-文本）的统一表示，进一步提升了模型的泛化能力。

结论

论文提出了一种新的跨模态泛化任务，并通过 Uni-Code 框架实现了多模态的细粒度统一表示。实验结果表明，Uni-Code 在多种下游任务中均优于现有方法，尤其是在跨模态事件分类、定位和检索任务中表现出色。此外，Uni-Code 还支持三模态统一表示，进一步提升了跨模态对齐的效果。
未来工作

尽管 Uni-Code 在双模态和三模态对齐中表现出色，但其在更多模态组合（如视觉、音频、文本、触觉等）上的效果仍有待探索。此外，如何进一步优化离散码的对齐精度和减少预训练数据规模对性能的影响，也是未来的研究方向。

来源：程序园用户自行投稿发布，如果侵权，请联系站长删除
免责声明：如果侵犯了您的权益，请联系站长，我们会及时删除侵权内容，谢谢合作！

账号		自动登录	找回密码
密码			立即注册