diff --git a/.gitattributes b/.gitattributes index a6344aac8c09253b3b630fb776ae94478aa0275b..0e720e77b9ef63a5bf28d3285cd4ea0cc7a52e2d 100644 --- a/.gitattributes +++ b/.gitattributes @@ -1,35 +1,43 @@ -*.7z filter=lfs diff=lfs merge=lfs -text -*.arrow filter=lfs diff=lfs merge=lfs -text -*.bin filter=lfs diff=lfs merge=lfs -text -*.bz2 filter=lfs diff=lfs merge=lfs -text -*.ckpt filter=lfs diff=lfs merge=lfs -text -*.ftz filter=lfs diff=lfs merge=lfs -text -*.gz filter=lfs diff=lfs merge=lfs -text -*.h5 filter=lfs diff=lfs merge=lfs -text -*.joblib filter=lfs diff=lfs merge=lfs -text -*.lfs.* filter=lfs diff=lfs merge=lfs -text -*.mlmodel filter=lfs diff=lfs merge=lfs -text -*.model filter=lfs diff=lfs merge=lfs -text -*.msgpack filter=lfs diff=lfs merge=lfs -text -*.npy filter=lfs diff=lfs merge=lfs -text -*.npz filter=lfs diff=lfs merge=lfs -text -*.onnx filter=lfs diff=lfs merge=lfs -text -*.ot filter=lfs diff=lfs merge=lfs -text -*.parquet filter=lfs diff=lfs merge=lfs -text -*.pb filter=lfs diff=lfs merge=lfs -text -*.pickle filter=lfs diff=lfs merge=lfs -text -*.pkl filter=lfs diff=lfs merge=lfs -text -*.pt filter=lfs diff=lfs merge=lfs -text -*.pth filter=lfs diff=lfs merge=lfs -text -*.rar filter=lfs diff=lfs merge=lfs -text -*.safetensors filter=lfs diff=lfs merge=lfs -text -saved_model/**/* filter=lfs diff=lfs merge=lfs -text -*.tar.* filter=lfs diff=lfs merge=lfs -text -*.tar filter=lfs diff=lfs merge=lfs -text -*.tflite filter=lfs diff=lfs merge=lfs -text -*.tgz filter=lfs diff=lfs merge=lfs -text -*.wasm filter=lfs diff=lfs merge=lfs -text -*.xz filter=lfs diff=lfs merge=lfs -text -*.zip filter=lfs diff=lfs merge=lfs -text -*.zst filter=lfs diff=lfs merge=lfs -text -*tfevents* filter=lfs diff=lfs merge=lfs -text +*.7z filter=lfs diff=lfs merge=lfs -text +*.arrow filter=lfs diff=lfs merge=lfs -text +*.bin filter=lfs diff=lfs merge=lfs -text +*.bz2 filter=lfs diff=lfs merge=lfs -text +*.ckpt filter=lfs diff=lfs merge=lfs -text +*.ftz filter=lfs diff=lfs merge=lfs -text +*.gz filter=lfs diff=lfs merge=lfs -text +*.h5 filter=lfs diff=lfs merge=lfs -text +*.joblib filter=lfs diff=lfs merge=lfs -text +*.lfs.* filter=lfs diff=lfs merge=lfs -text +*.mlmodel filter=lfs diff=lfs merge=lfs -text +*.model filter=lfs diff=lfs merge=lfs -text +*.msgpack filter=lfs diff=lfs merge=lfs -text +*.npy filter=lfs diff=lfs merge=lfs -text +*.npz filter=lfs diff=lfs merge=lfs -text +*.onnx filter=lfs diff=lfs merge=lfs -text +*.ot filter=lfs diff=lfs merge=lfs -text +*.parquet filter=lfs diff=lfs merge=lfs -text +*.pb filter=lfs diff=lfs merge=lfs -text +*.pickle filter=lfs diff=lfs merge=lfs -text +*.pkl filter=lfs diff=lfs merge=lfs -text +*.pt filter=lfs diff=lfs merge=lfs -text +*.pth filter=lfs diff=lfs merge=lfs -text +*.rar filter=lfs diff=lfs merge=lfs -text +*.safetensors filter=lfs diff=lfs merge=lfs -text +saved_model/**/* filter=lfs diff=lfs merge=lfs -text +*.tar.* filter=lfs diff=lfs merge=lfs -text +*.tar filter=lfs diff=lfs merge=lfs -text +*.tflite filter=lfs diff=lfs merge=lfs -text +*.tgz filter=lfs diff=lfs merge=lfs -text +*.wasm filter=lfs diff=lfs merge=lfs -text +*.xz filter=lfs diff=lfs merge=lfs -text +*.zip filter=lfs diff=lfs merge=lfs -text +*.zst filter=lfs diff=lfs merge=lfs -text +*tfevents* filter=lfs diff=lfs merge=lfs -text +checkpoints_v2m_part1/visualizations/base_weights_heatmap.png filter=lfs diff=lfs merge=lfs -text +checkpoints_v2m_part1/visualizations/fine_tuned_weights_heatmap.png filter=lfs diff=lfs merge=lfs -text +checkpoints_v2m_part1/visualizations/mapping_error_distribution.png filter=lfs diff=lfs merge=lfs -text +checkpoints_v2m_part1/visualizations/weight_changes_heatmap.png filter=lfs diff=lfs merge=lfs -text +checkpoints_v2m_part2/visualizations/base_weights_heatmap.png filter=lfs diff=lfs merge=lfs -text +checkpoints_v2m_part2/visualizations/fine_tuned_weights_heatmap.png filter=lfs diff=lfs merge=lfs -text +checkpoints_v2m_part2/visualizations/mapping_error_distribution.png filter=lfs diff=lfs merge=lfs -text +checkpoints_v2m_part2/visualizations/weight_changes_heatmap.png filter=lfs diff=lfs merge=lfs -text diff --git a/checkpoints_v2m_part1/2025-04-11_15-04-17_train.log b/checkpoints_v2m_part1/2025-04-11_15-04-17_train.log new file mode 100644 index 0000000000000000000000000000000000000000..a9c176c291d2bd5d8e109219da75809ed64451e9 --- /dev/null +++ b/checkpoints_v2m_part1/2025-04-11_15-04-17_train.log @@ -0,0 +1,743 @@ +[2025-04-11 15:04:17] [INFO] 使用设备: cuda:0 +[2025-04-11 15:04:17] [INFO] 训练集注释文件: /data0/work/DuYiFan/projects/traffic_classify/full_classes/TsignRecgTrainAnnotation.txt +[2025-04-11 15:04:17] [INFO] 测试集注释文件: /data0/work/DuYiFan/projects/traffic_classify/full_classes/TsignRecgTestAnnotation.txt +[2025-04-11 15:04:17] [INFO] 训练图像目录: /data0/work/DuYiFan/projects/traffic_classify/full_classes/train +[2025-04-11 15:04:17] [INFO] 测试图像目录: /data0/work/DuYiFan/projects/traffic_classify/full_classes/test +[2025-04-11 15:04:17] [INFO] 创建数据集和数据加载器 +[2025-04-11 15:04:17] [INFO] 创建efficientnet-v2-m模型，类别数: 58 +[2025-04-11 15:04:19] [INFO] 设置损失函数、优化器和学习率调度器，初始学习率: 0.0001 +[2025-04-11 15:04:19] [INFO] 开始训练，总共 50 轮 +[2025-04-11 15:04:19] [INFO] 当前学习率: 0.000100 +[2025-04-11 15:04:19] [INFO] Epoch 1/50 开始训练 +[2025-04-11 15:04:38] [INFO] Epoch 1/50 开始验证 +[2025-04-11 15:04:40] [INFO] Epoch 1/50 - Train Loss: 3.5120, Train Acc: 0.1151, Val Loss: 3.4758, Val Acc: 0.1424 +[2025-04-11 15:04:41] [INFO] 已保存最佳模型，准确率: 0.1424 +[2025-04-11 15:04:41] [INFO] 当前学习率: 0.000100 +[2025-04-11 15:04:42] [INFO] Epoch 2/50 开始训练 +[2025-04-11 15:05:00] [INFO] Epoch 2/50 开始验证 +[2025-04-11 15:05:02] [INFO] Epoch 2/50 - Train Loss: 2.7881, Train Acc: 0.2928, Val Loss: 2.7346, Val Acc: 0.3210 +[2025-04-11 15:05:03] [INFO] 已保存最佳模型，准确率: 0.3210 +[2025-04-11 15:05:04] [INFO] 当前学习率: 0.000100 +[2025-04-11 15:05:04] [INFO] Epoch 3/50 开始训练 +[2025-04-11 15:05:22] [INFO] Epoch 3/50 开始验证 +[2025-04-11 15:05:25] [INFO] Epoch 3/50 - Train Loss: 1.9895, Train Acc: 0.5070, Val Loss: 1.8575, Val Acc: 0.4975 +[2025-04-11 15:05:25] [INFO] 已保存最佳模型，准确率: 0.4975 +[2025-04-11 15:05:26] [INFO] 当前学习率: 0.000099 +[2025-04-11 15:05:26] [INFO] Epoch 4/50 开始训练 +[2025-04-11 15:05:44] [INFO] Epoch 4/50 开始验证 +[2025-04-11 15:05:47] [INFO] Epoch 4/50 - Train Loss: 1.5039, Train Acc: 0.6036, Val Loss: 1.4558, Val Acc: 0.5647 +[2025-04-11 15:05:47] [INFO] 已保存最佳模型，准确率: 0.5647 +[2025-04-11 15:05:48] [INFO] 当前学习率: 0.000098 +[2025-04-11 15:05:48] [INFO] Epoch 5/50 开始训练 +[2025-04-11 15:06:07] [INFO] Epoch 5/50 开始验证 +[2025-04-11 15:06:09] [INFO] Epoch 5/50 - Train Loss: 1.1698, Train Acc: 0.6796, Val Loss: 1.2661, Val Acc: 0.6800 +[2025-04-11 15:06:10] [INFO] 已保存最佳模型，准确率: 0.6800 +[2025-04-11 15:06:11] [INFO] 当前学习率: 0.000098 +[2025-04-11 15:06:11] [INFO] Epoch 6/50 开始训练 +[2025-04-11 15:06:29] [INFO] Epoch 6/50 开始验证 +[2025-04-11 15:06:31] [INFO] Epoch 6/50 - Train Loss: 0.9343, Train Acc: 0.7590, Val Loss: 1.0889, Val Acc: 0.7272 +[2025-04-11 15:06:32] [INFO] 已保存最佳模型，准确率: 0.7272 +[2025-04-11 15:06:33] [INFO] 当前学习率: 0.000097 +[2025-04-11 15:06:33] [INFO] Epoch 7/50 开始训练 +[2025-04-11 15:06:51] [INFO] Epoch 7/50 开始验证 +[2025-04-11 15:06:54] [INFO] Epoch 7/50 - Train Loss: 0.8004, Train Acc: 0.8151, Val Loss: 0.9226, Val Acc: 0.7773 +[2025-04-11 15:06:54] [INFO] 已保存最佳模型，准确率: 0.7773 +[2025-04-11 15:06:55] [INFO] 当前学习率: 0.000095 +[2025-04-11 15:06:55] [INFO] Epoch 8/50 开始训练 +[2025-04-11 15:07:14] [INFO] Epoch 8/50 开始验证 +[2025-04-11 15:07:16] [INFO] Epoch 8/50 - Train Loss: 0.6381, Train Acc: 0.8717, Val Loss: 0.8111, Val Acc: 0.7934 +[2025-04-11 15:07:16] [INFO] 已保存最佳模型，准确率: 0.7934 +[2025-04-11 15:07:17] [INFO] 当前学习率: 0.000094 +[2025-04-11 15:07:17] [INFO] Epoch 9/50 开始训练 +[2025-04-11 15:07:36] [INFO] Epoch 9/50 开始验证 +[2025-04-11 15:07:39] [INFO] Epoch 9/50 - Train Loss: 0.5269, Train Acc: 0.8974, Val Loss: 0.6490, Val Acc: 0.8355 +[2025-04-11 15:07:39] [INFO] 已保存最佳模型，准确率: 0.8355 +[2025-04-11 15:07:40] [INFO] 当前学习率: 0.000092 +[2025-04-11 15:07:40] [INFO] Epoch 10/50 开始训练 +[2025-04-11 15:07:59] [INFO] Epoch 10/50 开始验证 +[2025-04-11 15:08:01] [INFO] Epoch 10/50 - Train Loss: 0.3630, Train Acc: 0.9374, Val Loss: 0.5186, Val Acc: 0.8786 +[2025-04-11 15:08:01] [INFO] 已保存最佳模型，准确率: 0.8786 +[2025-04-11 15:08:02] [INFO] 当前学习率: 0.000091 +[2025-04-11 15:08:02] [INFO] Epoch 11/50 开始训练 +[2025-04-11 15:08:21] [INFO] Epoch 11/50 开始验证 +[2025-04-11 15:08:23] [INFO] Epoch 11/50 - Train Loss: 0.2655, Train Acc: 0.9508, Val Loss: 0.4091, Val Acc: 0.9017 +[2025-04-11 15:08:24] [INFO] 已保存最佳模型，准确率: 0.9017 +[2025-04-11 15:08:25] [INFO] 当前学习率: 0.000089 +[2025-04-11 15:08:25] [INFO] Epoch 12/50 开始训练 +[2025-04-11 15:08:43] [INFO] Epoch 12/50 开始验证 +[2025-04-11 15:08:46] [INFO] Epoch 12/50 - Train Loss: 0.2500, Train Acc: 0.9549, Val Loss: 0.5203, Val Acc: 0.8626 +[2025-04-11 15:08:47] [INFO] 当前学习率: 0.000087 +[2025-04-11 15:08:47] [INFO] Epoch 13/50 开始训练 +[2025-04-11 15:09:05] [INFO] Epoch 13/50 开始验证 +[2025-04-11 15:09:08] [INFO] Epoch 13/50 - Train Loss: 0.2335, Train Acc: 0.9530, Val Loss: 0.4292, Val Acc: 0.8877 +[2025-04-11 15:09:09] [INFO] 当前学习率: 0.000084 +[2025-04-11 15:09:09] [INFO] Epoch 14/50 开始训练 +[2025-04-11 15:09:28] [INFO] Epoch 14/50 开始验证 +[2025-04-11 15:09:30] [INFO] Epoch 14/50 - Train Loss: 0.1710, Train Acc: 0.9657, Val Loss: 0.3083, Val Acc: 0.9228 +[2025-04-11 15:09:30] [INFO] 已保存最佳模型，准确率: 0.9228 +[2025-04-11 15:09:31] [INFO] 当前学习率: 0.000082 +[2025-04-11 15:09:31] [INFO] Epoch 15/50 开始训练 +[2025-04-11 15:09:50] [INFO] Epoch 15/50 开始验证 +[2025-04-11 15:09:52] [INFO] Epoch 15/50 - Train Loss: 0.1300, Train Acc: 0.9758, Val Loss: 0.3492, Val Acc: 0.9107 +[2025-04-11 15:09:53] [INFO] 当前学习率: 0.000080 +[2025-04-11 15:09:53] [INFO] Epoch 16/50 开始训练 +[2025-04-11 15:10:12] [INFO] Epoch 16/50 开始验证 +[2025-04-11 15:10:14] [INFO] Epoch 16/50 - Train Loss: 0.1371, Train Acc: 0.9729, Val Loss: 0.4028, Val Acc: 0.9047 +[2025-04-11 15:10:15] [INFO] 当前学习率: 0.000077 +[2025-04-11 15:10:15] [INFO] Epoch 17/50 开始训练 +[2025-04-11 15:10:34] [INFO] Epoch 17/50 开始验证 +[2025-04-11 15:10:36] [INFO] Epoch 17/50 - Train Loss: 0.1029, Train Acc: 0.9796, Val Loss: 0.2834, Val Acc: 0.9268 +[2025-04-11 15:10:37] [INFO] 已保存最佳模型，准确率: 0.9268 +[2025-04-11 15:10:38] [INFO] 当前学习率: 0.000074 +[2025-04-11 15:10:38] [INFO] Epoch 18/50 开始训练 +[2025-04-11 15:10:56] [INFO] Epoch 18/50 开始验证 +[2025-04-11 15:10:58] [INFO] Epoch 18/50 - Train Loss: 0.1109, Train Acc: 0.9777, Val Loss: 0.2368, Val Acc: 0.9268 +[2025-04-11 15:10:59] [INFO] 当前学习率: 0.000072 +[2025-04-11 15:10:59] [INFO] Epoch 19/50 开始训练 +[2025-04-11 15:11:18] [INFO] Epoch 19/50 开始验证 +[2025-04-11 15:11:20] [INFO] Epoch 19/50 - Train Loss: 0.1003, Train Acc: 0.9770, Val Loss: 0.2490, Val Acc: 0.9248 +[2025-04-11 15:11:21] [INFO] 当前学习率: 0.000069 +[2025-04-11 15:11:21] [INFO] Epoch 20/50 开始训练 +[2025-04-11 15:11:40] [INFO] Epoch 20/50 开始验证 +[2025-04-11 15:11:42] [INFO] Epoch 20/50 - Train Loss: 0.0969, Train Acc: 0.9784, Val Loss: 0.2582, Val Acc: 0.9298 +[2025-04-11 15:11:43] [INFO] 已保存最佳模型，准确率: 0.9298 +[2025-04-11 15:11:44] [INFO] 当前学习率: 0.000066 +[2025-04-11 15:11:44] [INFO] Epoch 21/50 开始训练 +[2025-04-11 15:12:02] [INFO] Epoch 21/50 开始验证 +[2025-04-11 15:12:05] [INFO] Epoch 21/50 - Train Loss: 0.0661, Train Acc: 0.9847, Val Loss: 0.2623, Val Acc: 0.9408 +[2025-04-11 15:12:05] [INFO] 已保存最佳模型，准确率: 0.9408 +[2025-04-11 15:12:06] [INFO] 当前学习率: 0.000063 +[2025-04-11 15:12:06] [INFO] Epoch 22/50 开始训练 +[2025-04-11 15:12:25] [INFO] Epoch 22/50 开始验证 +[2025-04-11 15:12:27] [INFO] Epoch 22/50 - Train Loss: 0.0559, Train Acc: 0.9849, Val Loss: 0.2204, Val Acc: 0.9488 +[2025-04-11 15:12:27] [INFO] 已保存最佳模型，准确率: 0.9488 +[2025-04-11 15:12:28] [INFO] 当前学习率: 0.000060 +[2025-04-11 15:12:28] [INFO] Epoch 23/50 开始训练 +[2025-04-11 15:12:47] [INFO] Epoch 23/50 开始验证 +[2025-04-11 15:12:49] [INFO] Epoch 23/50 - Train Loss: 0.0489, Train Acc: 0.9880, Val Loss: 0.2380, Val Acc: 0.9408 +[2025-04-11 15:12:50] [INFO] 当前学习率: 0.000057 +[2025-04-11 15:12:50] [INFO] Epoch 24/50 开始训练 +[2025-04-11 15:13:09] [INFO] Epoch 24/50 开始验证 +[2025-04-11 15:13:11] [INFO] Epoch 24/50 - Train Loss: 0.0533, Train Acc: 0.9892, Val Loss: 0.3553, Val Acc: 0.8927 +[2025-04-11 15:13:12] [INFO] 当前学习率: 0.000054 +[2025-04-11 15:13:12] [INFO] Epoch 25/50 开始训练 +[2025-04-11 15:13:31] [INFO] Epoch 25/50 开始验证 +[2025-04-11 15:13:33] [INFO] Epoch 25/50 - Train Loss: 0.0507, Train Acc: 0.9880, Val Loss: 0.2503, Val Acc: 0.9218 +[2025-04-11 15:13:34] [INFO] 当前学习率: 0.000050 +[2025-04-11 15:13:34] [INFO] Epoch 26/50 开始训练 +[2025-04-11 15:13:52] [INFO] Epoch 26/50 开始验证 +[2025-04-11 15:13:55] [INFO] Epoch 26/50 - Train Loss: 0.0491, Train Acc: 0.9902, Val Loss: 0.2637, Val Acc: 0.9188 +[2025-04-11 15:13:56] [INFO] 当前学习率: 0.000047 +[2025-04-11 15:13:56] [INFO] Epoch 27/50 开始训练 +[2025-04-11 15:14:14] [INFO] Epoch 27/50 开始验证 +[2025-04-11 15:14:17] [INFO] Epoch 27/50 - Train Loss: 0.0369, Train Acc: 0.9904, Val Loss: 0.2795, Val Acc: 0.9127 +[2025-04-11 15:14:18] [INFO] 当前学习率: 0.000044 +[2025-04-11 15:14:18] [INFO] Epoch 28/50 开始训练 +[2025-04-11 15:14:36] [INFO] Epoch 28/50 开始验证 +[2025-04-11 15:14:38] [INFO] Epoch 28/50 - Train Loss: 0.0289, Train Acc: 0.9959, Val Loss: 0.3164, Val Acc: 0.9107 +[2025-04-11 15:14:39] [INFO] 当前学习率: 0.000041 +[2025-04-11 15:14:39] [INFO] Epoch 29/50 开始训练 +[2025-04-11 15:14:57] [INFO] Epoch 29/50 开始验证 +[2025-04-11 15:14:59] [INFO] Epoch 29/50 - Train Loss: 0.0331, Train Acc: 0.9935, Val Loss: 0.3443, Val Acc: 0.9087 +[2025-04-11 15:15:00] [INFO] 当前学习率: 0.000038 +[2025-04-11 15:15:00] [INFO] Epoch 30/50 开始训练 +[2025-04-11 15:15:19] [INFO] Epoch 30/50 开始验证 +[2025-04-11 15:15:21] [INFO] Epoch 30/50 - Train Loss: 0.0241, Train Acc: 0.9971, Val Loss: 0.2360, Val Acc: 0.9448 +[2025-04-11 15:15:22] [INFO] 当前学习率: 0.000035 +[2025-04-11 15:15:22] [INFO] Epoch 31/50 开始训练 +[2025-04-11 15:15:41] [INFO] Epoch 31/50 开始验证 +[2025-04-11 15:15:43] [INFO] Epoch 31/50 - Train Loss: 0.0229, Train Acc: 0.9966, Val Loss: 0.2604, Val Acc: 0.9358 +[2025-04-11 15:15:44] [INFO] 当前学习率: 0.000032 +[2025-04-11 15:15:44] [INFO] Epoch 32/50 开始训练 +[2025-04-11 15:16:03] [INFO] Epoch 32/50 开始验证 +[2025-04-11 15:16:05] [INFO] Epoch 32/50 - Train Loss: 0.0340, Train Acc: 0.9940, Val Loss: 0.2789, Val Acc: 0.9258 +[2025-04-11 15:16:06] [INFO] 当前学习率: 0.000029 +[2025-04-11 15:16:06] [INFO] Epoch 33/50 开始训练 +[2025-04-11 15:16:25] [INFO] Epoch 33/50 开始验证 +[2025-04-11 15:16:27] [INFO] Epoch 33/50 - Train Loss: 0.0209, Train Acc: 0.9978, Val Loss: 0.3414, Val Acc: 0.9198 +[2025-04-11 15:16:28] [INFO] 当前学习率: 0.000027 +[2025-04-11 15:16:28] [INFO] Epoch 34/50 开始训练 +[2025-04-11 15:16:46] [INFO] Epoch 34/50 开始验证 +[2025-04-11 15:16:49] [INFO] Epoch 34/50 - Train Loss: 0.0187, Train Acc: 0.9983, Val Loss: 0.4294, Val Acc: 0.9067 +[2025-04-11 15:16:50] [INFO] 当前学习率: 0.000024 +[2025-04-11 15:16:50] [INFO] Epoch 35/50 开始训练 +[2025-04-11 15:17:08] [INFO] Epoch 35/50 开始验证 +[2025-04-11 15:17:11] [INFO] Epoch 35/50 - Train Loss: 0.0186, Train Acc: 0.9978, Val Loss: 0.3243, Val Acc: 0.9388 +[2025-04-11 15:17:12] [INFO] 当前学习率: 0.000021 +[2025-04-11 15:17:12] [INFO] Epoch 36/50 开始训练 +[2025-04-11 15:17:30] [INFO] Epoch 36/50 开始验证 +[2025-04-11 15:17:32] [INFO] Epoch 36/50 - Train Loss: 0.0122, Train Acc: 0.9993, Val Loss: 0.3431, Val Acc: 0.9278 +[2025-04-11 15:17:33] [INFO] 当前学习率: 0.000019 +[2025-04-11 15:17:33] [INFO] Epoch 37/50 开始训练 +[2025-04-11 15:17:52] [INFO] Epoch 37/50 开始验证 +[2025-04-11 15:17:54] [INFO] Epoch 37/50 - Train Loss: 0.0124, Train Acc: 0.9993, Val Loss: 0.3521, Val Acc: 0.9248 +[2025-04-11 15:17:55] [INFO] 当前学习率: 0.000017 +[2025-04-11 15:17:56] [INFO] Epoch 38/50 开始训练 +[2025-04-11 15:18:14] [INFO] Epoch 38/50 开始验证 +[2025-04-11 15:18:16] [INFO] Epoch 38/50 - Train Loss: 0.0160, Train Acc: 0.9981, Val Loss: 0.3614, Val Acc: 0.9298 +[2025-04-11 15:18:17] [INFO] 当前学习率: 0.000014 +[2025-04-11 15:18:17] [INFO] Epoch 39/50 开始训练 +[2025-04-11 15:18:36] [INFO] Epoch 39/50 开始验证 +[2025-04-11 15:18:38] [INFO] Epoch 39/50 - Train Loss: 0.0139, Train Acc: 0.9986, Val Loss: 0.4236, Val Acc: 0.9107 +[2025-04-11 15:18:39] [INFO] 当前学习率: 0.000012 +[2025-04-11 15:18:39] [INFO] Epoch 40/50 开始训练 +[2025-04-11 15:18:57] [INFO] Epoch 40/50 开始验证 +[2025-04-11 15:19:00] [INFO] Epoch 40/50 - Train Loss: 0.0116, Train Acc: 0.9998, Val Loss: 0.4263, Val Acc: 0.9127 +[2025-04-11 15:19:01] [INFO] 当前学习率: 0.000010 +[2025-04-11 15:19:01] [INFO] Epoch 41/50 开始训练 +[2025-04-11 15:19:19] [INFO] Epoch 41/50 开始验证 +[2025-04-11 15:19:22] [INFO] Epoch 41/50 - Train Loss: 0.0141, Train Acc: 0.9990, Val Loss: 0.4753, Val Acc: 0.9027 +[2025-04-11 15:19:23] [INFO] 当前学习率: 0.000009 +[2025-04-11 15:19:23] [INFO] Epoch 42/50 开始训练 +[2025-04-11 15:19:41] [INFO] Epoch 42/50 开始验证 +[2025-04-11 15:19:44] [INFO] Epoch 42/50 - Train Loss: 0.0128, Train Acc: 0.9995, Val Loss: 0.7317, Val Acc: 0.8857 +[2025-04-11 15:19:45] [INFO] 当前学习率: 0.000007 +[2025-04-11 15:19:45] [INFO] Epoch 43/50 开始训练 +[2025-04-11 15:20:03] [INFO] Epoch 43/50 开始验证 +[2025-04-11 15:20:06] [INFO] Epoch 43/50 - Train Loss: 0.0119, Train Acc: 0.9993, Val Loss: 0.7612, Val Acc: 0.8826 +[2025-04-11 15:20:07] [INFO] 当前学习率: 0.000006 +[2025-04-11 15:20:07] [INFO] Epoch 44/50 开始训练 +[2025-04-11 15:20:25] [INFO] Epoch 44/50 开始验证 +[2025-04-11 15:20:28] [INFO] Epoch 44/50 - Train Loss: 0.0101, Train Acc: 0.9995, Val Loss: 0.4204, Val Acc: 0.9107 +[2025-04-11 15:20:29] [INFO] 当前学习率: 0.000004 +[2025-04-11 15:20:29] [INFO] Epoch 45/50 开始训练 +[2025-04-11 15:20:47] [INFO] Epoch 45/50 开始验证 +[2025-04-11 15:20:49] [INFO] Epoch 45/50 - Train Loss: 0.0126, Train Acc: 1.0000, Val Loss: 0.4008, Val Acc: 0.9137 +[2025-04-11 15:20:50] [INFO] 当前学习率: 0.000003 +[2025-04-11 15:20:50] [INFO] Epoch 46/50 开始训练 +[2025-04-11 15:21:09] [INFO] Epoch 46/50 开始验证 +[2025-04-11 15:21:11] [INFO] Epoch 46/50 - Train Loss: 0.0120, Train Acc: 0.9988, Val Loss: 0.4120, Val Acc: 0.9178 +[2025-04-11 15:21:12] [INFO] 当前学习率: 0.000003 +[2025-04-11 15:21:12] [INFO] Epoch 47/50 开始训练 +[2025-04-11 15:21:31] [INFO] Epoch 47/50 开始验证 +[2025-04-11 15:21:33] [INFO] Epoch 47/50 - Train Loss: 0.0088, Train Acc: 1.0000, Val Loss: 0.4181, Val Acc: 0.9157 +[2025-04-11 15:21:34] [INFO] 当前学习率: 0.000002 +[2025-04-11 15:21:34] [INFO] Epoch 48/50 开始训练 +[2025-04-11 15:21:53] [INFO] Epoch 48/50 开始验证 +[2025-04-11 15:21:55] [INFO] Epoch 48/50 - Train Loss: 0.0114, Train Acc: 0.9995, Val Loss: 0.4700, Val Acc: 0.9057 +[2025-04-11 15:21:56] [INFO] 当前学习率: 0.000001 +[2025-04-11 15:21:56] [INFO] Epoch 49/50 开始训练 +[2025-04-11 15:22:15] [INFO] Epoch 49/50 开始验证 +[2025-04-11 15:22:17] [INFO] Epoch 49/50 - Train Loss: 0.0107, Train Acc: 0.9993, Val Loss: 0.4309, Val Acc: 0.9137 +[2025-04-11 15:22:18] [INFO] 当前学习率: 0.000001 +[2025-04-11 15:22:18] [INFO] Epoch 50/50 开始训练 +[2025-04-11 15:22:37] [INFO] Epoch 50/50 开始验证 +[2025-04-11 15:22:39] [INFO] Epoch 50/50 - Train Loss: 0.0122, Train Acc: 0.9990, Val Loss: 0.4090, Val Acc: 0.9087 +[2025-04-11 15:22:40] [INFO] 绘制训练过程图表 +[2025-04-11 15:22:41] [INFO] 标准训练完成! +[2025-04-11 15:22:41] [INFO] 评估原始模型性能... +[2025-04-11 15:22:43] [INFO] 评估结果 - Loss: 0.4090, Accuracy: 0.9087 +[2025-04-11 15:22:43] [INFO] 开始执行RRAM映射... +[2025-04-11 15:22:43] [INFO] 加载了 100 个RRAM电导值 +[2025-04-11 15:22:43] [INFO] features.0.0.weight 的平均映射误差: 0.018905 +[2025-04-11 15:22:43] [INFO] features.0.1.weight 的平均映射误差: 0.031780 +[2025-04-11 15:22:43] [INFO] features.1.0.block.0.0.weight 的平均映射误差: 0.005872 +[2025-04-11 15:22:43] [INFO] features.1.0.block.0.1.weight 的平均映射误差: 0.033922 +[2025-04-11 15:22:43] [INFO] features.1.1.block.0.0.weight 的平均映射误差: 0.004029 +[2025-04-11 15:22:43] [INFO] features.1.1.block.0.1.weight 的平均映射误差: 0.032434 +[2025-04-11 15:22:43] [INFO] features.1.2.block.0.0.weight 的平均映射误差: 0.003632 +[2025-04-11 15:22:43] [INFO] features.1.2.block.0.1.weight 的平均映射误差: 0.033747 +[2025-04-11 15:22:43] [INFO] features.2.0.block.0.0.weight 的平均映射误差: 0.003266 +[2025-04-11 15:22:43] [INFO] features.2.0.block.0.1.weight 的平均映射误差: 0.032917 +[2025-04-11 15:22:43] [INFO] features.2.0.block.1.0.weight 的平均映射误差: 0.006467 +[2025-04-11 15:22:43] [INFO] features.2.0.block.1.1.weight 的平均映射误差: 0.033586 +[2025-04-11 15:22:43] [INFO] features.2.1.block.0.0.weight 的平均映射误差: 0.001786 +[2025-04-11 15:22:43] [INFO] features.2.1.block.0.1.weight 的平均映射误差: 0.033159 +[2025-04-11 15:22:43] [INFO] features.2.1.block.1.0.weight 的平均映射误差: 0.003039 +[2025-04-11 15:22:43] [INFO] features.2.1.block.1.1.weight 的平均映射误差: 0.034852 +[2025-04-11 15:22:43] [INFO] features.2.2.block.0.0.weight 的平均映射误差: 0.001770 +[2025-04-11 15:22:43] [INFO] features.2.2.block.0.1.weight 的平均映射误差: 0.033568 +[2025-04-11 15:22:43] [INFO] features.2.2.block.1.0.weight 的平均映射误差: 0.002761 +[2025-04-11 15:22:43] [INFO] features.2.2.block.1.1.weight 的平均映射误差: 0.032742 +[2025-04-11 15:22:43] [INFO] features.2.3.block.0.0.weight 的平均映射误差: 0.001789 +[2025-04-11 15:22:43] [INFO] features.2.3.block.0.1.weight 的平均映射误差: 0.034785 +[2025-04-11 15:22:43] [INFO] features.2.3.block.1.0.weight 的平均映射误差: 0.002686 +[2025-04-11 15:22:43] [INFO] features.2.3.block.1.1.weight 的平均映射误差: 0.031939 +[2025-04-11 15:22:43] [INFO] features.2.4.block.0.0.weight 的平均映射误差: 0.001811 +[2025-04-11 15:22:43] [INFO] features.2.4.block.0.1.weight 的平均映射误差: 0.037460 +[2025-04-11 15:22:43] [INFO] features.2.4.block.1.0.weight 的平均映射误差: 0.002625 +[2025-04-11 15:22:43] [INFO] features.2.4.block.1.1.weight 的平均映射误差: 0.034390 +[2025-04-11 15:22:43] [INFO] features.3.0.block.0.0.weight 的平均映射误差: 0.002088 +[2025-04-11 15:22:43] [INFO] features.3.0.block.0.1.weight 的平均映射误差: 0.032756 +[2025-04-11 15:22:43] [INFO] features.3.0.block.1.0.weight 的平均映射误差: 0.003913 +[2025-04-11 15:22:43] [INFO] features.3.0.block.1.1.weight 的平均映射误差: 0.034175 +[2025-04-11 15:22:43] [INFO] features.3.1.block.0.0.weight 的平均映射误差: 0.001622 +[2025-04-11 15:22:43] [INFO] features.3.1.block.0.1.weight 的平均映射误差: 0.036317 +[2025-04-11 15:22:43] [INFO] features.3.1.block.1.0.weight 的平均映射误差: 0.002014 +[2025-04-11 15:22:43] [INFO] features.3.1.block.1.1.weight 的平均映射误差: 0.034580 +[2025-04-11 15:22:43] [INFO] features.3.2.block.0.0.weight 的平均映射误差: 0.001615 +[2025-04-11 15:22:43] [INFO] features.3.2.block.0.1.weight 的平均映射误差: 0.045919 +[2025-04-11 15:22:43] [INFO] features.3.2.block.1.0.weight 的平均映射误差: 0.001930 +[2025-04-11 15:22:43] [INFO] features.3.2.block.1.1.weight 的平均映射误差: 0.032456 +[2025-04-11 15:22:43] [INFO] features.3.3.block.0.0.weight 的平均映射误差: 0.001619 +[2025-04-11 15:22:43] [INFO] features.3.3.block.0.1.weight 的平均映射误差: 0.048257 +[2025-04-11 15:22:43] [INFO] features.3.3.block.1.0.weight 的平均映射误差: 0.001930 +[2025-04-11 15:22:43] [INFO] features.3.3.block.1.1.weight 的平均映射误差: 0.035222 +[2025-04-11 15:22:43] [INFO] features.3.4.block.0.0.weight 的平均映射误差: 0.001612 +[2025-04-11 15:22:43] [INFO] features.3.4.block.0.1.weight 的平均映射误差: 0.039553 +[2025-04-11 15:22:43] [INFO] features.3.4.block.1.0.weight 的平均映射误差: 0.001844 +[2025-04-11 15:22:43] [INFO] features.3.4.block.1.1.weight 的平均映射误差: 0.032761 +[2025-04-11 15:22:43] [INFO] features.4.0.block.0.0.weight 的平均映射误差: 0.003869 +[2025-04-11 15:22:43] [INFO] features.4.0.block.0.1.weight 的平均映射误差: 0.040167 +[2025-04-11 15:22:43] [INFO] features.4.0.block.1.0.weight 的平均映射误差: 0.004816 +[2025-04-11 15:22:43] [INFO] features.4.0.block.1.1.weight 的平均映射误差: 0.047085 +[2025-04-11 15:22:43] [INFO] features.4.0.block.2.fc1.weight 的平均映射误差: 0.001470 +[2025-04-11 15:22:43] [INFO] features.4.0.block.2.fc2.weight 的平均映射误差: 0.001572 +[2025-04-11 15:22:43] [INFO] features.4.0.block.3.0.weight 的平均映射误差: 0.002932 +[2025-04-11 15:22:43] [INFO] features.4.0.block.3.1.weight 的平均映射误差: 0.034668 +[2025-04-11 15:22:43] [INFO] features.4.1.block.0.0.weight 的平均映射误差: 0.001683 +[2025-04-11 15:22:43] [INFO] features.4.1.block.0.1.weight 的平均映射误差: 0.034876 +[2025-04-11 15:22:43] [INFO] features.4.1.block.1.0.weight 的平均映射误差: 0.002791 +[2025-04-11 15:22:43] [INFO] features.4.1.block.1.1.weight 的平均映射误差: 0.034487 +[2025-04-11 15:22:43] [INFO] features.4.1.block.2.fc1.weight 的平均映射误差: 0.001438 +[2025-04-11 15:22:43] [INFO] features.4.1.block.2.fc2.weight 的平均映射误差: 0.002088 +[2025-04-11 15:22:43] [INFO] features.4.1.block.3.0.weight 的平均映射误差: 0.001680 +[2025-04-11 15:22:43] [INFO] features.4.1.block.3.1.weight 的平均映射误差: 0.036592 +[2025-04-11 15:22:43] [INFO] features.4.2.block.0.0.weight 的平均映射误差: 0.001685 +[2025-04-11 15:22:43] [INFO] features.4.2.block.0.1.weight 的平均映射误差: 0.036517 +[2025-04-11 15:22:43] [INFO] features.4.2.block.1.0.weight 的平均映射误差: 0.002741 +[2025-04-11 15:22:43] [INFO] features.4.2.block.1.1.weight 的平均映射误差: 0.034851 +[2025-04-11 15:22:43] [INFO] features.4.2.block.2.fc1.weight 的平均映射误差: 0.001254 +[2025-04-11 15:22:43] [INFO] features.4.2.block.2.fc2.weight 的平均映射误差: 0.001928 +[2025-04-11 15:22:43] [INFO] features.4.2.block.3.0.weight 的平均映射误差: 0.001649 +[2025-04-11 15:22:43] [INFO] features.4.2.block.3.1.weight 的平均映射误差: 0.034371 +[2025-04-11 15:22:43] [INFO] features.4.3.block.0.0.weight 的平均映射误差: 0.001666 +[2025-04-11 15:22:43] [INFO] features.4.3.block.0.1.weight 的平均映射误差: 0.035329 +[2025-04-11 15:22:43] [INFO] features.4.3.block.1.0.weight 的平均映射误差: 0.002622 +[2025-04-11 15:22:43] [INFO] features.4.3.block.1.1.weight 的平均映射误差: 0.035117 +[2025-04-11 15:22:43] [INFO] features.4.3.block.2.fc1.weight 的平均映射误差: 0.000960 +[2025-04-11 15:22:43] [INFO] features.4.3.block.2.fc2.weight 的平均映射误差: 0.001500 +[2025-04-11 15:22:43] [INFO] features.4.3.block.3.0.weight 的平均映射误差: 0.001638 +[2025-04-11 15:22:43] [INFO] features.4.3.block.3.1.weight 的平均映射误差: 0.035966 +[2025-04-11 15:22:43] [INFO] features.4.4.block.0.0.weight 的平均映射误差: 0.001661 +[2025-04-11 15:22:43] [INFO] features.4.4.block.0.1.weight 的平均映射误差: 0.036042 +[2025-04-11 15:22:43] [INFO] features.4.4.block.1.0.weight 的平均映射误差: 0.002573 +[2025-04-11 15:22:43] [INFO] features.4.4.block.1.1.weight 的平均映射误差: 0.034348 +[2025-04-11 15:22:43] [INFO] features.4.4.block.2.fc1.weight 的平均映射误差: 0.000826 +[2025-04-11 15:22:43] [INFO] features.4.4.block.2.fc2.weight 的平均映射误差: 0.001089 +[2025-04-11 15:22:43] [INFO] features.4.4.block.3.0.weight 的平均映射误差: 0.001627 +[2025-04-11 15:22:43] [INFO] features.4.4.block.3.1.weight 的平均映射误差: 0.037581 +[2025-04-11 15:22:43] [INFO] features.4.5.block.0.0.weight 的平均映射误差: 0.001658 +[2025-04-11 15:22:43] [INFO] features.4.5.block.0.1.weight 的平均映射误差: 0.037093 +[2025-04-11 15:22:43] [INFO] features.4.5.block.1.0.weight 的平均映射误差: 0.002317 +[2025-04-11 15:22:43] [INFO] features.4.5.block.1.1.weight 的平均映射误差: 0.034326 +[2025-04-11 15:22:43] [INFO] features.4.5.block.2.fc1.weight 的平均映射误差: 0.001352 +[2025-04-11 15:22:43] [INFO] features.4.5.block.2.fc2.weight 的平均映射误差: 0.001066 +[2025-04-11 15:22:43] [INFO] features.4.5.block.3.0.weight 的平均映射误差: 0.001631 +[2025-04-11 15:22:43] [INFO] features.4.5.block.3.1.weight 的平均映射误差: 0.036209 +[2025-04-11 15:22:43] [INFO] features.4.6.block.0.0.weight 的平均映射误差: 0.001665 +[2025-04-11 15:22:43] [INFO] features.4.6.block.0.1.weight 的平均映射误差: 0.036320 +[2025-04-11 15:22:43] [INFO] features.4.6.block.1.0.weight 的平均映射误差: 0.002217 +[2025-04-11 15:22:43] [INFO] features.4.6.block.1.1.weight 的平均映射误差: 0.041461 +[2025-04-11 15:22:43] [INFO] features.4.6.block.2.fc1.weight 的平均映射误差: 0.000861 +[2025-04-11 15:22:43] [INFO] features.4.6.block.2.fc2.weight 的平均映射误差: 0.001161 +[2025-04-11 15:22:43] [INFO] features.4.6.block.3.0.weight 的平均映射误差: 0.001622 +[2025-04-11 15:22:43] [INFO] features.4.6.block.3.1.weight 的平均映射误差: 0.035874 +[2025-04-11 15:22:43] [INFO] features.5.0.block.0.0.weight 的平均映射误差: 0.002147 +[2025-04-11 15:22:43] [INFO] features.5.0.block.0.1.weight 的平均映射误差: 0.034739 +[2025-04-11 15:22:43] [INFO] features.5.0.block.1.0.weight 的平均映射误差: 0.003584 +[2025-04-11 15:22:43] [INFO] features.5.0.block.1.1.weight 的平均映射误差: 0.038090 +[2025-04-11 15:22:43] [INFO] features.5.0.block.2.fc1.weight 的平均映射误差: 0.001862 +[2025-04-11 15:22:43] [INFO] features.5.0.block.2.fc2.weight 的平均映射误差: 0.002018 +[2025-04-11 15:22:43] [INFO] features.5.0.block.3.0.weight 的平均映射误差: 0.001985 +[2025-04-11 15:22:43] [INFO] features.5.0.block.3.1.weight 的平均映射误差: 0.034421 +[2025-04-11 15:22:43] [INFO] features.5.1.block.0.0.weight 的平均映射误差: 0.001631 +[2025-04-11 15:22:43] [INFO] features.5.1.block.0.1.weight 的平均映射误差: 0.037801 +[2025-04-11 15:22:43] [INFO] features.5.1.block.1.0.weight 的平均映射误差: 0.002205 +[2025-04-11 15:22:43] [INFO] features.5.1.block.1.1.weight 的平均映射误差: 0.038944 +[2025-04-11 15:22:43] [INFO] features.5.1.block.2.fc1.weight 的平均映射误差: 0.000995 +[2025-04-11 15:22:43] [INFO] features.5.1.block.2.fc2.weight 的平均映射误差: 0.001878 +[2025-04-11 15:22:43] [INFO] features.5.1.block.3.0.weight 的平均映射误差: 0.001603 +[2025-04-11 15:22:43] [INFO] features.5.1.block.3.1.weight 的平均映射误差: 0.041624 +[2025-04-11 15:22:43] [INFO] features.5.2.block.0.0.weight 的平均映射误差: 0.001614 +[2025-04-11 15:22:43] [INFO] features.5.2.block.0.1.weight 的平均映射误差: 0.036024 +[2025-04-11 15:22:43] [INFO] features.5.2.block.1.0.weight 的平均映射误差: 0.002093 +[2025-04-11 15:22:43] [INFO] features.5.2.block.1.1.weight 的平均映射误差: 0.038020 +[2025-04-11 15:22:43] [INFO] features.5.2.block.2.fc1.weight 的平均映射误差: 0.001059 +[2025-04-11 15:22:43] [INFO] features.5.2.block.2.fc2.weight 的平均映射误差: 0.001676 +[2025-04-11 15:22:43] [INFO] features.5.2.block.3.0.weight 的平均映射误差: 0.001597 +[2025-04-11 15:22:43] [INFO] features.5.2.block.3.1.weight 的平均映射误差: 0.034641 +[2025-04-11 15:22:43] [INFO] features.5.3.block.0.0.weight 的平均映射误差: 0.001602 +[2025-04-11 15:22:43] [INFO] features.5.3.block.0.1.weight 的平均映射误差: 0.037186 +[2025-04-11 15:22:43] [INFO] features.5.3.block.1.0.weight 的平均映射误差: 0.001999 +[2025-04-11 15:22:43] [INFO] features.5.3.block.1.1.weight 的平均映射误差: 0.039849 +[2025-04-11 15:22:43] [INFO] features.5.3.block.2.fc1.weight 的平均映射误差: 0.000896 +[2025-04-11 15:22:43] [INFO] features.5.3.block.2.fc2.weight 的平均映射误差: 0.001413 +[2025-04-11 15:22:43] [INFO] features.5.3.block.3.0.weight 的平均映射误差: 0.001572 +[2025-04-11 15:22:43] [INFO] features.5.3.block.3.1.weight 的平均映射误差: 0.033534 +[2025-04-11 15:22:43] [INFO] features.5.4.block.0.0.weight 的平均映射误差: 0.001617 +[2025-04-11 15:22:43] [INFO] features.5.4.block.0.1.weight 的平均映射误差: 0.037505 +[2025-04-11 15:22:43] [INFO] features.5.4.block.1.0.weight 的平均映射误差: 0.002018 +[2025-04-11 15:22:43] [INFO] features.5.4.block.1.1.weight 的平均映射误差: 0.041453 +[2025-04-11 15:22:43] [INFO] features.5.4.block.2.fc1.weight 的平均映射误差: 0.000889 +[2025-04-11 15:22:43] [INFO] features.5.4.block.2.fc2.weight 的平均映射误差: 0.001361 +[2025-04-11 15:22:43] [INFO] features.5.4.block.3.0.weight 的平均映射误差: 0.001573 +[2025-04-11 15:22:43] [INFO] features.5.4.block.3.1.weight 的平均映射误差: 0.032919 +[2025-04-11 15:22:43] [INFO] features.5.5.block.0.0.weight 的平均映射误差: 0.001618 +[2025-04-11 15:22:43] [INFO] features.5.5.block.0.1.weight 的平均映射误差: 0.036770 +[2025-04-11 15:22:43] [INFO] features.5.5.block.1.0.weight 的平均映射误差: 0.001944 +[2025-04-11 15:22:43] [INFO] features.5.5.block.1.1.weight 的平均映射误差: 0.041871 +[2025-04-11 15:22:43] [INFO] features.5.5.block.2.fc1.weight 的平均映射误差: 0.000864 +[2025-04-11 15:22:43] [INFO] features.5.5.block.2.fc2.weight 的平均映射误差: 0.001136 +[2025-04-11 15:22:43] [INFO] features.5.5.block.3.0.weight 的平均映射误差: 0.001548 +[2025-04-11 15:22:43] [INFO] features.5.5.block.3.1.weight 的平均映射误差: 0.030748 +[2025-04-11 15:22:43] [INFO] features.5.6.block.0.0.weight 的平均映射误差: 0.001610 +[2025-04-11 15:22:43] [INFO] features.5.6.block.0.1.weight 的平均映射误差: 0.036023 +[2025-04-11 15:22:43] [INFO] features.5.6.block.1.0.weight 的平均映射误差: 0.001823 +[2025-04-11 15:22:43] [INFO] features.5.6.block.1.1.weight 的平均映射误差: 0.042716 +[2025-04-11 15:22:43] [INFO] features.5.6.block.2.fc1.weight 的平均映射误差: 0.001042 +[2025-04-11 15:22:43] [INFO] features.5.6.block.2.fc2.weight 的平均映射误差: 0.001352 +[2025-04-11 15:22:43] [INFO] features.5.6.block.3.0.weight 的平均映射误差: 0.001538 +[2025-04-11 15:22:43] [INFO] features.5.6.block.3.1.weight 的平均映射误差: 0.032176 +[2025-04-11 15:22:43] [INFO] features.5.7.block.0.0.weight 的平均映射误差: 0.001594 +[2025-04-11 15:22:43] [INFO] features.5.7.block.0.1.weight 的平均映射误差: 0.036011 +[2025-04-11 15:22:43] [INFO] features.5.7.block.1.0.weight 的平均映射误差: 0.001902 +[2025-04-11 15:22:43] [INFO] features.5.7.block.1.1.weight 的平均映射误差: 0.043315 +[2025-04-11 15:22:43] [INFO] features.5.7.block.2.fc1.weight 的平均映射误差: 0.000777 +[2025-04-11 15:22:43] [INFO] features.5.7.block.2.fc2.weight 的平均映射误差: 0.000951 +[2025-04-11 15:22:43] [INFO] features.5.7.block.3.0.weight 的平均映射误差: 0.001511 +[2025-04-11 15:22:43] [INFO] features.5.7.block.3.1.weight 的平均映射误差: 0.030084 +[2025-04-11 15:22:43] [INFO] features.5.8.block.0.0.weight 的平均映射误差: 0.001581 +[2025-04-11 15:22:43] [INFO] features.5.8.block.0.1.weight 的平均映射误差: 0.035677 +[2025-04-11 15:22:43] [INFO] features.5.8.block.1.0.weight 的平均映射误差: 0.001825 +[2025-04-11 15:22:43] [INFO] features.5.8.block.1.1.weight 的平均映射误差: 0.043251 +[2025-04-11 15:22:43] [INFO] features.5.8.block.2.fc1.weight 的平均映射误差: 0.000841 +[2025-04-11 15:22:43] [INFO] features.5.8.block.2.fc2.weight 的平均映射误差: 0.001042 +[2025-04-11 15:22:43] [INFO] features.5.8.block.3.0.weight 的平均映射误差: 0.001534 +[2025-04-11 15:22:43] [INFO] features.5.8.block.3.1.weight 的平均映射误差: 0.031754 +[2025-04-11 15:22:43] [INFO] features.5.9.block.0.0.weight 的平均映射误差: 0.001591 +[2025-04-11 15:22:43] [INFO] features.5.9.block.0.1.weight 的平均映射误差: 0.035097 +[2025-04-11 15:22:43] [INFO] features.5.9.block.1.0.weight 的平均映射误差: 0.001813 +[2025-04-11 15:22:43] [INFO] features.5.9.block.1.1.weight 的平均映射误差: 0.044321 +[2025-04-11 15:22:43] [INFO] features.5.9.block.2.fc1.weight 的平均映射误差: 0.000888 +[2025-04-11 15:22:43] [INFO] features.5.9.block.2.fc2.weight 的平均映射误差: 0.001103 +[2025-04-11 15:22:43] [INFO] features.5.9.block.3.0.weight 的平均映射误差: 0.001522 +[2025-04-11 15:22:43] [INFO] features.5.9.block.3.1.weight 的平均映射误差: 0.030649 +[2025-04-11 15:22:43] [INFO] features.5.10.block.0.0.weight 的平均映射误差: 0.001591 +[2025-04-11 15:22:43] [INFO] features.5.10.block.0.1.weight 的平均映射误差: 0.034323 +[2025-04-11 15:22:43] [INFO] features.5.10.block.1.0.weight 的平均映射误差: 0.001863 +[2025-04-11 15:22:43] [INFO] features.5.10.block.1.1.weight 的平均映射误差: 0.043093 +[2025-04-11 15:22:43] [INFO] features.5.10.block.2.fc1.weight 的平均映射误差: 0.001425 +[2025-04-11 15:22:43] [INFO] features.5.10.block.2.fc2.weight 的平均映射误差: 0.001068 +[2025-04-11 15:22:43] [INFO] features.5.10.block.3.0.weight 的平均映射误差: 0.001566 +[2025-04-11 15:22:43] [INFO] features.5.10.block.3.1.weight 的平均映射误差: 0.038591 +[2025-04-11 15:22:43] [INFO] features.5.11.block.0.0.weight 的平均映射误差: 0.001607 +[2025-04-11 15:22:43] [INFO] features.5.11.block.0.1.weight 的平均映射误差: 0.035258 +[2025-04-11 15:22:43] [INFO] features.5.11.block.1.0.weight 的平均映射误差: 0.001852 +[2025-04-11 15:22:43] [INFO] features.5.11.block.1.1.weight 的平均映射误差: 0.045503 +[2025-04-11 15:22:43] [INFO] features.5.11.block.2.fc1.weight 的平均映射误差: 0.000805 +[2025-04-11 15:22:43] [INFO] features.5.11.block.2.fc2.weight 的平均映射误差: 0.000946 +[2025-04-11 15:22:43] [INFO] features.5.11.block.3.0.weight 的平均映射误差: 0.001570 +[2025-04-11 15:22:43] [INFO] features.5.11.block.3.1.weight 的平均映射误差: 0.038366 +[2025-04-11 15:22:43] [INFO] features.5.12.block.0.0.weight 的平均映射误差: 0.001592 +[2025-04-11 15:22:43] [INFO] features.5.12.block.0.1.weight 的平均映射误差: 0.035148 +[2025-04-11 15:22:43] [INFO] features.5.12.block.1.0.weight 的平均映射误差: 0.001816 +[2025-04-11 15:22:43] [INFO] features.5.12.block.1.1.weight 的平均映射误差: 0.046690 +[2025-04-11 15:22:43] [INFO] features.5.12.block.2.fc1.weight 的平均映射误差: 0.000782 +[2025-04-11 15:22:43] [INFO] features.5.12.block.2.fc2.weight 的平均映射误差: 0.000999 +[2025-04-11 15:22:43] [INFO] features.5.12.block.3.0.weight 的平均映射误差: 0.001560 +[2025-04-11 15:22:43] [INFO] features.5.12.block.3.1.weight 的平均映射误差: 0.039560 +[2025-04-11 15:22:43] [INFO] features.5.13.block.0.0.weight 的平均映射误差: 0.001599 +[2025-04-11 15:22:43] [INFO] features.5.13.block.0.1.weight 的平均映射误差: 0.034791 +[2025-04-11 15:22:43] [INFO] features.5.13.block.1.0.weight 的平均映射误差: 0.001807 +[2025-04-11 15:22:43] [INFO] features.5.13.block.1.1.weight 的平均映射误差: 0.046048 +[2025-04-11 15:22:43] [INFO] features.5.13.block.2.fc1.weight 的平均映射误差: 0.000815 +[2025-04-11 15:22:43] [INFO] features.5.13.block.2.fc2.weight 的平均映射误差: 0.000966 +[2025-04-11 15:22:43] [INFO] features.5.13.block.3.0.weight 的平均映射误差: 0.001565 +[2025-04-11 15:22:43] [INFO] features.5.13.block.3.1.weight 的平均映射误差: 0.041566 +[2025-04-11 15:22:43] [INFO] features.6.0.block.0.0.weight 的平均映射误差: 0.002128 +[2025-04-11 15:22:43] [INFO] features.6.0.block.0.1.weight 的平均映射误差: 0.038207 +[2025-04-11 15:22:43] [INFO] features.6.0.block.1.0.weight 的平均映射误差: 0.003156 +[2025-04-11 15:22:43] [INFO] features.6.0.block.1.1.weight 的平均映射误差: 0.039394 +[2025-04-11 15:22:43] [INFO] features.6.0.block.2.fc1.weight 的平均映射误差: 0.000644 +[2025-04-11 15:22:43] [INFO] features.6.0.block.2.fc2.weight 的平均映射误差: 0.001260 +[2025-04-11 15:22:43] [INFO] features.6.0.block.3.0.weight 的平均映射误差: 0.001897 +[2025-04-11 15:22:43] [INFO] features.6.0.block.3.1.weight 的平均映射误差: 0.034737 +[2025-04-11 15:22:43] [INFO] features.6.1.block.0.0.weight 的平均映射误差: 0.001576 +[2025-04-11 15:22:43] [INFO] features.6.1.block.0.1.weight 的平均映射误差: 0.038259 +[2025-04-11 15:22:43] [INFO] features.6.1.block.1.0.weight 的平均映射误差: 0.001995 +[2025-04-11 15:22:43] [INFO] features.6.1.block.1.1.weight 的平均映射误差: 0.038364 +[2025-04-11 15:22:43] [INFO] features.6.1.block.2.fc1.weight 的平均映射误差: 0.000814 +[2025-04-11 15:22:43] [INFO] features.6.1.block.2.fc2.weight 的平均映射误差: 0.001554 +[2025-04-11 15:22:43] [INFO] features.6.1.block.3.0.weight 的平均映射误差: 0.001574 +[2025-04-11 15:22:43] [INFO] features.6.1.block.3.1.weight 的平均映射误差: 0.047152 +[2025-04-11 15:22:43] [INFO] features.6.2.block.0.0.weight 的平均映射误差: 0.001573 +[2025-04-11 15:22:43] [INFO] features.6.2.block.0.1.weight 的平均映射误差: 0.038126 +[2025-04-11 15:22:43] [INFO] features.6.2.block.1.0.weight 的平均映射误差: 0.001996 +[2025-04-11 15:22:43] [INFO] features.6.2.block.1.1.weight 的平均映射误差: 0.039445 +[2025-04-11 15:22:43] [INFO] features.6.2.block.2.fc1.weight 的平均映射误差: 0.000905 +[2025-04-11 15:22:43] [INFO] features.6.2.block.2.fc2.weight 的平均映射误差: 0.001438 +[2025-04-11 15:22:43] [INFO] features.6.2.block.3.0.weight 的平均映射误差: 0.001564 +[2025-04-11 15:22:43] [INFO] features.6.2.block.3.1.weight 的平均映射误差: 0.044286 +[2025-04-11 15:22:43] [INFO] features.6.3.block.0.0.weight 的平均映射误差: 0.001560 +[2025-04-11 15:22:43] [INFO] features.6.3.block.0.1.weight 的平均映射误差: 0.038106 +[2025-04-11 15:22:43] [INFO] features.6.3.block.1.0.weight 的平均映射误差: 0.001935 +[2025-04-11 15:22:43] [INFO] features.6.3.block.1.1.weight 的平均映射误差: 0.043263 +[2025-04-11 15:22:43] [INFO] features.6.3.block.2.fc1.weight 的平均映射误差: 0.000859 +[2025-04-11 15:22:43] [INFO] features.6.3.block.2.fc2.weight 的平均映射误差: 0.001293 +[2025-04-11 15:22:43] [INFO] features.6.3.block.3.0.weight 的平均映射误差: 0.001541 +[2025-04-11 15:22:43] [INFO] features.6.3.block.3.1.weight 的平均映射误差: 0.044877 +[2025-04-11 15:22:43] [INFO] features.6.4.block.0.0.weight 的平均映射误差: 0.001566 +[2025-04-11 15:22:43] [INFO] features.6.4.block.0.1.weight 的平均映射误差: 0.037219 +[2025-04-11 15:22:43] [INFO] features.6.4.block.1.0.weight 的平均映射误差: 0.001892 +[2025-04-11 15:22:43] [INFO] features.6.4.block.1.1.weight 的平均映射误差: 0.044145 +[2025-04-11 15:22:43] [INFO] features.6.4.block.2.fc1.weight 的平均映射误差: 0.000968 +[2025-04-11 15:22:43] [INFO] features.6.4.block.2.fc2.weight 的平均映射误差: 0.001300 +[2025-04-11 15:22:43] [INFO] features.6.4.block.3.0.weight 的平均映射误差: 0.001538 +[2025-04-11 15:22:43] [INFO] features.6.4.block.3.1.weight 的平均映射误差: 0.044810 +[2025-04-11 15:22:43] [INFO] features.6.5.block.0.0.weight 的平均映射误差: 0.001566 +[2025-04-11 15:22:43] [INFO] features.6.5.block.0.1.weight 的平均映射误差: 0.038088 +[2025-04-11 15:22:43] [INFO] features.6.5.block.1.0.weight 的平均映射误差: 0.001856 +[2025-04-11 15:22:43] [INFO] features.6.5.block.1.1.weight 的平均映射误差: 0.044284 +[2025-04-11 15:22:43] [INFO] features.6.5.block.2.fc1.weight 的平均映射误差: 0.001037 +[2025-04-11 15:22:43] [INFO] features.6.5.block.2.fc2.weight 的平均映射误差: 0.001247 +[2025-04-11 15:22:43] [INFO] features.6.5.block.3.0.weight 的平均映射误差: 0.001545 +[2025-04-11 15:22:43] [INFO] features.6.5.block.3.1.weight 的平均映射误差: 0.043557 +[2025-04-11 15:22:43] [INFO] features.6.6.block.0.0.weight 的平均映射误差: 0.001558 +[2025-04-11 15:22:43] [INFO] features.6.6.block.0.1.weight 的平均映射误差: 0.036510 +[2025-04-11 15:22:43] [INFO] features.6.6.block.1.0.weight 的平均映射误差: 0.001854 +[2025-04-11 15:22:43] [INFO] features.6.6.block.1.1.weight 的平均映射误差: 0.046440 +[2025-04-11 15:22:43] [INFO] features.6.6.block.2.fc1.weight 的平均映射误差: 0.000939 +[2025-04-11 15:22:43] [INFO] features.6.6.block.2.fc2.weight 的平均映射误差: 0.001142 +[2025-04-11 15:22:43] [INFO] features.6.6.block.3.0.weight 的平均映射误差: 0.001527 +[2025-04-11 15:22:43] [INFO] features.6.6.block.3.1.weight 的平均映射误差: 0.040909 +[2025-04-11 15:22:43] [INFO] features.6.7.block.0.0.weight 的平均映射误差: 0.001561 +[2025-04-11 15:22:43] [INFO] features.6.7.block.0.1.weight 的平均映射误差: 0.038060 +[2025-04-11 15:22:43] [INFO] features.6.7.block.1.0.weight 的平均映射误差: 0.001828 +[2025-04-11 15:22:43] [INFO] features.6.7.block.1.1.weight 的平均映射误差: 0.047162 +[2025-04-11 15:22:43] [INFO] features.6.7.block.2.fc1.weight 的平均映射误差: 0.000894 +[2025-04-11 15:22:43] [INFO] features.6.7.block.2.fc2.weight 的平均映射误差: 0.001232 +[2025-04-11 15:22:43] [INFO] features.6.7.block.3.0.weight 的平均映射误差: 0.001537 +[2025-04-11 15:22:43] [INFO] features.6.7.block.3.1.weight 的平均映射误差: 0.044196 +[2025-04-11 15:22:43] [INFO] features.6.8.block.0.0.weight 的平均映射误差: 0.001570 +[2025-04-11 15:22:43] [INFO] features.6.8.block.0.1.weight 的平均映射误差: 0.038231 +[2025-04-11 15:22:43] [INFO] features.6.8.block.1.0.weight 的平均映射误差: 0.001820 +[2025-04-11 15:22:43] [INFO] features.6.8.block.1.1.weight 的平均映射误差: 0.046868 +[2025-04-11 15:22:43] [INFO] features.6.8.block.2.fc1.weight 的平均映射误差: 0.000841 +[2025-04-11 15:22:43] [INFO] features.6.8.block.2.fc2.weight 的平均映射误差: 0.001083 +[2025-04-11 15:22:43] [INFO] features.6.8.block.3.0.weight 的平均映射误差: 0.001541 +[2025-04-11 15:22:43] [INFO] features.6.8.block.3.1.weight 的平均映射误差: 0.043348 +[2025-04-11 15:22:43] [INFO] features.6.9.block.0.0.weight 的平均映射误差: 0.001569 +[2025-04-11 15:22:43] [INFO] features.6.9.block.0.1.weight 的平均映射误差: 0.037609 +[2025-04-11 15:22:43] [INFO] features.6.9.block.1.0.weight 的平均映射误差: 0.001843 +[2025-04-11 15:22:43] [INFO] features.6.9.block.1.1.weight 的平均映射误差: 0.042367 +[2025-04-11 15:22:43] [INFO] features.6.9.block.2.fc1.weight 的平均映射误差: 0.000817 +[2025-04-11 15:22:43] [INFO] features.6.9.block.2.fc2.weight 的平均映射误差: 0.001089 +[2025-04-11 15:22:43] [INFO] features.6.9.block.3.0.weight 的平均映射误差: 0.001542 +[2025-04-11 15:22:43] [INFO] features.6.9.block.3.1.weight 的平均映射误差: 0.042464 +[2025-04-11 15:22:43] [INFO] features.6.10.block.0.0.weight 的平均映射误差: 0.001575 +[2025-04-11 15:22:43] [INFO] features.6.10.block.0.1.weight 的平均映射误差: 0.038996 +[2025-04-11 15:22:43] [INFO] features.6.10.block.1.0.weight 的平均映射误差: 0.001793 +[2025-04-11 15:22:43] [INFO] features.6.10.block.1.1.weight 的平均映射误差: 0.041568 +[2025-04-11 15:22:43] [INFO] features.6.10.block.2.fc1.weight 的平均映射误差: 0.000797 +[2025-04-11 15:22:43] [INFO] features.6.10.block.2.fc2.weight 的平均映射误差: 0.001142 +[2025-04-11 15:22:43] [INFO] features.6.10.block.3.0.weight 的平均映射误差: 0.001551 +[2025-04-11 15:22:43] [INFO] features.6.10.block.3.1.weight 的平均映射误差: 0.042276 +[2025-04-11 15:22:43] [INFO] features.6.11.block.0.0.weight 的平均映射误差: 0.001575 +[2025-04-11 15:22:43] [INFO] features.6.11.block.0.1.weight 的平均映射误差: 0.039237 +[2025-04-11 15:22:43] [INFO] features.6.11.block.1.0.weight 的平均映射误差: 0.001833 +[2025-04-11 15:22:43] [INFO] features.6.11.block.1.1.weight 的平均映射误差: 0.042174 +[2025-04-11 15:22:43] [INFO] features.6.11.block.2.fc1.weight 的平均映射误差: 0.000966 +[2025-04-11 15:22:43] [INFO] features.6.11.block.2.fc2.weight 的平均映射误差: 0.001163 +[2025-04-11 15:22:43] [INFO] features.6.11.block.3.0.weight 的平均映射误差: 0.001558 +[2025-04-11 15:22:43] [INFO] features.6.11.block.3.1.weight 的平均映射误差: 0.041001 +[2025-04-11 15:22:43] [INFO] features.6.12.block.0.0.weight 的平均映射误差: 0.001584 +[2025-04-11 15:22:43] [INFO] features.6.12.block.0.1.weight 的平均映射误差: 0.039881 +[2025-04-11 15:22:43] [INFO] features.6.12.block.1.0.weight 的平均映射误差: 0.001756 +[2025-04-11 15:22:43] [INFO] features.6.12.block.1.1.weight 的平均映射误差: 0.036506 +[2025-04-11 15:22:43] [INFO] features.6.12.block.2.fc1.weight 的平均映射误差: 0.000809 +[2025-04-11 15:22:43] [INFO] features.6.12.block.2.fc2.weight 的平均映射误差: 0.000973 +[2025-04-11 15:22:43] [INFO] features.6.12.block.3.0.weight 的平均映射误差: 0.001570 +[2025-04-11 15:22:43] [INFO] features.6.12.block.3.1.weight 的平均映射误差: 0.036072 +[2025-04-11 15:22:43] [INFO] features.6.13.block.0.0.weight 的平均映射误差: 0.001580 +[2025-04-11 15:22:43] [INFO] features.6.13.block.0.1.weight 的平均映射误差: 0.039214 +[2025-04-11 15:22:43] [INFO] features.6.13.block.1.0.weight 的平均映射误差: 0.001777 +[2025-04-11 15:22:43] [INFO] features.6.13.block.1.1.weight 的平均映射误差: 0.041932 +[2025-04-11 15:22:43] [INFO] features.6.13.block.2.fc1.weight 的平均映射误差: 0.000697 +[2025-04-11 15:22:43] [INFO] features.6.13.block.2.fc2.weight 的平均映射误差: 0.001245 +[2025-04-11 15:22:43] [INFO] features.6.13.block.3.0.weight 的平均映射误差: 0.001559 +[2025-04-11 15:22:43] [INFO] features.6.13.block.3.1.weight 的平均映射误差: 0.038902 +[2025-04-11 15:22:43] [INFO] features.6.14.block.0.0.weight 的平均映射误差: 0.001583 +[2025-04-11 15:22:43] [INFO] features.6.14.block.0.1.weight 的平均映射误差: 0.042021 +[2025-04-11 15:22:43] [INFO] features.6.14.block.1.0.weight 的平均映射误差: 0.001735 +[2025-04-11 15:22:43] [INFO] features.6.14.block.1.1.weight 的平均映射误差: 0.040372 +[2025-04-11 15:22:43] [INFO] features.6.14.block.2.fc1.weight 的平均映射误差: 0.000737 +[2025-04-11 15:22:43] [INFO] features.6.14.block.2.fc2.weight 的平均映射误差: 0.001043 +[2025-04-11 15:22:43] [INFO] features.6.14.block.3.0.weight 的平均映射误差: 0.001564 +[2025-04-11 15:22:43] [INFO] features.6.14.block.3.1.weight 的平均映射误差: 0.035191 +[2025-04-11 15:22:43] [INFO] features.6.15.block.0.0.weight 的平均映射误差: 0.001583 +[2025-04-11 15:22:43] [INFO] features.6.15.block.0.1.weight 的平均映射误差: 0.044032 +[2025-04-11 15:22:43] [INFO] features.6.15.block.1.0.weight 的平均映射误差: 0.001738 +[2025-04-11 15:22:43] [INFO] features.6.15.block.1.1.weight 的平均映射误差: 0.040565 +[2025-04-11 15:22:43] [INFO] features.6.15.block.2.fc1.weight 的平均映射误差: 0.000735 +[2025-04-11 15:22:43] [INFO] features.6.15.block.2.fc2.weight 的平均映射误差: 0.000839 +[2025-04-11 15:22:43] [INFO] features.6.15.block.3.0.weight 的平均映射误差: 0.001570 +[2025-04-11 15:22:43] [INFO] features.6.15.block.3.1.weight 的平均映射误差: 0.033758 +[2025-04-11 15:22:43] [INFO] features.6.16.block.0.0.weight 的平均映射误差: 0.001568 +[2025-04-11 15:22:43] [INFO] features.6.16.block.0.1.weight 的平均映射误差: 0.041166 +[2025-04-11 15:22:43] [INFO] features.6.16.block.1.0.weight 的平均映射误差: 0.001717 +[2025-04-11 15:22:43] [INFO] features.6.16.block.1.1.weight 的平均映射误差: 0.047619 +[2025-04-11 15:22:43] [INFO] features.6.16.block.2.fc1.weight 的平均映射误差: 0.000693 +[2025-04-11 15:22:43] [INFO] features.6.16.block.2.fc2.weight 的平均映射误差: 0.001177 +[2025-04-11 15:22:43] [INFO] features.6.16.block.3.0.weight 的平均映射误差: 0.001543 +[2025-04-11 15:22:43] [INFO] features.6.16.block.3.1.weight 的平均映射误差: 0.036831 +[2025-04-11 15:22:43] [INFO] features.6.17.block.0.0.weight 的平均映射误差: 0.001559 +[2025-04-11 15:22:43] [INFO] features.6.17.block.0.1.weight 的平均映射误差: 0.042565 +[2025-04-11 15:22:43] [INFO] features.6.17.block.1.0.weight 的平均映射误差: 0.001687 +[2025-04-11 15:22:43] [INFO] features.6.17.block.1.1.weight 的平均映射误差: 0.048230 +[2025-04-11 15:22:43] [INFO] features.6.17.block.2.fc1.weight 的平均映射误差: 0.000823 +[2025-04-11 15:22:43] [INFO] features.6.17.block.2.fc2.weight 的平均映射误差: 0.001306 +[2025-04-11 15:22:43] [INFO] features.6.17.block.3.0.weight 的平均映射误差: 0.001531 +[2025-04-11 15:22:43] [INFO] features.6.17.block.3.1.weight 的平均映射误差: 0.037562 +[2025-04-11 15:22:43] [INFO] features.7.0.block.0.0.weight 的平均映射误差: 0.001858 +[2025-04-11 15:22:43] [INFO] features.7.0.block.0.1.weight 的平均映射误差: 0.032060 +[2025-04-11 15:22:43] [INFO] features.7.0.block.1.0.weight 的平均映射误差: 0.002120 +[2025-04-11 15:22:43] [INFO] features.7.0.block.1.1.weight 的平均映射误差: 0.032251 +[2025-04-11 15:22:43] [INFO] features.7.0.block.2.fc1.weight 的平均映射误差: 0.001538 +[2025-04-11 15:22:43] [INFO] features.7.0.block.2.fc2.weight 的平均映射误差: 0.001701 +[2025-04-11 15:22:43] [INFO] features.7.0.block.3.0.weight 的平均映射误差: 0.001624 +[2025-04-11 15:22:43] [INFO] features.7.0.block.3.1.weight 的平均映射误差: 0.034763 +[2025-04-11 15:22:43] [INFO] features.7.1.block.0.0.weight 的平均映射误差: 0.001547 +[2025-04-11 15:22:43] [INFO] features.7.1.block.0.1.weight 的平均映射误差: 0.040559 +[2025-04-11 15:22:43] [INFO] features.7.1.block.1.0.weight 的平均映射误差: 0.001798 +[2025-04-11 15:22:43] [INFO] features.7.1.block.1.1.weight 的平均映射误差: 0.039211 +[2025-04-11 15:22:43] [INFO] features.7.1.block.2.fc1.weight 的平均映射误差: 0.001259 +[2025-04-11 15:22:43] [INFO] features.7.1.block.2.fc2.weight 的平均映射误差: 0.001628 +[2025-04-11 15:22:43] [INFO] features.7.1.block.3.0.weight 的平均映射误差: 0.001517 +[2025-04-11 15:22:43] [INFO] features.7.1.block.3.1.weight 的平均映射误差: 0.046713 +[2025-04-11 15:22:43] [INFO] features.7.2.block.0.0.weight 的平均映射误差: 0.001524 +[2025-04-11 15:22:43] [INFO] features.7.2.block.0.1.weight 的平均映射误差: 0.046619 +[2025-04-11 15:22:43] [INFO] features.7.2.block.1.0.weight 的平均映射误差: 0.002099 +[2025-04-11 15:22:43] [INFO] features.7.2.block.1.1.weight 的平均映射误差: 0.043570 +[2025-04-11 15:22:43] [INFO] features.7.2.block.2.fc1.weight 的平均映射误差: 0.001207 +[2025-04-11 15:22:43] [INFO] features.7.2.block.2.fc2.weight 的平均映射误差: 0.001289 +[2025-04-11 15:22:43] [INFO] features.7.2.block.3.0.weight 的平均映射误差: 0.001486 +[2025-04-11 15:22:43] [INFO] features.7.2.block.3.1.weight 的平均映射误差: 0.031953 +[2025-04-11 15:22:43] [INFO] features.7.3.block.0.0.weight 的平均映射误差: 0.001456 +[2025-04-11 15:22:43] [INFO] features.7.3.block.0.1.weight 的平均映射误差: 0.044966 +[2025-04-11 15:22:43] [INFO] features.7.3.block.1.0.weight 的平均映射误差: 0.002393 +[2025-04-11 15:22:43] [INFO] features.7.3.block.1.1.weight 的平均映射误差: 0.038847 +[2025-04-11 15:22:43] [INFO] features.7.3.block.2.fc1.weight 的平均映射误差: 0.001172 +[2025-04-11 15:22:43] [INFO] features.7.3.block.2.fc2.weight 的平均映射误差: 0.001262 +[2025-04-11 15:22:43] [INFO] features.7.3.block.3.0.weight 的平均映射误差: 0.001414 +[2025-04-11 15:22:43] [INFO] features.7.3.block.3.1.weight 的平均映射误差: 0.032231 +[2025-04-11 15:22:43] [INFO] features.7.4.block.0.0.weight 的平均映射误差: 0.001427 +[2025-04-11 15:22:43] [INFO] features.7.4.block.0.1.weight 的平均映射误差: 0.035093 +[2025-04-11 15:22:43] [INFO] features.7.4.block.1.0.weight 的平均映射误差: 0.002034 +[2025-04-11 15:22:43] [INFO] features.7.4.block.1.1.weight 的平均映射误差: 0.032436 +[2025-04-11 15:22:43] [INFO] features.7.4.block.2.fc1.weight 的平均映射误差: 0.001459 +[2025-04-11 15:22:43] [INFO] features.7.4.block.2.fc2.weight 的平均映射误差: 0.001212 +[2025-04-11 15:22:43] [INFO] features.7.4.block.3.0.weight 的平均映射误差: 0.001388 +[2025-04-11 15:22:43] [INFO] features.7.4.block.3.1.weight 的平均映射误差: 0.037126 +[2025-04-11 15:22:43] [INFO] features.8.0.weight 的平均映射误差: 0.001619 +[2025-04-11 15:22:43] [INFO] features.8.1.weight 的平均映射误差: 0.035800 +[2025-04-11 15:22:43] [INFO] classifier.1.weight 的平均映射误差: 0.001411 +[2025-04-11 15:22:45] [INFO] 评估结果 - Loss: 0.8530, Accuracy: 0.8285 +[2025-04-11 15:22:46] [INFO] RRAM映射模型已保存到 checkpoints/rram_mapped_model.pth +[2025-04-11 15:22:46] [INFO] RRAM映射前后精度对比: 原始 0.9087 vs RRAM映射后 0.8285, 变化: -0.0802 +[2025-04-11 15:22:46] [INFO] 开始微调全连接层 (epochs=50, lr=5e-05)... +[2025-04-11 15:22:46] [INFO] 微调过程中的模型将保存到: checkpoints/fine_tune_checkpoints +[2025-04-11 15:23:06] [INFO] Fine-tuning Epoch 1/50 - Train Acc: 0.9966, Val Acc: 0.9057 +[2025-04-11 15:23:06] [INFO] 已保存第 1 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_1.pth +[2025-04-11 15:23:27] [INFO] Fine-tuning Epoch 2/50 - Train Acc: 0.9952, Val Acc: 0.9178 +[2025-04-11 15:23:28] [INFO] 已保存第 2 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_2.pth +[2025-04-11 15:23:47] [INFO] Fine-tuning Epoch 3/50 - Train Acc: 0.9964, Val Acc: 0.9107 +[2025-04-11 15:23:48] [INFO] 已保存第 3 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_3.pth +[2025-04-11 15:24:08] [INFO] Fine-tuning Epoch 4/50 - Train Acc: 0.9942, Val Acc: 0.9178 +[2025-04-11 15:24:09] [INFO] 已保存第 4 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_4.pth +[2025-04-11 15:24:29] [INFO] Fine-tuning Epoch 5/50 - Train Acc: 0.9950, Val Acc: 0.9228 +[2025-04-11 15:24:30] [INFO] 已保存第 5 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_5.pth +[2025-04-11 15:24:50] [INFO] Fine-tuning Epoch 6/50 - Train Acc: 0.9976, Val Acc: 0.9168 +[2025-04-11 15:24:50] [INFO] 已保存第 6 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_6.pth +[2025-04-11 15:25:11] [INFO] Fine-tuning Epoch 7/50 - Train Acc: 0.9988, Val Acc: 0.9007 +[2025-04-11 15:25:11] [INFO] 已保存第 7 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_7.pth +[2025-04-11 15:25:32] [INFO] Fine-tuning Epoch 8/50 - Train Acc: 0.9990, Val Acc: 0.9147 +[2025-04-11 15:25:33] [INFO] 已保存第 8 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_8.pth +[2025-04-11 15:25:54] [INFO] Fine-tuning Epoch 9/50 - Train Acc: 0.9971, Val Acc: 0.8696 +[2025-04-11 15:25:55] [INFO] 已保存第 9 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_9.pth +[2025-04-11 15:26:16] [INFO] Fine-tuning Epoch 10/50 - Train Acc: 0.9974, Val Acc: 0.8977 +[2025-04-11 15:26:17] [INFO] 已保存第 10 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_10.pth +[2025-04-11 15:26:38] [INFO] Fine-tuning Epoch 11/50 - Train Acc: 0.9921, Val Acc: 0.8556 +[2025-04-11 15:26:39] [INFO] 已保存第 11 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_11.pth +[2025-04-11 15:27:01] [INFO] Fine-tuning Epoch 12/50 - Train Acc: 0.9942, Val Acc: 0.8927 +[2025-04-11 15:27:01] [INFO] 已保存第 12 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_12.pth +[2025-04-11 15:27:23] [INFO] Fine-tuning Epoch 13/50 - Train Acc: 0.9952, Val Acc: 0.9077 +[2025-04-11 15:27:24] [INFO] 已保存第 13 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_13.pth +[2025-04-11 15:27:45] [INFO] Fine-tuning Epoch 14/50 - Train Acc: 0.9966, Val Acc: 0.8706 +[2025-04-11 15:27:46] [INFO] 已保存第 14 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_14.pth +[2025-04-11 15:28:07] [INFO] Fine-tuning Epoch 15/50 - Train Acc: 0.9957, Val Acc: 0.8857 +[2025-04-11 15:28:08] [INFO] 已保存第 15 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_15.pth +[2025-04-11 15:28:29] [INFO] Fine-tuning Epoch 16/50 - Train Acc: 0.9983, Val Acc: 0.9077 +[2025-04-11 15:28:30] [INFO] 已保存第 16 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_16.pth +[2025-04-11 15:28:52] [INFO] Fine-tuning Epoch 17/50 - Train Acc: 0.9952, Val Acc: 0.8897 +[2025-04-11 15:28:52] [INFO] 已保存第 17 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_17.pth +[2025-04-11 15:29:14] [INFO] Fine-tuning Epoch 18/50 - Train Acc: 0.9954, Val Acc: 0.8756 +[2025-04-11 15:29:15] [INFO] 已保存第 18 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_18.pth +[2025-04-11 15:29:36] [INFO] Fine-tuning Epoch 19/50 - Train Acc: 0.9966, Val Acc: 0.9017 +[2025-04-11 15:29:37] [INFO] 已保存第 19 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_19.pth +[2025-04-11 15:29:58] [INFO] Fine-tuning Epoch 20/50 - Train Acc: 0.9990, Val Acc: 0.8907 +[2025-04-11 15:29:59] [INFO] 已保存第 20 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_20.pth +[2025-04-11 15:30:21] [INFO] Fine-tuning Epoch 21/50 - Train Acc: 0.9995, Val Acc: 0.8987 +[2025-04-11 15:30:22] [INFO] 已保存第 21 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_21.pth +[2025-04-11 15:30:43] [INFO] Fine-tuning Epoch 22/50 - Train Acc: 0.9983, Val Acc: 0.8887 +[2025-04-11 15:30:44] [INFO] 已保存第 22 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_22.pth +[2025-04-11 15:31:06] [INFO] Fine-tuning Epoch 23/50 - Train Acc: 0.9971, Val Acc: 0.8957 +[2025-04-11 15:31:07] [INFO] 已保存第 23 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_23.pth +[2025-04-11 15:31:28] [INFO] Fine-tuning Epoch 24/50 - Train Acc: 0.9986, Val Acc: 0.8957 +[2025-04-11 15:31:29] [INFO] 已保存第 24 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_24.pth +[2025-04-11 15:31:50] [INFO] Fine-tuning Epoch 25/50 - Train Acc: 0.9981, Val Acc: 0.8947 +[2025-04-11 15:31:51] [INFO] 已保存第 25 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_25.pth +[2025-04-11 15:32:13] [INFO] Fine-tuning Epoch 26/50 - Train Acc: 0.9976, Val Acc: 0.9468 +[2025-04-11 15:32:13] [INFO] 已保存第 26 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_26.pth +[2025-04-11 15:32:35] [INFO] Fine-tuning Epoch 27/50 - Train Acc: 0.9981, Val Acc: 0.9288 +[2025-04-11 15:32:36] [INFO] 已保存第 27 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_27.pth +[2025-04-11 15:32:56] [INFO] Fine-tuning Epoch 28/50 - Train Acc: 0.9993, Val Acc: 0.9338 +[2025-04-11 15:32:57] [INFO] 已保存第 28 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_28.pth +[2025-04-11 15:33:17] [INFO] Fine-tuning Epoch 29/50 - Train Acc: 0.9986, Val Acc: 0.9238 +[2025-04-11 15:33:18] [INFO] 已保存第 29 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_29.pth +[2025-04-11 15:33:39] [INFO] Fine-tuning Epoch 30/50 - Train Acc: 0.9971, Val Acc: 0.9408 +[2025-04-11 15:33:40] [INFO] 已保存第 30 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_30.pth +[2025-04-11 15:34:01] [INFO] Fine-tuning Epoch 31/50 - Train Acc: 0.9976, Val Acc: 0.8967 +[2025-04-11 15:34:02] [INFO] 已保存第 31 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_31.pth +[2025-04-11 15:34:22] [INFO] Fine-tuning Epoch 32/50 - Train Acc: 0.9964, Val Acc: 0.9278 +[2025-04-11 15:34:23] [INFO] 已保存第 32 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_32.pth +[2025-04-11 15:34:43] [INFO] Fine-tuning Epoch 33/50 - Train Acc: 0.9966, Val Acc: 0.9248 +[2025-04-11 15:34:44] [INFO] 已保存第 33 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_33.pth +[2025-04-11 15:35:04] [INFO] Fine-tuning Epoch 34/50 - Train Acc: 0.9986, Val Acc: 0.8987 +[2025-04-11 15:35:05] [INFO] 已保存第 34 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_34.pth +[2025-04-11 15:35:25] [INFO] Fine-tuning Epoch 35/50 - Train Acc: 0.9976, Val Acc: 0.9278 +[2025-04-11 15:35:26] [INFO] 已保存第 35 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_35.pth +[2025-04-11 15:35:46] [INFO] Fine-tuning Epoch 36/50 - Train Acc: 0.9976, Val Acc: 0.9137 +[2025-04-11 15:35:47] [INFO] 已保存第 36 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_36.pth +[2025-04-11 15:36:07] [INFO] Fine-tuning Epoch 37/50 - Train Acc: 0.9986, Val Acc: 0.9258 +[2025-04-11 15:36:08] [INFO] 已保存第 37 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_37.pth +[2025-04-11 15:36:28] [INFO] Fine-tuning Epoch 38/50 - Train Acc: 0.9998, Val Acc: 0.9127 +[2025-04-11 15:36:29] [INFO] 已保存第 38 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_38.pth +[2025-04-11 15:36:50] [INFO] Fine-tuning Epoch 39/50 - Train Acc: 0.9986, Val Acc: 0.9298 +[2025-04-11 15:36:50] [INFO] 已保存第 39 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_39.pth +[2025-04-11 15:37:11] [INFO] Fine-tuning Epoch 40/50 - Train Acc: 0.9983, Val Acc: 0.9097 +[2025-04-11 15:37:12] [INFO] 已保存第 40 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_40.pth +[2025-04-11 15:37:33] [INFO] Fine-tuning Epoch 41/50 - Train Acc: 0.9983, Val Acc: 0.8997 +[2025-04-11 15:37:34] [INFO] 已保存第 41 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_41.pth +[2025-04-11 15:37:54] [INFO] Fine-tuning Epoch 42/50 - Train Acc: 0.9995, Val Acc: 0.9097 +[2025-04-11 15:37:55] [INFO] 已保存第 42 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_42.pth +[2025-04-11 15:38:15] [INFO] Fine-tuning Epoch 43/50 - Train Acc: 0.9986, Val Acc: 0.9107 +[2025-04-11 15:38:16] [INFO] 已保存第 43 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_43.pth +[2025-04-11 15:38:36] [INFO] Fine-tuning Epoch 44/50 - Train Acc: 0.9962, Val Acc: 0.9007 +[2025-04-11 15:38:37] [INFO] 已保存第 44 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_44.pth +[2025-04-11 15:38:58] [INFO] Fine-tuning Epoch 45/50 - Train Acc: 0.9981, Val Acc: 0.9117 +[2025-04-11 15:38:59] [INFO] 已保存第 45 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_45.pth +[2025-04-11 15:39:20] [INFO] Fine-tuning Epoch 46/50 - Train Acc: 0.9993, Val Acc: 0.9208 +[2025-04-11 15:39:21] [INFO] 已保存第 46 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_46.pth +[2025-04-11 15:39:42] [INFO] Fine-tuning Epoch 47/50 - Train Acc: 0.9998, Val Acc: 0.9178 +[2025-04-11 15:39:43] [INFO] 已保存第 47 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_47.pth +[2025-04-11 15:40:04] [INFO] Fine-tuning Epoch 48/50 - Train Acc: 1.0000, Val Acc: 0.9208 +[2025-04-11 15:40:05] [INFO] 已保存第 48 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_48.pth +[2025-04-11 15:40:26] [INFO] Fine-tuning Epoch 49/50 - Train Acc: 0.9998, Val Acc: 0.9198 +[2025-04-11 15:40:27] [INFO] 已保存第 49 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_49.pth +[2025-04-11 15:40:48] [INFO] Fine-tuning Epoch 50/50 - Train Acc: 0.9990, Val Acc: 0.9308 +[2025-04-11 15:40:49] [INFO] 已保存第 50 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_50.pth +[2025-04-11 15:40:51] [INFO] 评估结果 - Loss: 0.4584, Accuracy: 0.9308 +[2025-04-11 15:40:52] [INFO] 微调模型已保存到 checkpoints/fine_tuned_model.pth +[2025-04-11 15:40:52] [INFO] 微调前后精度对比: RRAM映射 0.8285 vs 微调后 0.9308, 变化: 0.1023 +[2025-04-11 15:40:52] [INFO] 所有处理完成! diff --git a/checkpoints_v2m_part1/base_training_metrics.csv b/checkpoints_v2m_part1/base_training_metrics.csv new file mode 100644 index 0000000000000000000000000000000000000000..8edae1912ef7ce93c00c9bec45229999742e910a --- /dev/null +++ b/checkpoints_v2m_part1/base_training_metrics.csv @@ -0,0 +1,51 @@ +epoch,train_loss,train_acc,val_loss,val_acc +1,3.512,0.1151,3.4758,0.1424 +2,2.7881,0.2928,2.7346,0.321 +3,1.9895,0.507,1.8575,0.4975 +4,1.5039,0.6036,1.4558,0.5647 +5,1.1698,0.6796,1.2661,0.68 +6,0.9343,0.759,1.0889,0.7272 +7,0.8004,0.8151,0.9226,0.7773 +8,0.6381,0.8717,0.8111,0.7934 +9,0.5269,0.8974,0.649,0.8355 +10,0.363,0.9374,0.5186,0.8786 +11,0.2655,0.9508,0.4091,0.9017 +12,0.25,0.9549,0.5203,0.8626 +13,0.2335,0.953,0.4292,0.8877 +14,0.171,0.9657,0.3083,0.9228 +15,0.13,0.9758,0.3492,0.9107 +16,0.1371,0.9729,0.4028,0.9047 +17,0.1029,0.9796,0.2834,0.9268 +18,0.1109,0.9777,0.2368,0.9268 +19,0.1003,0.977,0.249,0.9248 +20,0.0969,0.9784,0.2582,0.9298 +21,0.0661,0.9847,0.2623,0.9408 +22,0.0559,0.9849,0.2204,0.9488 +23,0.0489,0.988,0.238,0.9408 +24,0.0533,0.9892,0.3553,0.8927 +25,0.0507,0.988,0.2503,0.9218 +26,0.0491,0.9902,0.2637,0.9188 +27,0.0369,0.9904,0.2795,0.9127 +28,0.0289,0.9959,0.3164,0.9107 +29,0.0331,0.9935,0.3443,0.9087 +30,0.0241,0.9971,0.236,0.9448 +31,0.0229,0.9966,0.2604,0.9358 +32,0.034,0.994,0.2789,0.9258 +33,0.0209,0.9978,0.3414,0.9198 +34,0.0187,0.9983,0.4294,0.9067 +35,0.0186,0.9978,0.3243,0.9388 +36,0.0122,0.9993,0.3431,0.9278 +37,0.0124,0.9993,0.3521,0.9248 +38,0.016,0.9981,0.3614,0.9298 +39,0.0139,0.9986,0.4236,0.9107 +40,0.0116,0.9998,0.4263,0.9127 +41,0.0141,0.999,0.4753,0.9027 +42,0.0128,0.9995,0.7317,0.8857 +43,0.0119,0.9993,0.7612,0.8826 +44,0.0101,0.9995,0.4204,0.9107 +45,0.0126,1.0,0.4008,0.9137 +46,0.012,0.9988,0.412,0.9178 +47,0.0088,1.0,0.4181,0.9157 +48,0.0114,0.9995,0.47,0.9057 +49,0.0107,0.9993,0.4309,0.9137 +50,0.0122,0.999,0.409,0.9087 diff --git a/checkpoints_v2m_part1/best_model.pth b/checkpoints_v2m_part1/best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..0de0dbba3c40d94448a3617d3b77035c9be2957f --- /dev/null +++ b/checkpoints_v2m_part1/best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f422b030576efff0e4d0b0f464a47c81fcae5f4a5e9ba613fd91d4f66a5ec7d4 +size 213307478 diff --git a/checkpoints_v2m_part1/fine_tuned_model.pth b/checkpoints_v2m_part1/fine_tuned_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..899f07f84fb4da3003f853a80daf129db375e48c --- /dev/null +++ b/checkpoints_v2m_part1/fine_tuned_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca3d2f88aa5ae39ce1523cd000bd1eed02e07ac1809dca5ed715955dbfbcd491 +size 213314222 diff --git a/checkpoints_v2m_part1/fine_tuning_metrics.csv b/checkpoints_v2m_part1/fine_tuning_metrics.csv new file mode 100644 index 0000000000000000000000000000000000000000..db355ebb3b3d6b66988b0dad0dafc31fc3911ef2 --- /dev/null +++ b/checkpoints_v2m_part1/fine_tuning_metrics.csv @@ -0,0 +1,51 @@ +epoch,train_acc,val_acc +1,0.9966,0.9057 +2,0.9952,0.9178 +3,0.9964,0.9107 +4,0.9942,0.9178 +5,0.995,0.9228 +6,0.9976,0.9168 +7,0.9988,0.9007 +8,0.999,0.9147 +9,0.9971,0.8696 +10,0.9974,0.8977 +11,0.9921,0.8556 +12,0.9942,0.8927 +13,0.9952,0.9077 +14,0.9966,0.8706 +15,0.9957,0.8857 +16,0.9983,0.9077 +17,0.9952,0.8897 +18,0.9954,0.8756 +19,0.9966,0.9017 +20,0.999,0.8907 +21,0.9995,0.8987 +22,0.9983,0.8887 +23,0.9971,0.8957 +24,0.9986,0.8957 +25,0.9981,0.8947 +26,0.9976,0.9468 +27,0.9981,0.9288 +28,0.9993,0.9338 +29,0.9986,0.9238 +30,0.9971,0.9408 +31,0.9976,0.8967 +32,0.9964,0.9278 +33,0.9966,0.9248 +34,0.9986,0.8987 +35,0.9976,0.9278 +36,0.9976,0.9137 +37,0.9986,0.9258 +38,0.9998,0.9127 +39,0.9986,0.9298 +40,0.9983,0.9097 +41,0.9983,0.8997 +42,0.9995,0.9097 +43,0.9986,0.9107 +44,0.9962,0.9007 +45,0.9981,0.9117 +46,0.9993,0.9208 +47,0.9998,0.9178 +48,1.0,0.9208 +49,0.9998,0.9198 +50,0.999,0.9308 diff --git a/checkpoints_v2m_part1/last_checkpoint.pth b/checkpoints_v2m_part1/last_checkpoint.pth new file mode 100644 index 0000000000000000000000000000000000000000..4661ad00f0d0a8c967c425af97f0a69c3396eca4 --- /dev/null +++ b/checkpoints_v2m_part1/last_checkpoint.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a28d7988ddef13c22b609079ddfa21623fbba804c6a5ca27a29526e4433faaff +size 637341255 diff --git a/checkpoints_v2m_part1/parse_log.py b/checkpoints_v2m_part1/parse_log.py new file mode 100644 index 0000000000000000000000000000000000000000..d6fa873b793827c44bd9882739d8ca0f2ffc51ba --- /dev/null +++ b/checkpoints_v2m_part1/parse_log.py @@ -0,0 +1,62 @@ +import re +import pandas as pd + +def parse_log_file(log_file_path): + # 初始化存储数据的列表 + base_training_data = [] + fine_tuning_data = [] + + with open(log_file_path, 'r', encoding='utf-8') as f: + content = f.read() + + # 提取基础训练数据 + base_pattern = r'Epoch (\d+)/50 - Train Loss: ([\d.]+), Train Acc: ([\d.]+), Val Loss: ([\d.]+), Val Acc: ([\d.]+)' + base_matches = re.finditer(base_pattern, content) + + for match in base_matches: + epoch = int(match.group(1)) + train_loss = float(match.group(2)) + train_acc = float(match.group(3)) + val_loss = float(match.group(4)) + val_acc = float(match.group(5)) + + # 如果epoch小于等于50，认为是基础训练数据 + if epoch <= 50: + base_training_data.append({ + 'epoch': epoch, + 'train_loss': train_loss, + 'train_acc': train_acc, + 'val_loss': val_loss, + 'val_acc': val_acc + }) + + # 提取微调训练数据 + fine_tune_pattern = r'Fine-tuning Epoch (\d+)/50 - Train Acc: ([\d.]+), Val Acc: ([\d.]+)' + fine_tune_matches = re.finditer(fine_tune_pattern, content) + + for match in fine_tune_matches: + epoch = int(match.group(1)) + train_acc = float(match.group(2)) + val_acc = float(match.group(3)) + + fine_tuning_data.append({ + 'epoch': epoch, + 'train_acc': train_acc, + 'val_acc': val_acc + }) + + # 转换为DataFrame并保存为CSV + if base_training_data: + base_df = pd.DataFrame(base_training_data) + base_df.to_csv('base_training_metrics.csv', index=False) + print(f"基础训练数据已保存到 base_training_metrics.csv，共 {len(base_training_data)} 条记录") + + if fine_tuning_data: + fine_tune_df = pd.DataFrame(fine_tuning_data) + fine_tune_df.to_csv('fine_tuning_metrics.csv', index=False) + print(f"微调训练数据已保存到 fine_tuning_metrics.csv，共 {len(fine_tuning_data)} 条记录") + +if __name__ == '__main__': + # 指定日志文件路径 + log_file_path = '2025-04-11_15-04-17_train.log' + parse_log_file(log_file_path) \ No newline at end of file diff --git a/checkpoints_v2m_part1/rram_mapped_model.pth b/checkpoints_v2m_part1/rram_mapped_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..9d117417c134df847f5dbc6a1b0674f39364fb5b --- /dev/null +++ b/checkpoints_v2m_part1/rram_mapped_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d9d056e0c3a51f405f0faf25528f904b7faa5f1fbeff601685ba248349c446a +size 213315346 diff --git a/checkpoints_v2m_part1/training_plot.png b/checkpoints_v2m_part1/training_plot.png new file mode 100644 index 0000000000000000000000000000000000000000..d0b99f48c63ad7da76d7243da7a5307007b153a4 Binary files /dev/null and b/checkpoints_v2m_part1/training_plot.png differ diff --git a/checkpoints_v2m_part1/visualizations/base_weights_heatmap.png b/checkpoints_v2m_part1/visualizations/base_weights_heatmap.png new file mode 100644 index 0000000000000000000000000000000000000000..ee53f92f26a25dd2f3124252ca915cf74f968b32 --- /dev/null +++ b/checkpoints_v2m_part1/visualizations/base_weights_heatmap.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7143e615e44c45779bde04eb69027e1b2a36ac8ee43e35c348a2917ae1374d05 +size 460081 diff --git a/checkpoints_v2m_part1/visualizations/fine_tuned_weights_heatmap.png b/checkpoints_v2m_part1/visualizations/fine_tuned_weights_heatmap.png new file mode 100644 index 0000000000000000000000000000000000000000..77034861cf62d7a4aabe9440658ecc9f5bad9ff7 --- /dev/null +++ b/checkpoints_v2m_part1/visualizations/fine_tuned_weights_heatmap.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3caeef0d07460a7597fc013ad1d5bb64d5345fd031b74251ed52babf8f5d6905 +size 429113 diff --git a/checkpoints_v2m_part1/visualizations/mapping_error_distribution.png b/checkpoints_v2m_part1/visualizations/mapping_error_distribution.png new file mode 100644 index 0000000000000000000000000000000000000000..118f446d40b5b1b45d6d20169217c2a300794fc6 --- /dev/null +++ b/checkpoints_v2m_part1/visualizations/mapping_error_distribution.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b253a3c5f917f173bb8e91aac33cf895b62a24b4b85ee1a5e8f9458e456d3aa +size 405890 diff --git a/checkpoints_v2m_part1/visualizations/weight_changes_heatmap.png b/checkpoints_v2m_part1/visualizations/weight_changes_heatmap.png new file mode 100644 index 0000000000000000000000000000000000000000..0ee3ee6a281cc8f7d200a21b45c74ce24abcf41d --- /dev/null +++ b/checkpoints_v2m_part1/visualizations/weight_changes_heatmap.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:478a7c02308f0f6dc320d94f88be365cea26df21149f1b1448238f9fa0764e40 +size 383829 diff --git a/checkpoints_v2m_part2/2025-04-11_14-13-49_train.log b/checkpoints_v2m_part2/2025-04-11_14-13-49_train.log new file mode 100644 index 0000000000000000000000000000000000000000..9d8921c1fd99ac61bf5690a4f35b69d809cd164d --- /dev/null +++ b/checkpoints_v2m_part2/2025-04-11_14-13-49_train.log @@ -0,0 +1,730 @@ +[2025-04-11 14:13:49] [INFO] 使用设备: cuda:0 +[2025-04-11 14:13:49] [INFO] 训练集注释文件: /data0/work/DuYiFan/projects/traffic_classify/4_directions/TsignRecgTrainAnnotation.txt +[2025-04-11 14:13:49] [INFO] 测试集注释文件: /data0/work/DuYiFan/projects/traffic_classify/4_directions/TsignRecgTestAnnotation.txt +[2025-04-11 14:13:49] [INFO] 训练图像目录: /data0/work/DuYiFan/projects/traffic_classify/4_directions/train +[2025-04-11 14:13:49] [INFO] 测试图像目录: /data0/work/DuYiFan/projects/traffic_classify/4_directions/test +[2025-04-11 14:13:49] [INFO] 创建数据集和数据加载器 +[2025-04-11 14:13:49] [INFO] 创建efficientnet-v2-m模型，类别数: 4 +[2025-04-11 14:13:50] [INFO] 设置损失函数、优化器和学习率调度器，初始学习率: 0.0001 +[2025-04-11 14:13:50] [INFO] 开始训练，总共 50 轮 +[2025-04-11 14:13:50] [INFO] 当前学习率: 0.000100 +[2025-04-11 14:13:50] [INFO] Epoch 1/50 开始训练 +[2025-04-11 14:13:52] [INFO] Epoch 1/50 开始验证 +[2025-04-11 14:13:52] [INFO] Epoch 1/50 - Train Loss: 1.4098, Train Acc: 0.2183, Val Loss: 1.3692, Val Acc: 0.4483 +[2025-04-11 14:13:52] [INFO] 已保存最佳模型，准确率: 0.4483 +[2025-04-11 14:13:53] [INFO] 当前学习率: 0.000100 +[2025-04-11 14:13:53] [INFO] Epoch 2/50 开始训练 +[2025-04-11 14:13:54] [INFO] Epoch 2/50 开始验证 +[2025-04-11 14:13:55] [INFO] Epoch 2/50 - Train Loss: 1.3061, Train Acc: 0.5352, Val Loss: 1.3508, Val Acc: 0.4483 +[2025-04-11 14:13:56] [INFO] 当前学习率: 0.000100 +[2025-04-11 14:13:56] [INFO] Epoch 3/50 开始训练 +[2025-04-11 14:13:56] [INFO] Epoch 3/50 开始验证 +[2025-04-11 14:13:57] [INFO] Epoch 3/50 - Train Loss: 1.2258, Train Acc: 0.7042, Val Loss: 1.3288, Val Acc: 0.4483 +[2025-04-11 14:13:58] [INFO] 当前学习率: 0.000099 +[2025-04-11 14:13:58] [INFO] Epoch 4/50 开始训练 +[2025-04-11 14:13:59] [INFO] Epoch 4/50 开始验证 +[2025-04-11 14:13:59] [INFO] Epoch 4/50 - Train Loss: 1.1423, Train Acc: 0.7042, Val Loss: 1.3089, Val Acc: 0.4483 +[2025-04-11 14:14:00] [INFO] 当前学习率: 0.000098 +[2025-04-11 14:14:00] [INFO] Epoch 5/50 开始训练 +[2025-04-11 14:14:01] [INFO] Epoch 5/50 开始验证 +[2025-04-11 14:14:01] [INFO] Epoch 5/50 - Train Loss: 1.0667, Train Acc: 0.7042, Val Loss: 1.2984, Val Acc: 0.4483 +[2025-04-11 14:14:03] [INFO] 当前学习率: 0.000098 +[2025-04-11 14:14:03] [INFO] Epoch 6/50 开始训练 +[2025-04-11 14:14:03] [INFO] Epoch 6/50 开始验证 +[2025-04-11 14:14:04] [INFO] Epoch 6/50 - Train Loss: 0.9744, Train Acc: 0.7042, Val Loss: 1.2977, Val Acc: 0.4483 +[2025-04-11 14:14:05] [INFO] 当前学习率: 0.000097 +[2025-04-11 14:14:05] [INFO] Epoch 7/50 开始训练 +[2025-04-11 14:14:06] [INFO] Epoch 7/50 开始验证 +[2025-04-11 14:14:06] [INFO] Epoch 7/50 - Train Loss: 0.9159, Train Acc: 0.7042, Val Loss: 1.3075, Val Acc: 0.4483 +[2025-04-11 14:14:07] [INFO] 当前学习率: 0.000095 +[2025-04-11 14:14:07] [INFO] Epoch 8/50 开始训练 +[2025-04-11 14:14:08] [INFO] Epoch 8/50 开始验证 +[2025-04-11 14:14:08] [INFO] Epoch 8/50 - Train Loss: 0.8672, Train Acc: 0.7042, Val Loss: 1.3161, Val Acc: 0.4483 +[2025-04-11 14:14:09] [INFO] 当前学习率: 0.000094 +[2025-04-11 14:14:09] [INFO] Epoch 9/50 开始训练 +[2025-04-11 14:14:10] [INFO] Epoch 9/50 开始验证 +[2025-04-11 14:14:10] [INFO] Epoch 9/50 - Train Loss: 0.8349, Train Acc: 0.7042, Val Loss: 1.3154, Val Acc: 0.4483 +[2025-04-11 14:14:11] [INFO] 当前学习率: 0.000092 +[2025-04-11 14:14:11] [INFO] Epoch 10/50 开始训练 +[2025-04-11 14:14:12] [INFO] Epoch 10/50 开始验证 +[2025-04-11 14:14:13] [INFO] Epoch 10/50 - Train Loss: 0.8062, Train Acc: 0.7042, Val Loss: 1.3091, Val Acc: 0.4483 +[2025-04-11 14:14:14] [INFO] 当前学习率: 0.000091 +[2025-04-11 14:14:14] [INFO] Epoch 11/50 开始训练 +[2025-04-11 14:14:15] [INFO] Epoch 11/50 开始验证 +[2025-04-11 14:14:15] [INFO] Epoch 11/50 - Train Loss: 0.7764, Train Acc: 0.7042, Val Loss: 1.2824, Val Acc: 0.4483 +[2025-04-11 14:14:16] [INFO] 当前学习率: 0.000089 +[2025-04-11 14:14:16] [INFO] Epoch 12/50 开始训练 +[2025-04-11 14:14:17] [INFO] Epoch 12/50 开始验证 +[2025-04-11 14:14:17] [INFO] Epoch 12/50 - Train Loss: 0.7459, Train Acc: 0.7042, Val Loss: 1.2373, Val Acc: 0.4483 +[2025-04-11 14:14:18] [INFO] 当前学习率: 0.000087 +[2025-04-11 14:14:18] [INFO] Epoch 13/50 开始训练 +[2025-04-11 14:14:19] [INFO] Epoch 13/50 开始验证 +[2025-04-11 14:14:20] [INFO] Epoch 13/50 - Train Loss: 0.7026, Train Acc: 0.7042, Val Loss: 1.2288, Val Acc: 0.4483 +[2025-04-11 14:14:21] [INFO] 当前学习率: 0.000084 +[2025-04-11 14:14:21] [INFO] Epoch 14/50 开始训练 +[2025-04-11 14:14:21] [INFO] Epoch 14/50 开始验证 +[2025-04-11 14:14:22] [INFO] Epoch 14/50 - Train Loss: 0.6678, Train Acc: 0.7042, Val Loss: 1.2415, Val Acc: 0.4483 +[2025-04-11 14:14:23] [INFO] 当前学习率: 0.000082 +[2025-04-11 14:14:23] [INFO] Epoch 15/50 开始训练 +[2025-04-11 14:14:24] [INFO] Epoch 15/50 开始验证 +[2025-04-11 14:14:24] [INFO] Epoch 15/50 - Train Loss: 0.6473, Train Acc: 0.7042, Val Loss: 1.2037, Val Acc: 0.4483 +[2025-04-11 14:14:25] [INFO] 当前学习率: 0.000080 +[2025-04-11 14:14:25] [INFO] Epoch 16/50 开始训练 +[2025-04-11 14:14:26] [INFO] Epoch 16/50 开始验证 +[2025-04-11 14:14:26] [INFO] Epoch 16/50 - Train Loss: 0.6035, Train Acc: 0.7394, Val Loss: 1.1331, Val Acc: 0.4483 +[2025-04-11 14:14:27] [INFO] 当前学习率: 0.000077 +[2025-04-11 14:14:27] [INFO] Epoch 17/50 开始训练 +[2025-04-11 14:14:28] [INFO] Epoch 17/50 开始验证 +[2025-04-11 14:14:29] [INFO] Epoch 17/50 - Train Loss: 0.5741, Train Acc: 0.7535, Val Loss: 1.1263, Val Acc: 0.4483 +[2025-04-11 14:14:30] [INFO] 当前学习率: 0.000074 +[2025-04-11 14:14:30] [INFO] Epoch 18/50 开始训练 +[2025-04-11 14:14:31] [INFO] Epoch 18/50 开始验证 +[2025-04-11 14:14:31] [INFO] Epoch 18/50 - Train Loss: 0.5583, Train Acc: 0.7887, Val Loss: 1.1548, Val Acc: 0.4483 +[2025-04-11 14:14:32] [INFO] 当前学习率: 0.000072 +[2025-04-11 14:14:32] [INFO] Epoch 19/50 开始训练 +[2025-04-11 14:14:33] [INFO] Epoch 19/50 开始验证 +[2025-04-11 14:14:33] [INFO] Epoch 19/50 - Train Loss: 0.5233, Train Acc: 0.8028, Val Loss: 1.1269, Val Acc: 0.5172 +[2025-04-11 14:14:34] [INFO] 已保存最佳模型，准确率: 0.5172 +[2025-04-11 14:14:35] [INFO] 当前学习率: 0.000069 +[2025-04-11 14:14:35] [INFO] Epoch 20/50 开始训练 +[2025-04-11 14:14:35] [INFO] Epoch 20/50 开始验证 +[2025-04-11 14:14:36] [INFO] Epoch 20/50 - Train Loss: 0.5189, Train Acc: 0.7746, Val Loss: 1.1425, Val Acc: 0.3793 +[2025-04-11 14:14:37] [INFO] 当前学习率: 0.000066 +[2025-04-11 14:14:37] [INFO] Epoch 21/50 开始训练 +[2025-04-11 14:14:38] [INFO] Epoch 21/50 开始验证 +[2025-04-11 14:14:38] [INFO] Epoch 21/50 - Train Loss: 0.5140, Train Acc: 0.7958, Val Loss: 1.1988, Val Acc: 0.3448 +[2025-04-11 14:14:39] [INFO] 当前学习率: 0.000063 +[2025-04-11 14:14:39] [INFO] Epoch 22/50 开始训练 +[2025-04-11 14:14:40] [INFO] Epoch 22/50 开始验证 +[2025-04-11 14:14:40] [INFO] Epoch 22/50 - Train Loss: 0.5165, Train Acc: 0.7817, Val Loss: 1.2972, Val Acc: 0.3448 +[2025-04-11 14:14:41] [INFO] 当前学习率: 0.000060 +[2025-04-11 14:14:41] [INFO] Epoch 23/50 开始训练 +[2025-04-11 14:14:42] [INFO] Epoch 23/50 开始验证 +[2025-04-11 14:14:43] [INFO] Epoch 23/50 - Train Loss: 0.4809, Train Acc: 0.7958, Val Loss: 1.1992, Val Acc: 0.3448 +[2025-04-11 14:14:44] [INFO] 当前学习率: 0.000057 +[2025-04-11 14:14:44] [INFO] Epoch 24/50 开始训练 +[2025-04-11 14:14:45] [INFO] Epoch 24/50 开始验证 +[2025-04-11 14:14:45] [INFO] Epoch 24/50 - Train Loss: 0.4835, Train Acc: 0.7887, Val Loss: 1.1661, Val Acc: 0.3793 +[2025-04-11 14:14:46] [INFO] 当前学习率: 0.000054 +[2025-04-11 14:14:46] [INFO] Epoch 25/50 开始训练 +[2025-04-11 14:14:47] [INFO] Epoch 25/50 开始验证 +[2025-04-11 14:14:47] [INFO] Epoch 25/50 - Train Loss: 0.4557, Train Acc: 0.7958, Val Loss: 1.2521, Val Acc: 0.3448 +[2025-04-11 14:14:48] [INFO] 当前学习率: 0.000050 +[2025-04-11 14:14:48] [INFO] Epoch 26/50 开始训练 +[2025-04-11 14:14:49] [INFO] Epoch 26/50 开始验证 +[2025-04-11 14:14:50] [INFO] Epoch 26/50 - Train Loss: 0.4590, Train Acc: 0.8169, Val Loss: 1.4326, Val Acc: 0.2759 +[2025-04-11 14:14:51] [INFO] 当前学习率: 0.000047 +[2025-04-11 14:14:51] [INFO] Epoch 27/50 开始训练 +[2025-04-11 14:14:51] [INFO] Epoch 27/50 开始验证 +[2025-04-11 14:14:52] [INFO] Epoch 27/50 - Train Loss: 0.4430, Train Acc: 0.8239, Val Loss: 1.4415, Val Acc: 0.2759 +[2025-04-11 14:14:53] [INFO] 当前学习率: 0.000044 +[2025-04-11 14:14:53] [INFO] Epoch 28/50 开始训练 +[2025-04-11 14:14:54] [INFO] Epoch 28/50 开始验证 +[2025-04-11 14:14:54] [INFO] Epoch 28/50 - Train Loss: 0.4579, Train Acc: 0.8310, Val Loss: 1.3938, Val Acc: 0.3448 +[2025-04-11 14:14:55] [INFO] 当前学习率: 0.000041 +[2025-04-11 14:14:55] [INFO] Epoch 29/50 开始训练 +[2025-04-11 14:14:56] [INFO] Epoch 29/50 开始验证 +[2025-04-11 14:14:56] [INFO] Epoch 29/50 - Train Loss: 0.4221, Train Acc: 0.8662, Val Loss: 1.2294, Val Acc: 0.3448 +[2025-04-11 14:14:57] [INFO] 当前学习率: 0.000038 +[2025-04-11 14:14:57] [INFO] Epoch 30/50 开始训练 +[2025-04-11 14:14:58] [INFO] Epoch 30/50 开始验证 +[2025-04-11 14:14:59] [INFO] Epoch 30/50 - Train Loss: 0.4264, Train Acc: 0.8310, Val Loss: 1.0815, Val Acc: 0.4138 +[2025-04-11 14:15:00] [INFO] 当前学习率: 0.000035 +[2025-04-11 14:15:00] [INFO] Epoch 31/50 开始训练 +[2025-04-11 14:15:01] [INFO] Epoch 31/50 开始验证 +[2025-04-11 14:15:01] [INFO] Epoch 31/50 - Train Loss: 0.3995, Train Acc: 0.8662, Val Loss: 1.0540, Val Acc: 0.5172 +[2025-04-11 14:15:02] [INFO] 当前学习率: 0.000032 +[2025-04-11 14:15:02] [INFO] Epoch 32/50 开始训练 +[2025-04-11 14:15:03] [INFO] Epoch 32/50 开始验证 +[2025-04-11 14:15:03] [INFO] Epoch 32/50 - Train Loss: 0.3980, Train Acc: 0.8380, Val Loss: 1.0917, Val Acc: 0.5172 +[2025-04-11 14:15:04] [INFO] 当前学习率: 0.000029 +[2025-04-11 14:15:04] [INFO] Epoch 33/50 开始训练 +[2025-04-11 14:15:05] [INFO] Epoch 33/50 开始验证 +[2025-04-11 14:15:06] [INFO] Epoch 33/50 - Train Loss: 0.3591, Train Acc: 0.8803, Val Loss: 1.0186, Val Acc: 0.5172 +[2025-04-11 14:15:07] [INFO] 当前学习率: 0.000027 +[2025-04-11 14:15:07] [INFO] Epoch 34/50 开始训练 +[2025-04-11 14:15:07] [INFO] Epoch 34/50 开始验证 +[2025-04-11 14:15:08] [INFO] Epoch 34/50 - Train Loss: 0.3420, Train Acc: 0.8803, Val Loss: 1.0275, Val Acc: 0.4483 +[2025-04-11 14:15:09] [INFO] 当前学习率: 0.000024 +[2025-04-11 14:15:09] [INFO] Epoch 35/50 开始训练 +[2025-04-11 14:15:10] [INFO] Epoch 35/50 开始验证 +[2025-04-11 14:15:10] [INFO] Epoch 35/50 - Train Loss: 0.3773, Train Acc: 0.8592, Val Loss: 1.0903, Val Acc: 0.4828 +[2025-04-11 14:15:11] [INFO] 当前学习率: 0.000021 +[2025-04-11 14:15:11] [INFO] Epoch 36/50 开始训练 +[2025-04-11 14:15:12] [INFO] Epoch 36/50 开始验证 +[2025-04-11 14:15:12] [INFO] Epoch 36/50 - Train Loss: 0.3629, Train Acc: 0.8873, Val Loss: 1.1087, Val Acc: 0.5172 +[2025-04-11 14:15:13] [INFO] 当前学习率: 0.000019 +[2025-04-11 14:15:13] [INFO] Epoch 37/50 开始训练 +[2025-04-11 14:15:14] [INFO] Epoch 37/50 开始验证 +[2025-04-11 14:15:15] [INFO] Epoch 37/50 - Train Loss: 0.3062, Train Acc: 0.8944, Val Loss: 1.1035, Val Acc: 0.5172 +[2025-04-11 14:15:16] [INFO] 当前学习率: 0.000017 +[2025-04-11 14:15:16] [INFO] Epoch 38/50 开始训练 +[2025-04-11 14:15:16] [INFO] Epoch 38/50 开始验证 +[2025-04-11 14:15:17] [INFO] Epoch 38/50 - Train Loss: 0.3355, Train Acc: 0.9085, Val Loss: 1.0940, Val Acc: 0.5172 +[2025-04-11 14:15:18] [INFO] 当前学习率: 0.000014 +[2025-04-11 14:15:18] [INFO] Epoch 39/50 开始训练 +[2025-04-11 14:15:19] [INFO] Epoch 39/50 开始验证 +[2025-04-11 14:15:19] [INFO] Epoch 39/50 - Train Loss: 0.3338, Train Acc: 0.8803, Val Loss: 1.0815, Val Acc: 0.5172 +[2025-04-11 14:15:20] [INFO] 当前学习率: 0.000012 +[2025-04-11 14:15:20] [INFO] Epoch 40/50 开始训练 +[2025-04-11 14:15:21] [INFO] Epoch 40/50 开始验证 +[2025-04-11 14:15:21] [INFO] Epoch 40/50 - Train Loss: 0.3105, Train Acc: 0.8803, Val Loss: 1.0742, Val Acc: 0.5172 +[2025-04-11 14:15:22] [INFO] 当前学习率: 0.000010 +[2025-04-11 14:15:22] [INFO] Epoch 41/50 开始训练 +[2025-04-11 14:15:23] [INFO] Epoch 41/50 开始验证 +[2025-04-11 14:15:24] [INFO] Epoch 41/50 - Train Loss: 0.3438, Train Acc: 0.8873, Val Loss: 1.0633, Val Acc: 0.5862 +[2025-04-11 14:15:24] [INFO] 已保存最佳模型，准确率: 0.5862 +[2025-04-11 14:15:25] [INFO] 当前学习率: 0.000009 +[2025-04-11 14:15:25] [INFO] Epoch 42/50 开始训练 +[2025-04-11 14:15:26] [INFO] Epoch 42/50 开始验证 +[2025-04-11 14:15:26] [INFO] Epoch 42/50 - Train Loss: 0.3150, Train Acc: 0.8944, Val Loss: 1.0631, Val Acc: 0.5862 +[2025-04-11 14:15:27] [INFO] 当前学习率: 0.000007 +[2025-04-11 14:15:28] [INFO] Epoch 43/50 开始训练 +[2025-04-11 14:15:28] [INFO] Epoch 43/50 开始验证 +[2025-04-11 14:15:29] [INFO] Epoch 43/50 - Train Loss: 0.3168, Train Acc: 0.8944, Val Loss: 1.0575, Val Acc: 0.5862 +[2025-04-11 14:15:30] [INFO] 当前学习率: 0.000006 +[2025-04-11 14:15:30] [INFO] Epoch 44/50 开始训练 +[2025-04-11 14:15:31] [INFO] Epoch 44/50 开始验证 +[2025-04-11 14:15:31] [INFO] Epoch 44/50 - Train Loss: 0.2939, Train Acc: 0.9085, Val Loss: 1.0698, Val Acc: 0.5862 +[2025-04-11 14:15:32] [INFO] 当前学习率: 0.000004 +[2025-04-11 14:15:32] [INFO] Epoch 45/50 开始训练 +[2025-04-11 14:15:33] [INFO] Epoch 45/50 开始验证 +[2025-04-11 14:15:33] [INFO] Epoch 45/50 - Train Loss: 0.3333, Train Acc: 0.8662, Val Loss: 1.0725, Val Acc: 0.5862 +[2025-04-11 14:15:34] [INFO] 当前学习率: 0.000003 +[2025-04-11 14:15:34] [INFO] Epoch 46/50 开始训练 +[2025-04-11 14:15:35] [INFO] Epoch 46/50 开始验证 +[2025-04-11 14:15:35] [INFO] Epoch 46/50 - Train Loss: 0.3176, Train Acc: 0.8803, Val Loss: 1.0823, Val Acc: 0.5862 +[2025-04-11 14:15:36] [INFO] 当前学习率: 0.000003 +[2025-04-11 14:15:36] [INFO] Epoch 47/50 开始训练 +[2025-04-11 14:15:37] [INFO] Epoch 47/50 开始验证 +[2025-04-11 14:15:38] [INFO] Epoch 47/50 - Train Loss: 0.2840, Train Acc: 0.9225, Val Loss: 1.0824, Val Acc: 0.5862 +[2025-04-11 14:15:39] [INFO] 当前学习率: 0.000002 +[2025-04-11 14:15:39] [INFO] Epoch 48/50 开始训练 +[2025-04-11 14:15:40] [INFO] Epoch 48/50 开始验证 +[2025-04-11 14:15:40] [INFO] Epoch 48/50 - Train Loss: 0.2919, Train Acc: 0.9014, Val Loss: 1.0881, Val Acc: 0.5862 +[2025-04-11 14:15:41] [INFO] 当前学习率: 0.000001 +[2025-04-11 14:15:41] [INFO] Epoch 49/50 开始训练 +[2025-04-11 14:15:42] [INFO] Epoch 49/50 开始验证 +[2025-04-11 14:15:42] [INFO] Epoch 49/50 - Train Loss: 0.2736, Train Acc: 0.9085, Val Loss: 1.0791, Val Acc: 0.5862 +[2025-04-11 14:15:43] [INFO] 当前学习率: 0.000001 +[2025-04-11 14:15:43] [INFO] Epoch 50/50 开始训练 +[2025-04-11 14:15:44] [INFO] Epoch 50/50 开始验证 +[2025-04-11 14:15:45] [INFO] Epoch 50/50 - Train Loss: 0.3232, Train Acc: 0.9014, Val Loss: 1.0816, Val Acc: 0.5862 +[2025-04-11 14:15:46] [INFO] 绘制训练过程图表 +[2025-04-11 14:15:46] [INFO] 标准训练完成! +[2025-04-11 14:15:46] [INFO] 评估原始模型性能... +[2025-04-11 14:15:47] [INFO] 评估结果 - Loss: 1.0816, Accuracy: 0.5862 +[2025-04-11 14:15:47] [INFO] 开始执行RRAM映射... +[2025-04-11 14:15:47] [INFO] 加载了 100 个RRAM电导值 +[2025-04-11 14:15:47] [INFO] features.0.0.weight 的平均映射误差: 0.018786 +[2025-04-11 14:15:47] [INFO] features.0.1.weight 的平均映射误差: 0.033996 +[2025-04-11 14:15:47] [INFO] features.1.0.block.0.0.weight 的平均映射误差: 0.005890 +[2025-04-11 14:15:47] [INFO] features.1.0.block.0.1.weight 的平均映射误差: 0.035606 +[2025-04-11 14:15:47] [INFO] features.1.1.block.0.0.weight 的平均映射误差: 0.004046 +[2025-04-11 14:15:47] [INFO] features.1.1.block.0.1.weight 的平均映射误差: 0.034027 +[2025-04-11 14:15:47] [INFO] features.1.2.block.0.0.weight 的平均映射误差: 0.003640 +[2025-04-11 14:15:47] [INFO] features.1.2.block.0.1.weight 的平均映射误差: 0.035464 +[2025-04-11 14:15:47] [INFO] features.2.0.block.0.0.weight 的平均映射误差: 0.003260 +[2025-04-11 14:15:47] [INFO] features.2.0.block.0.1.weight 的平均映射误差: 0.035470 +[2025-04-11 14:15:47] [INFO] features.2.0.block.1.0.weight 的平均映射误差: 0.006487 +[2025-04-11 14:15:47] [INFO] features.2.0.block.1.1.weight 的平均映射误差: 0.035480 +[2025-04-11 14:15:47] [INFO] features.2.1.block.0.0.weight 的平均映射误差: 0.001782 +[2025-04-11 14:15:47] [INFO] features.2.1.block.0.1.weight 的平均映射误差: 0.035523 +[2025-04-11 14:15:47] [INFO] features.2.1.block.1.0.weight 的平均映射误差: 0.003041 +[2025-04-11 14:15:47] [INFO] features.2.1.block.1.1.weight 的平均映射误差: 0.037262 +[2025-04-11 14:15:47] [INFO] features.2.2.block.0.0.weight 的平均映射误差: 0.001776 +[2025-04-11 14:15:47] [INFO] features.2.2.block.0.1.weight 的平均映射误差: 0.036067 +[2025-04-11 14:15:47] [INFO] features.2.2.block.1.0.weight 的平均映射误差: 0.002761 +[2025-04-11 14:15:47] [INFO] features.2.2.block.1.1.weight 的平均映射误差: 0.035264 +[2025-04-11 14:15:47] [INFO] features.2.3.block.0.0.weight 的平均映射误差: 0.001791 +[2025-04-11 14:15:47] [INFO] features.2.3.block.0.1.weight 的平均映射误差: 0.036800 +[2025-04-11 14:15:47] [INFO] features.2.3.block.1.0.weight 的平均映射误差: 0.002672 +[2025-04-11 14:15:47] [INFO] features.2.3.block.1.1.weight 的平均映射误差: 0.034887 +[2025-04-11 14:15:47] [INFO] features.2.4.block.0.0.weight 的平均映射误差: 0.001800 +[2025-04-11 14:15:47] [INFO] features.2.4.block.0.1.weight 的平均映射误差: 0.039009 +[2025-04-11 14:15:47] [INFO] features.2.4.block.1.0.weight 的平均映射误差: 0.002604 +[2025-04-11 14:15:47] [INFO] features.2.4.block.1.1.weight 的平均映射误差: 0.036549 +[2025-04-11 14:15:47] [INFO] features.3.0.block.0.0.weight 的平均映射误差: 0.002077 +[2025-04-11 14:15:47] [INFO] features.3.0.block.0.1.weight 的平均映射误差: 0.035429 +[2025-04-11 14:15:47] [INFO] features.3.0.block.1.0.weight 的平均映射误差: 0.003936 +[2025-04-11 14:15:47] [INFO] features.3.0.block.1.1.weight 的平均映射误差: 0.035475 +[2025-04-11 14:15:47] [INFO] features.3.1.block.0.0.weight 的平均映射误差: 0.001614 +[2025-04-11 14:15:47] [INFO] features.3.1.block.0.1.weight 的平均映射误差: 0.037643 +[2025-04-11 14:15:47] [INFO] features.3.1.block.1.0.weight 的平均映射误差: 0.001997 +[2025-04-11 14:15:47] [INFO] features.3.1.block.1.1.weight 的平均映射误差: 0.035900 +[2025-04-11 14:15:47] [INFO] features.3.2.block.0.0.weight 的平均映射误差: 0.001611 +[2025-04-11 14:15:47] [INFO] features.3.2.block.0.1.weight 的平均映射误差: 0.046619 +[2025-04-11 14:15:47] [INFO] features.3.2.block.1.0.weight 的平均映射误差: 0.001936 +[2025-04-11 14:15:47] [INFO] features.3.2.block.1.1.weight 的平均映射误差: 0.035857 +[2025-04-11 14:15:47] [INFO] features.3.3.block.0.0.weight 的平均映射误差: 0.001618 +[2025-04-11 14:15:47] [INFO] features.3.3.block.0.1.weight 的平均映射误差: 0.048430 +[2025-04-11 14:15:47] [INFO] features.3.3.block.1.0.weight 的平均映射误差: 0.001900 +[2025-04-11 14:15:47] [INFO] features.3.3.block.1.1.weight 的平均映射误差: 0.037172 +[2025-04-11 14:15:47] [INFO] features.3.4.block.0.0.weight 的平均映射误差: 0.001610 +[2025-04-11 14:15:47] [INFO] features.3.4.block.0.1.weight 的平均映射误差: 0.040407 +[2025-04-11 14:15:47] [INFO] features.3.4.block.1.0.weight 的平均映射误差: 0.001847 +[2025-04-11 14:15:47] [INFO] features.3.4.block.1.1.weight 的平均映射误差: 0.035280 +[2025-04-11 14:15:47] [INFO] features.4.0.block.0.0.weight 的平均映射误差: 0.003857 +[2025-04-11 14:15:47] [INFO] features.4.0.block.0.1.weight 的平均映射误差: 0.041298 +[2025-04-11 14:15:47] [INFO] features.4.0.block.1.0.weight 的平均映射误差: 0.004809 +[2025-04-11 14:15:47] [INFO] features.4.0.block.1.1.weight 的平均映射误差: 0.047496 +[2025-04-11 14:15:47] [INFO] features.4.0.block.2.fc1.weight 的平均映射误差: 0.001521 +[2025-04-11 14:15:47] [INFO] features.4.0.block.2.fc2.weight 的平均映射误差: 0.001597 +[2025-04-11 14:15:47] [INFO] features.4.0.block.3.0.weight 的平均映射误差: 0.002922 +[2025-04-11 14:15:47] [INFO] features.4.0.block.3.1.weight 的平均映射误差: 0.036359 +[2025-04-11 14:15:47] [INFO] features.4.1.block.0.0.weight 的平均映射误差: 0.001677 +[2025-04-11 14:15:47] [INFO] features.4.1.block.0.1.weight 的平均映射误差: 0.035753 +[2025-04-11 14:15:47] [INFO] features.4.1.block.1.0.weight 的平均映射误差: 0.002718 +[2025-04-11 14:15:47] [INFO] features.4.1.block.1.1.weight 的平均映射误差: 0.036858 +[2025-04-11 14:15:47] [INFO] features.4.1.block.2.fc1.weight 的平均映射误差: 0.001398 +[2025-04-11 14:15:47] [INFO] features.4.1.block.2.fc2.weight 的平均映射误差: 0.002052 +[2025-04-11 14:15:47] [INFO] features.4.1.block.3.0.weight 的平均映射误差: 0.001679 +[2025-04-11 14:15:47] [INFO] features.4.1.block.3.1.weight 的平均映射误差: 0.039053 +[2025-04-11 14:15:47] [INFO] features.4.2.block.0.0.weight 的平均映射误差: 0.001682 +[2025-04-11 14:15:47] [INFO] features.4.2.block.0.1.weight 的平均映射误差: 0.037560 +[2025-04-11 14:15:47] [INFO] features.4.2.block.1.0.weight 的平均映射误差: 0.002701 +[2025-04-11 14:15:47] [INFO] features.4.2.block.1.1.weight 的平均映射误差: 0.036930 +[2025-04-11 14:15:47] [INFO] features.4.2.block.2.fc1.weight 的平均映射误差: 0.001286 +[2025-04-11 14:15:47] [INFO] features.4.2.block.2.fc2.weight 的平均映射误差: 0.001878 +[2025-04-11 14:15:47] [INFO] features.4.2.block.3.0.weight 的平均映射误差: 0.001646 +[2025-04-11 14:15:47] [INFO] features.4.2.block.3.1.weight 的平均映射误差: 0.035867 +[2025-04-11 14:15:47] [INFO] features.4.3.block.0.0.weight 的平均映射误差: 0.001659 +[2025-04-11 14:15:47] [INFO] features.4.3.block.0.1.weight 的平均映射误差: 0.036378 +[2025-04-11 14:15:47] [INFO] features.4.3.block.1.0.weight 的平均映射误差: 0.002571 +[2025-04-11 14:15:47] [INFO] features.4.3.block.1.1.weight 的平均映射误差: 0.036774 +[2025-04-11 14:15:47] [INFO] features.4.3.block.2.fc1.weight 的平均映射误差: 0.000938 +[2025-04-11 14:15:47] [INFO] features.4.3.block.2.fc2.weight 的平均映射误差: 0.001383 +[2025-04-11 14:15:47] [INFO] features.4.3.block.3.0.weight 的平均映射误差: 0.001630 +[2025-04-11 14:15:47] [INFO] features.4.3.block.3.1.weight 的平均映射误差: 0.036895 +[2025-04-11 14:15:47] [INFO] features.4.4.block.0.0.weight 的平均映射误差: 0.001658 +[2025-04-11 14:15:47] [INFO] features.4.4.block.0.1.weight 的平均映射误差: 0.037319 +[2025-04-11 14:15:47] [INFO] features.4.4.block.1.0.weight 的平均映射误差: 0.002598 +[2025-04-11 14:15:47] [INFO] features.4.4.block.1.1.weight 的平均映射误差: 0.034783 +[2025-04-11 14:15:47] [INFO] features.4.4.block.2.fc1.weight 的平均映射误差: 0.000828 +[2025-04-11 14:15:47] [INFO] features.4.4.block.2.fc2.weight 的平均映射误差: 0.000868 +[2025-04-11 14:15:47] [INFO] features.4.4.block.3.0.weight 的平均映射误差: 0.001626 +[2025-04-11 14:15:47] [INFO] features.4.4.block.3.1.weight 的平均映射误差: 0.039247 +[2025-04-11 14:15:47] [INFO] features.4.5.block.0.0.weight 的平均映射误差: 0.001654 +[2025-04-11 14:15:47] [INFO] features.4.5.block.0.1.weight 的平均映射误差: 0.038612 +[2025-04-11 14:15:47] [INFO] features.4.5.block.1.0.weight 的平均映射误差: 0.002319 +[2025-04-11 14:15:47] [INFO] features.4.5.block.1.1.weight 的平均映射误差: 0.035052 +[2025-04-11 14:15:47] [INFO] features.4.5.block.2.fc1.weight 的平均映射误差: 0.000829 +[2025-04-11 14:15:47] [INFO] features.4.5.block.2.fc2.weight 的平均映射误差: 0.000840 +[2025-04-11 14:15:47] [INFO] features.4.5.block.3.0.weight 的平均映射误差: 0.001627 +[2025-04-11 14:15:47] [INFO] features.4.5.block.3.1.weight 的平均映射误差: 0.037723 +[2025-04-11 14:15:47] [INFO] features.4.6.block.0.0.weight 的平均映射误差: 0.001670 +[2025-04-11 14:15:47] [INFO] features.4.6.block.0.1.weight 的平均映射误差: 0.037588 +[2025-04-11 14:15:47] [INFO] features.4.6.block.1.0.weight 的平均映射误差: 0.002228 +[2025-04-11 14:15:47] [INFO] features.4.6.block.1.1.weight 的平均映射误差: 0.042649 +[2025-04-11 14:15:47] [INFO] features.4.6.block.2.fc1.weight 的平均映射误差: 0.000811 +[2025-04-11 14:15:47] [INFO] features.4.6.block.2.fc2.weight 的平均映射误差: 0.000938 +[2025-04-11 14:15:47] [INFO] features.4.6.block.3.0.weight 的平均映射误差: 0.001618 +[2025-04-11 14:15:47] [INFO] features.4.6.block.3.1.weight 的平均映射误差: 0.036893 +[2025-04-11 14:15:47] [INFO] features.5.0.block.0.0.weight 的平均映射误差: 0.002143 +[2025-04-11 14:15:47] [INFO] features.5.0.block.0.1.weight 的平均映射误差: 0.036118 +[2025-04-11 14:15:47] [INFO] features.5.0.block.1.0.weight 的平均映射误差: 0.003581 +[2025-04-11 14:15:47] [INFO] features.5.0.block.1.1.weight 的平均映射误差: 0.039169 +[2025-04-11 14:15:47] [INFO] features.5.0.block.2.fc1.weight 的平均映射误差: 0.001807 +[2025-04-11 14:15:47] [INFO] features.5.0.block.2.fc2.weight 的平均映射误差: 0.002002 +[2025-04-11 14:15:47] [INFO] features.5.0.block.3.0.weight 的平均映射误差: 0.001975 +[2025-04-11 14:15:47] [INFO] features.5.0.block.3.1.weight 的平均映射误差: 0.035505 +[2025-04-11 14:15:47] [INFO] features.5.1.block.0.0.weight 的平均映射误差: 0.001630 +[2025-04-11 14:15:47] [INFO] features.5.1.block.0.1.weight 的平均映射误差: 0.039468 +[2025-04-11 14:15:47] [INFO] features.5.1.block.1.0.weight 的平均映射误差: 0.002178 +[2025-04-11 14:15:47] [INFO] features.5.1.block.1.1.weight 的平均映射误差: 0.040440 +[2025-04-11 14:15:47] [INFO] features.5.1.block.2.fc1.weight 的平均映射误差: 0.000979 +[2025-04-11 14:15:47] [INFO] features.5.1.block.2.fc2.weight 的平均映射误差: 0.001840 +[2025-04-11 14:15:47] [INFO] features.5.1.block.3.0.weight 的平均映射误差: 0.001607 +[2025-04-11 14:15:47] [INFO] features.5.1.block.3.1.weight 的平均映射误差: 0.043088 +[2025-04-11 14:15:47] [INFO] features.5.2.block.0.0.weight 的平均映射误差: 0.001612 +[2025-04-11 14:15:47] [INFO] features.5.2.block.0.1.weight 的平均映射误差: 0.037725 +[2025-04-11 14:15:47] [INFO] features.5.2.block.1.0.weight 的平均映射误差: 0.002096 +[2025-04-11 14:15:47] [INFO] features.5.2.block.1.1.weight 的平均映射误差: 0.039893 +[2025-04-11 14:15:47] [INFO] features.5.2.block.2.fc1.weight 的平均映射误差: 0.000939 +[2025-04-11 14:15:47] [INFO] features.5.2.block.2.fc2.weight 的平均映射误差: 0.001601 +[2025-04-11 14:15:47] [INFO] features.5.2.block.3.0.weight 的平均映射误差: 0.001593 +[2025-04-11 14:15:47] [INFO] features.5.2.block.3.1.weight 的平均映射误差: 0.035678 +[2025-04-11 14:15:47] [INFO] features.5.3.block.0.0.weight 的平均映射误差: 0.001601 +[2025-04-11 14:15:47] [INFO] features.5.3.block.0.1.weight 的平均映射误差: 0.038849 +[2025-04-11 14:15:47] [INFO] features.5.3.block.1.0.weight 的平均映射误差: 0.002018 +[2025-04-11 14:15:47] [INFO] features.5.3.block.1.1.weight 的平均映射误差: 0.041347 +[2025-04-11 14:15:47] [INFO] features.5.3.block.2.fc1.weight 的平均映射误差: 0.000869 +[2025-04-11 14:15:47] [INFO] features.5.3.block.2.fc2.weight 的平均映射误差: 0.001253 +[2025-04-11 14:15:47] [INFO] features.5.3.block.3.0.weight 的平均映射误差: 0.001566 +[2025-04-11 14:15:47] [INFO] features.5.3.block.3.1.weight 的平均映射误差: 0.034256 +[2025-04-11 14:15:47] [INFO] features.5.4.block.0.0.weight 的平均映射误差: 0.001611 +[2025-04-11 14:15:47] [INFO] features.5.4.block.0.1.weight 的平均映射误差: 0.039264 +[2025-04-11 14:15:47] [INFO] features.5.4.block.1.0.weight 的平均映射误差: 0.001988 +[2025-04-11 14:15:47] [INFO] features.5.4.block.1.1.weight 的平均映射误差: 0.042858 +[2025-04-11 14:15:47] [INFO] features.5.4.block.2.fc1.weight 的平均映射误差: 0.000758 +[2025-04-11 14:15:47] [INFO] features.5.4.block.2.fc2.weight 的平均映射误差: 0.001140 +[2025-04-11 14:15:47] [INFO] features.5.4.block.3.0.weight 的平均映射误差: 0.001560 +[2025-04-11 14:15:47] [INFO] features.5.4.block.3.1.weight 的平均映射误差: 0.034366 +[2025-04-11 14:15:47] [INFO] features.5.5.block.0.0.weight 的平均映射误差: 0.001611 +[2025-04-11 14:15:47] [INFO] features.5.5.block.0.1.weight 的平均映射误差: 0.038609 +[2025-04-11 14:15:47] [INFO] features.5.5.block.1.0.weight 的平均映射误差: 0.001954 +[2025-04-11 14:15:47] [INFO] features.5.5.block.1.1.weight 的平均映射误差: 0.043097 +[2025-04-11 14:15:47] [INFO] features.5.5.block.2.fc1.weight 的平均映射误差: 0.000893 +[2025-04-11 14:15:47] [INFO] features.5.5.block.2.fc2.weight 的平均映射误差: 0.000929 +[2025-04-11 14:15:47] [INFO] features.5.5.block.3.0.weight 的平均映射误差: 0.001533 +[2025-04-11 14:15:47] [INFO] features.5.5.block.3.1.weight 的平均映射误差: 0.032236 +[2025-04-11 14:15:47] [INFO] features.5.6.block.0.0.weight 的平均映射误差: 0.001604 +[2025-04-11 14:15:47] [INFO] features.5.6.block.0.1.weight 的平均映射误差: 0.038047 +[2025-04-11 14:15:47] [INFO] features.5.6.block.1.0.weight 的平均映射误差: 0.001843 +[2025-04-11 14:15:47] [INFO] features.5.6.block.1.1.weight 的平均映射误差: 0.044229 +[2025-04-11 14:15:47] [INFO] features.5.6.block.2.fc1.weight 的平均映射误差: 0.000958 +[2025-04-11 14:15:47] [INFO] features.5.6.block.2.fc2.weight 的平均映射误差: 0.001192 +[2025-04-11 14:15:47] [INFO] features.5.6.block.3.0.weight 的平均映射误差: 0.001524 +[2025-04-11 14:15:47] [INFO] features.5.6.block.3.1.weight 的平均映射误差: 0.033243 +[2025-04-11 14:15:47] [INFO] features.5.7.block.0.0.weight 的平均映射误差: 0.001585 +[2025-04-11 14:15:47] [INFO] features.5.7.block.0.1.weight 的平均映射误差: 0.037826 +[2025-04-11 14:15:47] [INFO] features.5.7.block.1.0.weight 的平均映射误差: 0.001887 +[2025-04-11 14:15:47] [INFO] features.5.7.block.1.1.weight 的平均映射误差: 0.044553 +[2025-04-11 14:15:47] [INFO] features.5.7.block.2.fc1.weight 的平均映射误差: 0.000707 +[2025-04-11 14:15:47] [INFO] features.5.7.block.2.fc2.weight 的平均映射误差: 0.000641 +[2025-04-11 14:15:47] [INFO] features.5.7.block.3.0.weight 的平均映射误差: 0.001499 +[2025-04-11 14:15:47] [INFO] features.5.7.block.3.1.weight 的平均映射误差: 0.031182 +[2025-04-11 14:15:47] [INFO] features.5.8.block.0.0.weight 的平均映射误差: 0.001576 +[2025-04-11 14:15:47] [INFO] features.5.8.block.0.1.weight 的平均映射误差: 0.037619 +[2025-04-11 14:15:47] [INFO] features.5.8.block.1.0.weight 的平均映射误差: 0.001825 +[2025-04-11 14:15:47] [INFO] features.5.8.block.1.1.weight 的平均映射误差: 0.044604 +[2025-04-11 14:15:47] [INFO] features.5.8.block.2.fc1.weight 的平均映射误差: 0.000778 +[2025-04-11 14:15:47] [INFO] features.5.8.block.2.fc2.weight 的平均映射误差: 0.000749 +[2025-04-11 14:15:47] [INFO] features.5.8.block.3.0.weight 的平均映射误差: 0.001522 +[2025-04-11 14:15:47] [INFO] features.5.8.block.3.1.weight 的平均映射误差: 0.032814 +[2025-04-11 14:15:47] [INFO] features.5.9.block.0.0.weight 的平均映射误差: 0.001587 +[2025-04-11 14:15:47] [INFO] features.5.9.block.0.1.weight 的平均映射误差: 0.037321 +[2025-04-11 14:15:47] [INFO] features.5.9.block.1.0.weight 的平均映射误差: 0.001797 +[2025-04-11 14:15:47] [INFO] features.5.9.block.1.1.weight 的平均映射误差: 0.045703 +[2025-04-11 14:15:47] [INFO] features.5.9.block.2.fc1.weight 的平均映射误差: 0.000869 +[2025-04-11 14:15:47] [INFO] features.5.9.block.2.fc2.weight 的平均映射误差: 0.000930 +[2025-04-11 14:15:47] [INFO] features.5.9.block.3.0.weight 的平均映射误差: 0.001509 +[2025-04-11 14:15:47] [INFO] features.5.9.block.3.1.weight 的平均映射误差: 0.031491 +[2025-04-11 14:15:47] [INFO] features.5.10.block.0.0.weight 的平均映射误差: 0.001597 +[2025-04-11 14:15:47] [INFO] features.5.10.block.0.1.weight 的平均映射误差: 0.036260 +[2025-04-11 14:15:47] [INFO] features.5.10.block.1.0.weight 的平均映射误差: 0.001875 +[2025-04-11 14:15:47] [INFO] features.5.10.block.1.1.weight 的平均映射误差: 0.044181 +[2025-04-11 14:15:47] [INFO] features.5.10.block.2.fc1.weight 的平均映射误差: 0.000694 +[2025-04-11 14:15:47] [INFO] features.5.10.block.2.fc2.weight 的平均映射误差: 0.000665 +[2025-04-11 14:15:47] [INFO] features.5.10.block.3.0.weight 的平均映射误差: 0.001552 +[2025-04-11 14:15:47] [INFO] features.5.10.block.3.1.weight 的平均映射误差: 0.040860 +[2025-04-11 14:15:47] [INFO] features.5.11.block.0.0.weight 的平均映射误差: 0.001599 +[2025-04-11 14:15:47] [INFO] features.5.11.block.0.1.weight 的平均映射误差: 0.036998 +[2025-04-11 14:15:47] [INFO] features.5.11.block.1.0.weight 的平均映射误差: 0.001869 +[2025-04-11 14:15:47] [INFO] features.5.11.block.1.1.weight 的平均映射误差: 0.046363 +[2025-04-11 14:15:47] [INFO] features.5.11.block.2.fc1.weight 的平均映射误差: 0.000844 +[2025-04-11 14:15:47] [INFO] features.5.11.block.2.fc2.weight 的平均映射误差: 0.000727 +[2025-04-11 14:15:47] [INFO] features.5.11.block.3.0.weight 的平均映射误差: 0.001565 +[2025-04-11 14:15:47] [INFO] features.5.11.block.3.1.weight 的平均映射误差: 0.039384 +[2025-04-11 14:15:47] [INFO] features.5.12.block.0.0.weight 的平均映射误差: 0.001591 +[2025-04-11 14:15:47] [INFO] features.5.12.block.0.1.weight 的平均映射误差: 0.037183 +[2025-04-11 14:15:47] [INFO] features.5.12.block.1.0.weight 的平均映射误差: 0.001795 +[2025-04-11 14:15:47] [INFO] features.5.12.block.1.1.weight 的平均映射误差: 0.047979 +[2025-04-11 14:15:47] [INFO] features.5.12.block.2.fc1.weight 的平均映射误差: 0.000804 +[2025-04-11 14:15:47] [INFO] features.5.12.block.2.fc2.weight 的平均映射误差: 0.000701 +[2025-04-11 14:15:47] [INFO] features.5.12.block.3.0.weight 的平均映射误差: 0.001556 +[2025-04-11 14:15:47] [INFO] features.5.12.block.3.1.weight 的平均映射误差: 0.040410 +[2025-04-11 14:15:47] [INFO] features.5.13.block.0.0.weight 的平均映射误差: 0.001591 +[2025-04-11 14:15:47] [INFO] features.5.13.block.0.1.weight 的平均映射误差: 0.036977 +[2025-04-11 14:15:47] [INFO] features.5.13.block.1.0.weight 的平均映射误差: 0.001808 +[2025-04-11 14:15:47] [INFO] features.5.13.block.1.1.weight 的平均映射误差: 0.046961 +[2025-04-11 14:15:47] [INFO] features.5.13.block.2.fc1.weight 的平均映射误差: 0.000692 +[2025-04-11 14:15:47] [INFO] features.5.13.block.2.fc2.weight 的平均映射误差: 0.000667 +[2025-04-11 14:15:47] [INFO] features.5.13.block.3.0.weight 的平均映射误差: 0.001560 +[2025-04-11 14:15:47] [INFO] features.5.13.block.3.1.weight 的平均映射误差: 0.043202 +[2025-04-11 14:15:47] [INFO] features.6.0.block.0.0.weight 的平均映射误差: 0.002123 +[2025-04-11 14:15:47] [INFO] features.6.0.block.0.1.weight 的平均映射误差: 0.039398 +[2025-04-11 14:15:47] [INFO] features.6.0.block.1.0.weight 的平均映射误差: 0.003175 +[2025-04-11 14:15:47] [INFO] features.6.0.block.1.1.weight 的平均映射误差: 0.039906 +[2025-04-11 14:15:47] [INFO] features.6.0.block.2.fc1.weight 的平均映射误差: 0.000684 +[2025-04-11 14:15:47] [INFO] features.6.0.block.2.fc2.weight 的平均映射误差: 0.000678 +[2025-04-11 14:15:47] [INFO] features.6.0.block.3.0.weight 的平均映射误差: 0.001892 +[2025-04-11 14:15:47] [INFO] features.6.0.block.3.1.weight 的平均映射误差: 0.035403 +[2025-04-11 14:15:47] [INFO] features.6.1.block.0.0.weight 的平均映射误差: 0.001572 +[2025-04-11 14:15:47] [INFO] features.6.1.block.0.1.weight 的平均映射误差: 0.039177 +[2025-04-11 14:15:47] [INFO] features.6.1.block.1.0.weight 的平均映射误差: 0.001994 +[2025-04-11 14:15:47] [INFO] features.6.1.block.1.1.weight 的平均映射误差: 0.039785 +[2025-04-11 14:15:47] [INFO] features.6.1.block.2.fc1.weight 的平均映射误差: 0.000736 +[2025-04-11 14:15:47] [INFO] features.6.1.block.2.fc2.weight 的平均映射误差: 0.001490 +[2025-04-11 14:15:47] [INFO] features.6.1.block.3.0.weight 的平均映射误差: 0.001569 +[2025-04-11 14:15:47] [INFO] features.6.1.block.3.1.weight 的平均映射误差: 0.047751 +[2025-04-11 14:15:47] [INFO] features.6.2.block.0.0.weight 的平均映射误差: 0.001566 +[2025-04-11 14:15:47] [INFO] features.6.2.block.0.1.weight 的平均映射误差: 0.039453 +[2025-04-11 14:15:47] [INFO] features.6.2.block.1.0.weight 的平均映射误差: 0.001995 +[2025-04-11 14:15:47] [INFO] features.6.2.block.1.1.weight 的平均映射误差: 0.040910 +[2025-04-11 14:15:47] [INFO] features.6.2.block.2.fc1.weight 的平均映射误差: 0.000733 +[2025-04-11 14:15:47] [INFO] features.6.2.block.2.fc2.weight 的平均映射误差: 0.001366 +[2025-04-11 14:15:47] [INFO] features.6.2.block.3.0.weight 的平均映射误差: 0.001559 +[2025-04-11 14:15:47] [INFO] features.6.2.block.3.1.weight 的平均映射误差: 0.045482 +[2025-04-11 14:15:47] [INFO] features.6.3.block.0.0.weight 的平均映射误差: 0.001553 +[2025-04-11 14:15:47] [INFO] features.6.3.block.0.1.weight 的平均映射误差: 0.039350 +[2025-04-11 14:15:47] [INFO] features.6.3.block.1.0.weight 的平均映射误差: 0.001940 +[2025-04-11 14:15:47] [INFO] features.6.3.block.1.1.weight 的平均映射误差: 0.044553 +[2025-04-11 14:15:47] [INFO] features.6.3.block.2.fc1.weight 的平均映射误差: 0.000709 +[2025-04-11 14:15:47] [INFO] features.6.3.block.2.fc2.weight 的平均映射误差: 0.001145 +[2025-04-11 14:15:47] [INFO] features.6.3.block.3.0.weight 的平均映射误差: 0.001530 +[2025-04-11 14:15:47] [INFO] features.6.3.block.3.1.weight 的平均映射误差: 0.045242 +[2025-04-11 14:15:47] [INFO] features.6.4.block.0.0.weight 的平均映射误差: 0.001556 +[2025-04-11 14:15:47] [INFO] features.6.4.block.0.1.weight 的平均映射误差: 0.038508 +[2025-04-11 14:15:47] [INFO] features.6.4.block.1.0.weight 的平均映射误差: 0.001892 +[2025-04-11 14:15:47] [INFO] features.6.4.block.1.1.weight 的平均映射误差: 0.045423 +[2025-04-11 14:15:47] [INFO] features.6.4.block.2.fc1.weight 的平均映射误差: 0.000796 +[2025-04-11 14:15:47] [INFO] features.6.4.block.2.fc2.weight 的平均映射误差: 0.001205 +[2025-04-11 14:15:47] [INFO] features.6.4.block.3.0.weight 的平均映射误差: 0.001528 +[2025-04-11 14:15:47] [INFO] features.6.4.block.3.1.weight 的平均映射误差: 0.045637 +[2025-04-11 14:15:47] [INFO] features.6.5.block.0.0.weight 的平均映射误差: 0.001557 +[2025-04-11 14:15:47] [INFO] features.6.5.block.0.1.weight 的平均映射误差: 0.039449 +[2025-04-11 14:15:47] [INFO] features.6.5.block.1.0.weight 的平均映射误差: 0.001847 +[2025-04-11 14:15:47] [INFO] features.6.5.block.1.1.weight 的平均映射误差: 0.045524 +[2025-04-11 14:15:47] [INFO] features.6.5.block.2.fc1.weight 的平均映射误差: 0.000767 +[2025-04-11 14:15:47] [INFO] features.6.5.block.2.fc2.weight 的平均映射误差: 0.001154 +[2025-04-11 14:15:47] [INFO] features.6.5.block.3.0.weight 的平均映射误差: 0.001532 +[2025-04-11 14:15:47] [INFO] features.6.5.block.3.1.weight 的平均映射误差: 0.043949 +[2025-04-11 14:15:47] [INFO] features.6.6.block.0.0.weight 的平均映射误差: 0.001549 +[2025-04-11 14:15:47] [INFO] features.6.6.block.0.1.weight 的平均映射误差: 0.037997 +[2025-04-11 14:15:47] [INFO] features.6.6.block.1.0.weight 的平均映射误差: 0.001848 +[2025-04-11 14:15:47] [INFO] features.6.6.block.1.1.weight 的平均映射误差: 0.047230 +[2025-04-11 14:15:47] [INFO] features.6.6.block.2.fc1.weight 的平均映射误差: 0.000746 +[2025-04-11 14:15:47] [INFO] features.6.6.block.2.fc2.weight 的平均映射误差: 0.000928 +[2025-04-11 14:15:47] [INFO] features.6.6.block.3.0.weight 的平均映射误差: 0.001515 +[2025-04-11 14:15:47] [INFO] features.6.6.block.3.1.weight 的平均映射误差: 0.041541 +[2025-04-11 14:15:47] [INFO] features.6.7.block.0.0.weight 的平均映射误差: 0.001554 +[2025-04-11 14:15:47] [INFO] features.6.7.block.0.1.weight 的平均映射误差: 0.039419 +[2025-04-11 14:15:47] [INFO] features.6.7.block.1.0.weight 的平均映射误差: 0.001852 +[2025-04-11 14:15:47] [INFO] features.6.7.block.1.1.weight 的平均映射误差: 0.048058 +[2025-04-11 14:15:47] [INFO] features.6.7.block.2.fc1.weight 的平均映射误差: 0.000769 +[2025-04-11 14:15:47] [INFO] features.6.7.block.2.fc2.weight 的平均映射误差: 0.001100 +[2025-04-11 14:15:47] [INFO] features.6.7.block.3.0.weight 的平均映射误差: 0.001524 +[2025-04-11 14:15:47] [INFO] features.6.7.block.3.1.weight 的平均映射误差: 0.045279 +[2025-04-11 14:15:47] [INFO] features.6.8.block.0.0.weight 的平均映射误差: 0.001559 +[2025-04-11 14:15:47] [INFO] features.6.8.block.0.1.weight 的平均映射误差: 0.039592 +[2025-04-11 14:15:47] [INFO] features.6.8.block.1.0.weight 的平均映射误差: 0.001833 +[2025-04-11 14:15:47] [INFO] features.6.8.block.1.1.weight 的平均映射误差: 0.047630 +[2025-04-11 14:15:47] [INFO] features.6.8.block.2.fc1.weight 的平均映射误差: 0.000738 +[2025-04-11 14:15:47] [INFO] features.6.8.block.2.fc2.weight 的平均映射误差: 0.000955 +[2025-04-11 14:15:47] [INFO] features.6.8.block.3.0.weight 的平均映射误差: 0.001532 +[2025-04-11 14:15:47] [INFO] features.6.8.block.3.1.weight 的平均映射误差: 0.044245 +[2025-04-11 14:15:47] [INFO] features.6.9.block.0.0.weight 的平均映射误差: 0.001565 +[2025-04-11 14:15:47] [INFO] features.6.9.block.0.1.weight 的平均映射误差: 0.038773 +[2025-04-11 14:15:47] [INFO] features.6.9.block.1.0.weight 的平均映射误差: 0.001822 +[2025-04-11 14:15:47] [INFO] features.6.9.block.1.1.weight 的平均映射误差: 0.043091 +[2025-04-11 14:15:47] [INFO] features.6.9.block.2.fc1.weight 的平均映射误差: 0.000778 +[2025-04-11 14:15:47] [INFO] features.6.9.block.2.fc2.weight 的平均映射误差: 0.000931 +[2025-04-11 14:15:47] [INFO] features.6.9.block.3.0.weight 的平均映射误差: 0.001537 +[2025-04-11 14:15:47] [INFO] features.6.9.block.3.1.weight 的平均映射误差: 0.043794 +[2025-04-11 14:15:47] [INFO] features.6.10.block.0.0.weight 的平均映射误差: 0.001569 +[2025-04-11 14:15:47] [INFO] features.6.10.block.0.1.weight 的平均映射误差: 0.040033 +[2025-04-11 14:15:47] [INFO] features.6.10.block.1.0.weight 的平均映射误差: 0.001818 +[2025-04-11 14:15:47] [INFO] features.6.10.block.1.1.weight 的平均映射误差: 0.042297 +[2025-04-11 14:15:47] [INFO] features.6.10.block.2.fc1.weight 的平均映射误差: 0.000792 +[2025-04-11 14:15:47] [INFO] features.6.10.block.2.fc2.weight 的平均映射误差: 0.000956 +[2025-04-11 14:15:47] [INFO] features.6.10.block.3.0.weight 的平均映射误差: 0.001544 +[2025-04-11 14:15:47] [INFO] features.6.10.block.3.1.weight 的平均映射误差: 0.043862 +[2025-04-11 14:15:47] [INFO] features.6.11.block.0.0.weight 的平均映射误差: 0.001571 +[2025-04-11 14:15:47] [INFO] features.6.11.block.0.1.weight 的平均映射误差: 0.040223 +[2025-04-11 14:15:47] [INFO] features.6.11.block.1.0.weight 的平均映射误差: 0.001820 +[2025-04-11 14:15:47] [INFO] features.6.11.block.1.1.weight 的平均映射误差: 0.042829 +[2025-04-11 14:15:47] [INFO] features.6.11.block.2.fc1.weight 的平均映射误差: 0.000648 +[2025-04-11 14:15:47] [INFO] features.6.11.block.2.fc2.weight 的平均映射误差: 0.000975 +[2025-04-11 14:15:47] [INFO] features.6.11.block.3.0.weight 的平均映射误差: 0.001556 +[2025-04-11 14:15:47] [INFO] features.6.11.block.3.1.weight 的平均映射误差: 0.042950 +[2025-04-11 14:15:47] [INFO] features.6.12.block.0.0.weight 的平均映射误差: 0.001578 +[2025-04-11 14:15:47] [INFO] features.6.12.block.0.1.weight 的平均映射误差: 0.040683 +[2025-04-11 14:15:47] [INFO] features.6.12.block.1.0.weight 的平均映射误差: 0.001772 +[2025-04-11 14:15:47] [INFO] features.6.12.block.1.1.weight 的平均映射误差: 0.036807 +[2025-04-11 14:15:47] [INFO] features.6.12.block.2.fc1.weight 的平均映射误差: 0.000588 +[2025-04-11 14:15:47] [INFO] features.6.12.block.2.fc2.weight 的平均映射误差: 0.000677 +[2025-04-11 14:15:47] [INFO] features.6.12.block.3.0.weight 的平均映射误差: 0.001563 +[2025-04-11 14:15:47] [INFO] features.6.12.block.3.1.weight 的平均映射误差: 0.039227 +[2025-04-11 14:15:47] [INFO] features.6.13.block.0.0.weight 的平均映射误差: 0.001573 +[2025-04-11 14:15:47] [INFO] features.6.13.block.0.1.weight 的平均映射误差: 0.040096 +[2025-04-11 14:15:47] [INFO] features.6.13.block.1.0.weight 的平均映射误差: 0.001760 +[2025-04-11 14:15:47] [INFO] features.6.13.block.1.1.weight 的平均映射误差: 0.042820 +[2025-04-11 14:15:47] [INFO] features.6.13.block.2.fc1.weight 的平均映射误差: 0.000681 +[2025-04-11 14:15:47] [INFO] features.6.13.block.2.fc2.weight 的平均映射误差: 0.001051 +[2025-04-11 14:15:47] [INFO] features.6.13.block.3.0.weight 的平均映射误差: 0.001553 +[2025-04-11 14:15:47] [INFO] features.6.13.block.3.1.weight 的平均映射误差: 0.041267 +[2025-04-11 14:15:47] [INFO] features.6.14.block.0.0.weight 的平均映射误差: 0.001584 +[2025-04-11 14:15:47] [INFO] features.6.14.block.0.1.weight 的平均映射误差: 0.042771 +[2025-04-11 14:15:47] [INFO] features.6.14.block.1.0.weight 的平均映射误差: 0.001726 +[2025-04-11 14:15:47] [INFO] features.6.14.block.1.1.weight 的平均映射误差: 0.040882 +[2025-04-11 14:15:47] [INFO] features.6.14.block.2.fc1.weight 的平均映射误差: 0.000707 +[2025-04-11 14:15:47] [INFO] features.6.14.block.2.fc2.weight 的平均映射误差: 0.000825 +[2025-04-11 14:15:47] [INFO] features.6.14.block.3.0.weight 的平均映射误差: 0.001558 +[2025-04-11 14:15:47] [INFO] features.6.14.block.3.1.weight 的平均映射误差: 0.037846 +[2025-04-11 14:15:47] [INFO] features.6.15.block.0.0.weight 的平均映射误差: 0.001579 +[2025-04-11 14:15:47] [INFO] features.6.15.block.0.1.weight 的平均映射误差: 0.044733 +[2025-04-11 14:15:47] [INFO] features.6.15.block.1.0.weight 的平均映射误差: 0.001702 +[2025-04-11 14:15:47] [INFO] features.6.15.block.1.1.weight 的平均映射误差: 0.040810 +[2025-04-11 14:15:47] [INFO] features.6.15.block.2.fc1.weight 的平均映射误差: 0.000664 +[2025-04-11 14:15:47] [INFO] features.6.15.block.2.fc2.weight 的平均映射误差: 0.000729 +[2025-04-11 14:15:47] [INFO] features.6.15.block.3.0.weight 的平均映射误差: 0.001565 +[2025-04-11 14:15:47] [INFO] features.6.15.block.3.1.weight 的平均映射误差: 0.037453 +[2025-04-11 14:15:47] [INFO] features.6.16.block.0.0.weight 的平均映射误差: 0.001561 +[2025-04-11 14:15:47] [INFO] features.6.16.block.0.1.weight 的平均映射误差: 0.041804 +[2025-04-11 14:15:47] [INFO] features.6.16.block.1.0.weight 的平均映射误差: 0.001678 +[2025-04-11 14:15:47] [INFO] features.6.16.block.1.1.weight 的平均映射误差: 0.048582 +[2025-04-11 14:15:47] [INFO] features.6.16.block.2.fc1.weight 的平均映射误差: 0.000723 +[2025-04-11 14:15:47] [INFO] features.6.16.block.2.fc2.weight 的平均映射误差: 0.000981 +[2025-04-11 14:15:47] [INFO] features.6.16.block.3.0.weight 的平均映射误差: 0.001532 +[2025-04-11 14:15:47] [INFO] features.6.16.block.3.1.weight 的平均映射误差: 0.039843 +[2025-04-11 14:15:47] [INFO] features.6.17.block.0.0.weight 的平均映射误差: 0.001553 +[2025-04-11 14:15:47] [INFO] features.6.17.block.0.1.weight 的平均映射误差: 0.043473 +[2025-04-11 14:15:47] [INFO] features.6.17.block.1.0.weight 的平均映射误差: 0.001663 +[2025-04-11 14:15:47] [INFO] features.6.17.block.1.1.weight 的平均映射误差: 0.049348 +[2025-04-11 14:15:47] [INFO] features.6.17.block.2.fc1.weight 的平均映射误差: 0.000675 +[2025-04-11 14:15:47] [INFO] features.6.17.block.2.fc2.weight 的平均映射误差: 0.001071 +[2025-04-11 14:15:47] [INFO] features.6.17.block.3.0.weight 的平均映射误差: 0.001521 +[2025-04-11 14:15:47] [INFO] features.6.17.block.3.1.weight 的平均映射误差: 0.040201 +[2025-04-11 14:15:47] [INFO] features.7.0.block.0.0.weight 的平均映射误差: 0.001853 +[2025-04-11 14:15:47] [INFO] features.7.0.block.0.1.weight 的平均映射误差: 0.032322 +[2025-04-11 14:15:47] [INFO] features.7.0.block.1.0.weight 的平均映射误差: 0.002048 +[2025-04-11 14:15:47] [INFO] features.7.0.block.1.1.weight 的平均映射误差: 0.033082 +[2025-04-11 14:15:47] [INFO] features.7.0.block.2.fc1.weight 的平均映射误差: 0.001504 +[2025-04-11 14:15:47] [INFO] features.7.0.block.2.fc2.weight 的平均映射误差: 0.001695 +[2025-04-11 14:15:47] [INFO] features.7.0.block.3.0.weight 的平均映射误差: 0.001625 +[2025-04-11 14:15:47] [INFO] features.7.0.block.3.1.weight 的平均映射误差: 0.034974 +[2025-04-11 14:15:47] [INFO] features.7.1.block.0.0.weight 的平均映射误差: 0.001534 +[2025-04-11 14:15:47] [INFO] features.7.1.block.0.1.weight 的平均映射误差: 0.041334 +[2025-04-11 14:15:47] [INFO] features.7.1.block.1.0.weight 的平均映射误差: 0.001756 +[2025-04-11 14:15:47] [INFO] features.7.1.block.1.1.weight 的平均映射误差: 0.040078 +[2025-04-11 14:15:47] [INFO] features.7.1.block.2.fc1.weight 的平均映射误差: 0.001124 +[2025-04-11 14:15:47] [INFO] features.7.1.block.2.fc2.weight 的平均映射误差: 0.001550 +[2025-04-11 14:15:47] [INFO] features.7.1.block.3.0.weight 的平均映射误差: 0.001506 +[2025-04-11 14:15:47] [INFO] features.7.1.block.3.1.weight 的平均映射误差: 0.048265 +[2025-04-11 14:15:47] [INFO] features.7.2.block.0.0.weight 的平均映射误差: 0.001509 +[2025-04-11 14:15:47] [INFO] features.7.2.block.0.1.weight 的平均映射误差: 0.047545 +[2025-04-11 14:15:47] [INFO] features.7.2.block.1.0.weight 的平均映射误差: 0.002202 +[2025-04-11 14:15:47] [INFO] features.7.2.block.1.1.weight 的平均映射误差: 0.044394 +[2025-04-11 14:15:47] [INFO] features.7.2.block.2.fc1.weight 的平均映射误差: 0.000866 +[2025-04-11 14:15:47] [INFO] features.7.2.block.2.fc2.weight 的平均映射误差: 0.001280 +[2025-04-11 14:15:47] [INFO] features.7.2.block.3.0.weight 的平均映射误差: 0.001464 +[2025-04-11 14:15:47] [INFO] features.7.2.block.3.1.weight 的平均映射误差: 0.037660 +[2025-04-11 14:15:47] [INFO] features.7.3.block.0.0.weight 的平均映射误差: 0.001420 +[2025-04-11 14:15:47] [INFO] features.7.3.block.0.1.weight 的平均映射误差: 0.045454 +[2025-04-11 14:15:47] [INFO] features.7.3.block.1.0.weight 的平均映射误差: 0.002457 +[2025-04-11 14:15:47] [INFO] features.7.3.block.1.1.weight 的平均映射误差: 0.039783 +[2025-04-11 14:15:47] [INFO] features.7.3.block.2.fc1.weight 的平均映射误差: 0.000898 +[2025-04-11 14:15:47] [INFO] features.7.3.block.2.fc2.weight 的平均映射误差: 0.001218 +[2025-04-11 14:15:47] [INFO] features.7.3.block.3.0.weight 的平均映射误差: 0.001374 +[2025-04-11 14:15:47] [INFO] features.7.3.block.3.1.weight 的平均映射误差: 0.038010 +[2025-04-11 14:15:47] [INFO] features.7.4.block.0.0.weight 的平均映射误差: 0.001374 +[2025-04-11 14:15:47] [INFO] features.7.4.block.0.1.weight 的平均映射误差: 0.035527 +[2025-04-11 14:15:47] [INFO] features.7.4.block.1.0.weight 的平均映射误差: 0.002129 +[2025-04-11 14:15:47] [INFO] features.7.4.block.1.1.weight 的平均映射误差: 0.034052 +[2025-04-11 14:15:47] [INFO] features.7.4.block.2.fc1.weight 的平均映射误差: 0.001206 +[2025-04-11 14:15:47] [INFO] features.7.4.block.2.fc2.weight 的平均映射误差: 0.001140 +[2025-04-11 14:15:47] [INFO] features.7.4.block.3.0.weight 的平均映射误差: 0.001325 +[2025-04-11 14:15:47] [INFO] features.7.4.block.3.1.weight 的平均映射误差: 0.039857 +[2025-04-11 14:15:47] [INFO] features.8.0.weight 的平均映射误差: 0.001615 +[2025-04-11 14:15:47] [INFO] features.8.1.weight 的平均映射误差: 0.035214 +[2025-04-11 14:15:47] [INFO] classifier.1.weight 的平均映射误差: 0.001788 +[2025-04-11 14:15:47] [INFO] 评估结果 - Loss: 1.3406, Accuracy: 0.2414 +[2025-04-11 14:15:47] [INFO] RRAM映射模型已保存到 checkpoints/rram_mapped_model.pth +[2025-04-11 14:15:47] [INFO] RRAM映射前后精度对比: 原始 0.5862 vs RRAM映射后 0.2414, 变化: -0.3448 +[2025-04-11 14:15:47] [INFO] 开始微调全连接层 (epochs=50, lr=5e-05)... +[2025-04-11 14:15:47] [INFO] 微调过程中的模型将保存到: checkpoints/fine_tune_checkpoints +[2025-04-11 14:15:49] [INFO] Fine-tuning Epoch 1/50 - Train Acc: 0.8521, Val Acc: 0.5172 +[2025-04-11 14:15:50] [INFO] 已保存第 1 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_1.pth +[2025-04-11 14:15:51] [INFO] Fine-tuning Epoch 2/50 - Train Acc: 0.9085, Val Acc: 0.5517 +[2025-04-11 14:15:52] [INFO] 已保存第 2 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_2.pth +[2025-04-11 14:15:53] [INFO] Fine-tuning Epoch 3/50 - Train Acc: 0.9085, Val Acc: 0.6207 +[2025-04-11 14:15:54] [INFO] 已保存第 3 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_3.pth +[2025-04-11 14:15:55] [INFO] Fine-tuning Epoch 4/50 - Train Acc: 0.8873, Val Acc: 0.7586 +[2025-04-11 14:15:56] [INFO] 已保存第 4 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_4.pth +[2025-04-11 14:15:57] [INFO] Fine-tuning Epoch 5/50 - Train Acc: 0.8944, Val Acc: 0.7586 +[2025-04-11 14:15:58] [INFO] 已保存第 5 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_5.pth +[2025-04-11 14:15:59] [INFO] Fine-tuning Epoch 6/50 - Train Acc: 0.9577, Val Acc: 0.8276 +[2025-04-11 14:16:00] [INFO] 已保存第 6 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_6.pth +[2025-04-11 14:16:01] [INFO] Fine-tuning Epoch 7/50 - Train Acc: 0.9014, Val Acc: 0.8621 +[2025-04-11 14:16:02] [INFO] 已保存第 7 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_7.pth +[2025-04-11 14:16:03] [INFO] Fine-tuning Epoch 8/50 - Train Acc: 0.9155, Val Acc: 0.7241 +[2025-04-11 14:16:04] [INFO] 已保存第 8 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_8.pth +[2025-04-11 14:16:05] [INFO] Fine-tuning Epoch 9/50 - Train Acc: 0.9225, Val Acc: 0.5862 +[2025-04-11 14:16:06] [INFO] 已保存第 9 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_9.pth +[2025-04-11 14:16:08] [INFO] Fine-tuning Epoch 10/50 - Train Acc: 0.9648, Val Acc: 0.5862 +[2025-04-11 14:16:08] [INFO] 已保存第 10 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_10.pth +[2025-04-11 14:16:10] [INFO] Fine-tuning Epoch 11/50 - Train Acc: 0.9577, Val Acc: 0.5862 +[2025-04-11 14:16:10] [INFO] 已保存第 11 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_11.pth +[2025-04-11 14:16:12] [INFO] Fine-tuning Epoch 12/50 - Train Acc: 0.9577, Val Acc: 0.6207 +[2025-04-11 14:16:12] [INFO] 已保存第 12 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_12.pth +[2025-04-11 14:16:14] [INFO] Fine-tuning Epoch 13/50 - Train Acc: 0.9789, Val Acc: 0.6207 +[2025-04-11 14:16:15] [INFO] 已保存第 13 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_13.pth +[2025-04-11 14:16:16] [INFO] Fine-tuning Epoch 14/50 - Train Acc: 0.9789, Val Acc: 0.7931 +[2025-04-11 14:16:17] [INFO] 已保存第 14 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_14.pth +[2025-04-11 14:16:18] [INFO] Fine-tuning Epoch 15/50 - Train Acc: 0.9718, Val Acc: 0.9310 +[2025-04-11 14:16:19] [INFO] 已保存第 15 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_15.pth +[2025-04-11 14:16:20] [INFO] Fine-tuning Epoch 16/50 - Train Acc: 0.9577, Val Acc: 0.8276 +[2025-04-11 14:16:21] [INFO] 已保存第 16 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_16.pth +[2025-04-11 14:16:22] [INFO] Fine-tuning Epoch 17/50 - Train Acc: 0.9789, Val Acc: 0.8966 +[2025-04-11 14:16:23] [INFO] 已保存第 17 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_17.pth +[2025-04-11 14:16:24] [INFO] Fine-tuning Epoch 18/50 - Train Acc: 0.9718, Val Acc: 0.8276 +[2025-04-11 14:16:25] [INFO] 已保存第 18 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_18.pth +[2025-04-11 14:16:26] [INFO] Fine-tuning Epoch 19/50 - Train Acc: 0.9718, Val Acc: 0.8276 +[2025-04-11 14:16:27] [INFO] 已保存第 19 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_19.pth +[2025-04-11 14:16:28] [INFO] Fine-tuning Epoch 20/50 - Train Acc: 0.9930, Val Acc: 0.7931 +[2025-04-11 14:16:29] [INFO] 已保存第 20 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_20.pth +[2025-04-11 14:16:30] [INFO] Fine-tuning Epoch 21/50 - Train Acc: 0.9577, Val Acc: 0.7931 +[2025-04-11 14:16:31] [INFO] 已保存第 21 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_21.pth +[2025-04-11 14:16:32] [INFO] Fine-tuning Epoch 22/50 - Train Acc: 0.9930, Val Acc: 0.7931 +[2025-04-11 14:16:33] [INFO] 已保存第 22 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_22.pth +[2025-04-11 14:16:34] [INFO] Fine-tuning Epoch 23/50 - Train Acc: 0.9789, Val Acc: 0.8276 +[2025-04-11 14:16:35] [INFO] 已保存第 23 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_23.pth +[2025-04-11 14:16:36] [INFO] Fine-tuning Epoch 24/50 - Train Acc: 0.9930, Val Acc: 0.8276 +[2025-04-11 14:16:37] [INFO] 已保存第 24 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_24.pth +[2025-04-11 14:16:38] [INFO] Fine-tuning Epoch 25/50 - Train Acc: 0.9859, Val Acc: 0.8621 +[2025-04-11 14:16:39] [INFO] 已保存第 25 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_25.pth +[2025-04-11 14:16:40] [INFO] Fine-tuning Epoch 26/50 - Train Acc: 0.9859, Val Acc: 0.7931 +[2025-04-11 14:16:41] [INFO] 已保存第 26 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_26.pth +[2025-04-11 14:16:43] [INFO] Fine-tuning Epoch 27/50 - Train Acc: 0.9930, Val Acc: 0.7931 +[2025-04-11 14:16:43] [INFO] 已保存第 27 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_27.pth +[2025-04-11 14:16:45] [INFO] Fine-tuning Epoch 28/50 - Train Acc: 1.0000, Val Acc: 0.7586 +[2025-04-11 14:16:45] [INFO] 已保存第 28 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_28.pth +[2025-04-11 14:16:47] [INFO] Fine-tuning Epoch 29/50 - Train Acc: 1.0000, Val Acc: 0.7586 +[2025-04-11 14:16:47] [INFO] 已保存第 29 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_29.pth +[2025-04-11 14:16:49] [INFO] Fine-tuning Epoch 30/50 - Train Acc: 0.9859, Val Acc: 0.7931 +[2025-04-11 14:16:49] [INFO] 已保存第 30 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_30.pth +[2025-04-11 14:16:51] [INFO] Fine-tuning Epoch 31/50 - Train Acc: 0.9930, Val Acc: 0.7931 +[2025-04-11 14:16:52] [INFO] 已保存第 31 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_31.pth +[2025-04-11 14:16:53] [INFO] Fine-tuning Epoch 32/50 - Train Acc: 0.9930, Val Acc: 0.8276 +[2025-04-11 14:16:54] [INFO] 已保存第 32 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_32.pth +[2025-04-11 14:16:55] [INFO] Fine-tuning Epoch 33/50 - Train Acc: 1.0000, Val Acc: 0.9310 +[2025-04-11 14:16:56] [INFO] 已保存第 33 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_33.pth +[2025-04-11 14:16:57] [INFO] Fine-tuning Epoch 34/50 - Train Acc: 1.0000, Val Acc: 0.9310 +[2025-04-11 14:16:58] [INFO] 已保存第 34 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_34.pth +[2025-04-11 14:16:59] [INFO] Fine-tuning Epoch 35/50 - Train Acc: 0.9859, Val Acc: 0.8621 +[2025-04-11 14:17:00] [INFO] 已保存第 35 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_35.pth +[2025-04-11 14:17:01] [INFO] Fine-tuning Epoch 36/50 - Train Acc: 1.0000, Val Acc: 0.8621 +[2025-04-11 14:17:02] [INFO] 已保存第 36 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_36.pth +[2025-04-11 14:17:03] [INFO] Fine-tuning Epoch 37/50 - Train Acc: 0.9718, Val Acc: 0.8966 +[2025-04-11 14:17:04] [INFO] 已保存第 37 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_37.pth +[2025-04-11 14:17:05] [INFO] Fine-tuning Epoch 38/50 - Train Acc: 1.0000, Val Acc: 0.8621 +[2025-04-11 14:17:06] [INFO] 已保存第 38 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_38.pth +[2025-04-11 14:17:07] [INFO] Fine-tuning Epoch 39/50 - Train Acc: 1.0000, Val Acc: 0.8276 +[2025-04-11 14:17:08] [INFO] 已保存第 39 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_39.pth +[2025-04-11 14:17:09] [INFO] Fine-tuning Epoch 40/50 - Train Acc: 1.0000, Val Acc: 0.8276 +[2025-04-11 14:17:10] [INFO] 已保存第 40 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_40.pth +[2025-04-11 14:17:11] [INFO] Fine-tuning Epoch 41/50 - Train Acc: 0.9930, Val Acc: 0.8276 +[2025-04-11 14:17:12] [INFO] 已保存第 41 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_41.pth +[2025-04-11 14:17:14] [INFO] Fine-tuning Epoch 42/50 - Train Acc: 1.0000, Val Acc: 0.8276 +[2025-04-11 14:17:14] [INFO] 已保存第 42 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_42.pth +[2025-04-11 14:17:16] [INFO] Fine-tuning Epoch 43/50 - Train Acc: 1.0000, Val Acc: 0.8276 +[2025-04-11 14:17:16] [INFO] 已保存第 43 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_43.pth +[2025-04-11 14:17:18] [INFO] Fine-tuning Epoch 44/50 - Train Acc: 0.9859, Val Acc: 0.8621 +[2025-04-11 14:17:18] [INFO] 已保存第 44 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_44.pth +[2025-04-11 14:17:20] [INFO] Fine-tuning Epoch 45/50 - Train Acc: 0.9930, Val Acc: 0.8621 +[2025-04-11 14:17:20] [INFO] 已保存第 45 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_45.pth +[2025-04-11 14:17:22] [INFO] Fine-tuning Epoch 46/50 - Train Acc: 0.9859, Val Acc: 0.8621 +[2025-04-11 14:17:22] [INFO] 已保存第 46 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_46.pth +[2025-04-11 14:17:24] [INFO] Fine-tuning Epoch 47/50 - Train Acc: 0.9930, Val Acc: 0.8621 +[2025-04-11 14:17:25] [INFO] 已保存第 47 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_47.pth +[2025-04-11 14:17:26] [INFO] Fine-tuning Epoch 48/50 - Train Acc: 1.0000, Val Acc: 0.8276 +[2025-04-11 14:17:27] [INFO] 已保存第 48 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_48.pth +[2025-04-11 14:17:28] [INFO] Fine-tuning Epoch 49/50 - Train Acc: 0.9789, Val Acc: 0.7931 +[2025-04-11 14:17:29] [INFO] 已保存第 49 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_49.pth +[2025-04-11 14:17:30] [INFO] Fine-tuning Epoch 50/50 - Train Acc: 0.9930, Val Acc: 0.8276 +[2025-04-11 14:17:31] [INFO] 已保存第 50 轮微调模型到: checkpoints/fine_tune_checkpoints/fine_tuned_model_epoch_50.pth +[2025-04-11 14:17:31] [INFO] 评估结果 - Loss: 0.5838, Accuracy: 0.8276 +[2025-04-11 14:17:32] [INFO] 微调模型已保存到 checkpoints/fine_tuned_model.pth +[2025-04-11 14:17:32] [INFO] 微调前后精度对比: RRAM映射 0.2414 vs 微调后 0.8276, 变化: 0.5862 +[2025-04-11 14:17:32] [INFO] 所有处理完成! diff --git a/checkpoints_v2m_part2/base_training_metrics.csv b/checkpoints_v2m_part2/base_training_metrics.csv new file mode 100644 index 0000000000000000000000000000000000000000..21bf13a6e06c377d1a6a1557764604cc258c8488 --- /dev/null +++ b/checkpoints_v2m_part2/base_training_metrics.csv @@ -0,0 +1,51 @@ +epoch,train_loss,train_acc,val_loss,val_acc +1,1.4098,0.2183,1.3692,0.4483 +2,1.3061,0.5352,1.3508,0.4483 +3,1.2258,0.7042,1.3288,0.4483 +4,1.1423,0.7042,1.3089,0.4483 +5,1.0667,0.7042,1.2984,0.4483 +6,0.9744,0.7042,1.2977,0.4483 +7,0.9159,0.7042,1.3075,0.4483 +8,0.8672,0.7042,1.3161,0.4483 +9,0.8349,0.7042,1.3154,0.4483 +10,0.8062,0.7042,1.3091,0.4483 +11,0.7764,0.7042,1.2824,0.4483 +12,0.7459,0.7042,1.2373,0.4483 +13,0.7026,0.7042,1.2288,0.4483 +14,0.6678,0.7042,1.2415,0.4483 +15,0.6473,0.7042,1.2037,0.4483 +16,0.6035,0.7394,1.1331,0.4483 +17,0.5741,0.7535,1.1263,0.4483 +18,0.5583,0.7887,1.1548,0.4483 +19,0.5233,0.8028,1.1269,0.5172 +20,0.5189,0.7746,1.1425,0.3793 +21,0.514,0.7958,1.1988,0.3448 +22,0.5165,0.7817,1.2972,0.3448 +23,0.4809,0.7958,1.1992,0.3448 +24,0.4835,0.7887,1.1661,0.3793 +25,0.4557,0.7958,1.2521,0.3448 +26,0.459,0.8169,1.4326,0.2759 +27,0.443,0.8239,1.4415,0.2759 +28,0.4579,0.831,1.3938,0.3448 +29,0.4221,0.8662,1.2294,0.3448 +30,0.4264,0.831,1.0815,0.4138 +31,0.3995,0.8662,1.054,0.5172 +32,0.398,0.838,1.0917,0.5172 +33,0.3591,0.8803,1.0186,0.5172 +34,0.342,0.8803,1.0275,0.4483 +35,0.3773,0.8592,1.0903,0.4828 +36,0.3629,0.8873,1.1087,0.5172 +37,0.3062,0.8944,1.1035,0.5172 +38,0.3355,0.9085,1.094,0.5172 +39,0.3338,0.8803,1.0815,0.5172 +40,0.3105,0.8803,1.0742,0.5172 +41,0.3438,0.8873,1.0633,0.5862 +42,0.315,0.8944,1.0631,0.5862 +43,0.3168,0.8944,1.0575,0.5862 +44,0.2939,0.9085,1.0698,0.5862 +45,0.3333,0.8662,1.0725,0.5862 +46,0.3176,0.8803,1.0823,0.5862 +47,0.284,0.9225,1.0824,0.5862 +48,0.2919,0.9014,1.0881,0.5862 +49,0.2736,0.9085,1.0791,0.5862 +50,0.3232,0.9014,1.0816,0.5862 diff --git a/checkpoints_v2m_part2/best_model.pth b/checkpoints_v2m_part2/best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..de0a66ba304add0c8e65d873621025ae07cc6985 --- /dev/null +++ b/checkpoints_v2m_part2/best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78751e22cb63a5a042ac070023872589a7fb4f4c968ad810fc15f4056f57d2c1 +size 213030806 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_1.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_1.pth new file mode 100644 index 0000000000000000000000000000000000000000..aaf038d96219e8a280bdf60943a76a551679f70e --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_1.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:047fab83ad6a6d8515341140acad35f4b687edf9ef575e70b8e2d2d6a72b48c8 +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_10.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_10.pth new file mode 100644 index 0000000000000000000000000000000000000000..e3711a3a5416c0085122fe7423683ede0c6b6a98 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:957937f69a332cb5826da2d252e6fd716968f909ed8b30f716731e4c6d1b993c +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_11.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_11.pth new file mode 100644 index 0000000000000000000000000000000000000000..1e5d9b153196c75ebee53000ce25193d7773d64f --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_11.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:179ac13c408fec0c6accdf5c1030dce2e3f514d1c6a3867df01e671074f947b4 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_12.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_12.pth new file mode 100644 index 0000000000000000000000000000000000000000..3b6b00b582a25b56d2c0c327a1026961987ec23b --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04622fc1e063aeff91f69a9717ec83317d9ed068a846a2c5de15817a7cfb6ea1 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_13.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_13.pth new file mode 100644 index 0000000000000000000000000000000000000000..08cfa0097f39dd9ab1ed0ac4cd67516dca8536c6 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_13.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f93260b95a17e30a35abc99ece75d7c4406abbc3ceed7400639185c11f670a66 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_14.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_14.pth new file mode 100644 index 0000000000000000000000000000000000000000..62eb8c13f3cb56c32ec10d94c8b0940087e033b8 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_14.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9f3c79bea7159c01716d5591ec7aa74603f295ce5edc0b7841fb26b0f79a7aa +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_15.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_15.pth new file mode 100644 index 0000000000000000000000000000000000000000..74d022eab8c9dff45a7d2bdf257738eb767a87f5 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e23c9ae623993207c66676e8cde4c8a3781c10313dea0470f9db8643abb04472 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_16.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_16.pth new file mode 100644 index 0000000000000000000000000000000000000000..b2679bec79dc5260cc6c9b09fd6931781f651a03 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_16.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd4e6b980b1ae10596e5cd742d27138e3fb3a55dac7bacd373996fb6764a9ef0 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_17.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_17.pth new file mode 100644 index 0000000000000000000000000000000000000000..ea7d26d2ac78d5df06cde6813e013c96fef8ec37 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:044cc72a3899421f111ff7a3b4fcd120f0f8948c154424b8ebd72b3d51873f0b +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_18.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_18.pth new file mode 100644 index 0000000000000000000000000000000000000000..772d3e4d6db316326498208168c6d870c82e0314 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_18.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:576d53b63d65fc4e1d3f64392abd6fc89ef3548f36ecb7365ca7353ba4f9f3c0 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_19.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_19.pth new file mode 100644 index 0000000000000000000000000000000000000000..0805576bc42192ec1bd4cb1d93f386856a200bc2 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_19.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79646d557f11d44a954d8cd2e8bf8a701f9200ebef4be565578e90a7199d1514 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_2.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_2.pth new file mode 100644 index 0000000000000000000000000000000000000000..da9d81502d773c904e6e5c125552d55007804e0b --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca18e17547d4e761f1817b1fca4f8a7858255a6396d95d782608faf0e3a8e682 +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_20.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_20.pth new file mode 100644 index 0000000000000000000000000000000000000000..8706ae9a7812d15b53e21ac83ba6c47845203e79 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7dbc9b91d8d8f2036c12d761f41d04a698382693c036f65064d247a6fa93368d +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_21.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_21.pth new file mode 100644 index 0000000000000000000000000000000000000000..831c06bc84701a2ee9a523f0bda3b4443734db6b --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_21.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:608e74c5353fb02cb72eaab4feb8091d5a23b48180c94f42f18e50346cb6ab86 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_22.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_22.pth new file mode 100644 index 0000000000000000000000000000000000000000..faea5a4a338e52ed2afd4180da8a7a8dbeb4382a --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_22.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c2914e6ec12a3028371f4a9668245fd91565939975cc7ba57b821103a073bc69 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_23.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_23.pth new file mode 100644 index 0000000000000000000000000000000000000000..b438bd810cf0559d2dc2e8918c6d5af64b18c7ef --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_23.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:385a5b88699c52d076ada1573a4de750c74a7adb3dd201fc6bfa970105cedcfe +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_24.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_24.pth new file mode 100644 index 0000000000000000000000000000000000000000..ec8e6906bbef95eb98f89cbb81e07533d2cad001 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_24.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6c8e80f7e421d6a2d4d920d079db47fdb7cd5c48dfa767060f6a1f8bcc0864b +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_25.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_25.pth new file mode 100644 index 0000000000000000000000000000000000000000..c844942d6f7405959ef760aff3a82458a6ca9d9f --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_25.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37974d475c255dcd69ea8b246b6e59c639553880823b488b8c81661ba0eee1ce +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_26.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_26.pth new file mode 100644 index 0000000000000000000000000000000000000000..17b402819d7d6d265e29df5d5b31296c44fd58e7 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_26.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bac25a5fe8047abc79ce220e74fde458a17d147ce7f8bd59caedb4388b493581 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_27.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_27.pth new file mode 100644 index 0000000000000000000000000000000000000000..33b7db142b5a1dc298c0097d799e438fe6d8891c --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_27.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc6d649f1c0e417f6528a97106c26ae26efe8a9a4683cffe52c596055d176e5d +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_28.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_28.pth new file mode 100644 index 0000000000000000000000000000000000000000..91a5afb7081337b4e2633d38c9f079b497a821a8 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_28.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a869d96d4601cdfacd352dd5cbc80188e3664fcb4fc66cb00b8dc5656e604b4 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_29.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_29.pth new file mode 100644 index 0000000000000000000000000000000000000000..65d683ffe4603b478aaa9ff0a49dcfe9182f5f8f --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_29.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b04e8af6f20227d43cb6bc0978b5da74d5cef6cf11dced91f1753da41fe32325 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_3.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_3.pth new file mode 100644 index 0000000000000000000000000000000000000000..6f9ab4c47c891d2af205d26533d412a14821a606 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_3.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ffbf3dd2a45fbc73e071f9ff0e48131d51d0514d137b79724129d8f595b2e654 +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_30.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_30.pth new file mode 100644 index 0000000000000000000000000000000000000000..f42757e4c0e7310ccabbebb235431543291cc2f0 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_30.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdba1bb54e1f3a9775d9ed46d45c824ae5377efa43e015082291c227c6b50f64 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_31.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_31.pth new file mode 100644 index 0000000000000000000000000000000000000000..fe87a71eabd455b2358d804adfec1ef6b8c1b09d --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_31.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14ba672839ab07771061aa2fd69cc08c090428141c42cd148e5b596c65032d85 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_32.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_32.pth new file mode 100644 index 0000000000000000000000000000000000000000..5e618a94fe80cfd830835840c5884b05d4fb29ba --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_32.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:326c202459a0605c926476fbcaa9a3c256732e8c3a23249c9e8c9e07322019d4 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_33.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_33.pth new file mode 100644 index 0000000000000000000000000000000000000000..fb2de4a6288c3bbc8c0c3fcc27eb777f967cc308 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_33.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:049a20faaf3e82358025c797748f0871b14b23513c49ae536ef82da7c625d551 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_34.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_34.pth new file mode 100644 index 0000000000000000000000000000000000000000..1062b7dbcc82edac1909aecc7ba99de831d5bd21 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_34.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:465cc2185424cc619425b13ed5b892e24741027660c882e080d676a82ec2cde2 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_35.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_35.pth new file mode 100644 index 0000000000000000000000000000000000000000..081fe55e0c6e4b15610c0071981aa03248b02468 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_35.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3dd22186cd6845aec5196f882a3ef9d8afb58d31603b9434bc6d34f91cf7619c +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_36.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_36.pth new file mode 100644 index 0000000000000000000000000000000000000000..9440bcaf7a112787712b81bf96e13ab5ea8c1517 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_36.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf42ff2c19a78714456e98082175793b9f818f76aceef34c128c1b20e66c2a98 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_37.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_37.pth new file mode 100644 index 0000000000000000000000000000000000000000..388eb7b30f00aeb504f0aac0ad4d7fbfc46ab73c --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_37.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7831bae97470dfc53008197548fa822790de753603830b065da0c210f6991c6c +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_38.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_38.pth new file mode 100644 index 0000000000000000000000000000000000000000..24b510c83b46934e8105503bc74dc5e21eeda510 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_38.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23137e59e1e266f8625ffed8e5efb58d25cbaecac2f3e7d8466e5f7ea3a542e2 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_39.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_39.pth new file mode 100644 index 0000000000000000000000000000000000000000..97200f7b752be8e9e4b7afa8bed38e2c1a6207ad --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_39.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c32bcf000740d246f45a63d10b483529beb29b4e129e6a77bfc1c7f12e34742c +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_4.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_4.pth new file mode 100644 index 0000000000000000000000000000000000000000..fdff6d4139456346aefd17f8647286f4fcb167a7 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_4.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ef265af1dd2c93709918fd360d55e6b0f33b6cc6a27a0277c4158820a727a63 +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_40.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_40.pth new file mode 100644 index 0000000000000000000000000000000000000000..044513db5265e0dbe8dd34f79c1c812b633eb4b3 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_40.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a990a491bf0caafb6701a01a39eed217fb855532a8a44a7954a4046f9d217199 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_41.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_41.pth new file mode 100644 index 0000000000000000000000000000000000000000..7c179290b740e2828f6f959c61d87000b5ee560e --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_41.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6fd99c82fb73d533096afe018b551b7414083cbb7b6c08d3a6f715e8658ec1c1 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_42.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_42.pth new file mode 100644 index 0000000000000000000000000000000000000000..3eff4c5e329e8d852bfa29b0f468e26bc75bfc7b --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_42.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1195dcef7a024d70147dc9f906521b4850aacd3b7a6974680206ecac92ebabd +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_43.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_43.pth new file mode 100644 index 0000000000000000000000000000000000000000..54d9513c182662990775453eed4aa97971dc5d56 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_43.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5dd572acdf08c1d775b6ef29826cf84d9bbb8abbe002da8fffb627571ead5c9c +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_44.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_44.pth new file mode 100644 index 0000000000000000000000000000000000000000..e7b33d266ca8a6b2b69bb3c273587fc2007302cb --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_44.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35809474efcaa30c6137bf5b639a111da86dd0e3c63eb1f3b951502ef728f05f +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_45.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_45.pth new file mode 100644 index 0000000000000000000000000000000000000000..5f6a4c95c1034a75120cf626c4ec63a7a29277ee --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_45.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8e01d6aaa786aae4d54e6519ef23849b6e87f7b8de9ff7a12631fb96887ff36 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_46.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_46.pth new file mode 100644 index 0000000000000000000000000000000000000000..469f1264e51f0fa45df5adecb09ee7ade308aa02 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_46.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15f6e92a983e391498e82968502ba0a182e0c92363c1211b36131bf9c04c534c +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_47.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_47.pth new file mode 100644 index 0000000000000000000000000000000000000000..0ce4f1e42add88df167a1bdec62e97fd38658115 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_47.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a116bb683a9fc8a1764758ce027debc884dbb34a257bc39230a5cec049fa56ad +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_48.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_48.pth new file mode 100644 index 0000000000000000000000000000000000000000..2c463ed916357d8d436aebce57df35a9678ca7c0 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_48.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86cb6261cd86751c6d27b52757e0148c1ea51368ab634ffa785f2daebdd24465 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_49.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_49.pth new file mode 100644 index 0000000000000000000000000000000000000000..55083922d01d0b56eaed6b27fcd4761c69bb4b5d --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_49.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80f0390fe67fbcfdf9fad89cbd8f1997744da3140286a226e4b5caae5e047ccf +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_5.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_5.pth new file mode 100644 index 0000000000000000000000000000000000000000..0cd81e4c11c9c13bad7fc43a4cbcbb72fd5deb77 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c938b3862af43195fc89e76847065ee3269c5c529d2ebbeb88001ca6178e8a3 +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_50.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_50.pth new file mode 100644 index 0000000000000000000000000000000000000000..696b188ee7a6960719fcefaa8b9268243837c465 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_50.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:68316d1c101376a78d0423a2015e81283dcfcee13260a6368104b8862f6ec672 +size 636546301 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_6.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_6.pth new file mode 100644 index 0000000000000000000000000000000000000000..6cb784e9bc31aeac6bf7ad4f8f56d80206f63d92 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_6.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5ded0ccf33a0a62ef1496bb75f0172ff1a219b4d1dd8bd95e0ce79b1a5d3692b +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_7.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_7.pth new file mode 100644 index 0000000000000000000000000000000000000000..0b02ce1e7b6282fc9aaf8e661be502439b48c689 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ff8ef6c74019c45e5e83ed83151cac24569a13c3fc0d768d8cebe37a6207049 +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_8.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_8.pth new file mode 100644 index 0000000000000000000000000000000000000000..ab69f9740f9a4b231e737bc537255d69c2d2d6f7 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_8.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1aa9130bd71b8e5fdb97f13520886fec88d39a7250b6de44e5c1b364edac11bc +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_9.pth b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_9.pth new file mode 100644 index 0000000000000000000000000000000000000000..0b3fb3f922deae2ffca84160b794e61901ec0d9c --- /dev/null +++ b/checkpoints_v2m_part2/fine_tune_checkpoints/fine_tuned_model_epoch_9.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d045f03e47fc9fac227868bf9b1f7e4d3c6658afb9d93a55f1122003041ce32 +size 636543230 diff --git a/checkpoints_v2m_part2/fine_tuned_model.pth b/checkpoints_v2m_part2/fine_tuned_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..ab40318e755a6397a0c505978464b75b2352e221 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tuned_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8561f4aab9fa3e866b4be16fee2d7ac8d5b824946f2fa187dbb686d46f94a853 +size 213037550 diff --git a/checkpoints_v2m_part2/fine_tuning_metrics.csv b/checkpoints_v2m_part2/fine_tuning_metrics.csv new file mode 100644 index 0000000000000000000000000000000000000000..509d370a33e8294c598d72eb9a3e8f01ed349754 --- /dev/null +++ b/checkpoints_v2m_part2/fine_tuning_metrics.csv @@ -0,0 +1,51 @@ +epoch,train_acc,val_acc +1,0.8521,0.5172 +2,0.9085,0.5517 +3,0.9085,0.6207 +4,0.8873,0.7586 +5,0.8944,0.7586 +6,0.9577,0.8276 +7,0.9014,0.8621 +8,0.9155,0.7241 +9,0.9225,0.5862 +10,0.9648,0.5862 +11,0.9577,0.5862 +12,0.9577,0.6207 +13,0.9789,0.6207 +14,0.9789,0.7931 +15,0.9718,0.931 +16,0.9577,0.8276 +17,0.9789,0.8966 +18,0.9718,0.8276 +19,0.9718,0.8276 +20,0.993,0.7931 +21,0.9577,0.7931 +22,0.993,0.7931 +23,0.9789,0.8276 +24,0.993,0.8276 +25,0.9859,0.8621 +26,0.9859,0.7931 +27,0.993,0.7931 +28,1.0,0.7586 +29,1.0,0.7586 +30,0.9859,0.7931 +31,0.993,0.7931 +32,0.993,0.8276 +33,1.0,0.931 +34,1.0,0.931 +35,0.9859,0.8621 +36,1.0,0.8621 +37,0.9718,0.8966 +38,1.0,0.8621 +39,1.0,0.8276 +40,1.0,0.8276 +41,0.993,0.8276 +42,1.0,0.8276 +43,1.0,0.8276 +44,0.9859,0.8621 +45,0.993,0.8621 +46,0.9859,0.8621 +47,0.993,0.8621 +48,1.0,0.8276 +49,0.9789,0.7931 +50,0.993,0.8276 diff --git a/checkpoints_v2m_part2/last_checkpoint.pth b/checkpoints_v2m_part2/last_checkpoint.pth new file mode 100644 index 0000000000000000000000000000000000000000..64deae20efca7bb5daf229a487e6b665bf1a9878 --- /dev/null +++ b/checkpoints_v2m_part2/last_checkpoint.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d58d646e1790485d62b732fe27d1767037ecff698afb90f8d3a1ba6c56b5566f +size 636511239 diff --git a/checkpoints_v2m_part2/parse_log.py b/checkpoints_v2m_part2/parse_log.py new file mode 100644 index 0000000000000000000000000000000000000000..208fef3d710149081abb114bc7d53ce454a536f9 --- /dev/null +++ b/checkpoints_v2m_part2/parse_log.py @@ -0,0 +1,62 @@ +import re +import pandas as pd + +def parse_log_file(log_file_path): + # 初始化存储数据的列表 + base_training_data = [] + fine_tuning_data = [] + + with open(log_file_path, 'r', encoding='utf-8') as f: + content = f.read() + + # 提取基础训练数据 + base_pattern = r'Epoch (\d+)/50 - Train Loss: ([\d.]+), Train Acc: ([\d.]+), Val Loss: ([\d.]+), Val Acc: ([\d.]+)' + base_matches = re.finditer(base_pattern, content) + + for match in base_matches: + epoch = int(match.group(1)) + train_loss = float(match.group(2)) + train_acc = float(match.group(3)) + val_loss = float(match.group(4)) + val_acc = float(match.group(5)) + + # 如果epoch小于等于50，认为是基础训练数据 + if epoch <= 50: + base_training_data.append({ + 'epoch': epoch, + 'train_loss': train_loss, + 'train_acc': train_acc, + 'val_loss': val_loss, + 'val_acc': val_acc + }) + + # 提取微调训练数据 + fine_tune_pattern = r'Fine-tuning Epoch (\d+)/50 - Train Acc: ([\d.]+), Val Acc: ([\d.]+)' + fine_tune_matches = re.finditer(fine_tune_pattern, content) + + for match in fine_tune_matches: + epoch = int(match.group(1)) + train_acc = float(match.group(2)) + val_acc = float(match.group(3)) + + fine_tuning_data.append({ + 'epoch': epoch, + 'train_acc': train_acc, + 'val_acc': val_acc + }) + + # 转换为DataFrame并保存为CSV + if base_training_data: + base_df = pd.DataFrame(base_training_data) + base_df.to_csv('base_training_metrics.csv', index=False) + print(f"基础训练数据已保存到 base_training_metrics.csv，共 {len(base_training_data)} 条记录") + + if fine_tuning_data: + fine_tune_df = pd.DataFrame(fine_tuning_data) + fine_tune_df.to_csv('fine_tuning_metrics.csv', index=False) + print(f"微调训练数据已保存到 fine_tuning_metrics.csv，共 {len(fine_tuning_data)} 条记录") + +if __name__ == '__main__': + # 指定日志文件路径 + log_file_path = '2025-04-11_14-13-49_train.log' + parse_log_file(log_file_path) \ No newline at end of file diff --git a/checkpoints_v2m_part2/rram_mapped_model.pth b/checkpoints_v2m_part2/rram_mapped_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..853df36152117c4a5eb3207800769736bbf844c6 --- /dev/null +++ b/checkpoints_v2m_part2/rram_mapped_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ce36a6f4aa209d3476ba1077c768ecd96d3a3f58a43cc41b8f721e5dce9dd1c +size 213038674 diff --git a/checkpoints_v2m_part2/training_plot.png b/checkpoints_v2m_part2/training_plot.png new file mode 100644 index 0000000000000000000000000000000000000000..911cba63928e5026ae3216b26475b63ea956a5e0 Binary files /dev/null and b/checkpoints_v2m_part2/training_plot.png differ diff --git a/checkpoints_v2m_part2/visualizations/base_weights_heatmap.png b/checkpoints_v2m_part2/visualizations/base_weights_heatmap.png new file mode 100644 index 0000000000000000000000000000000000000000..435b85eb985e57277a0fb9802f1acb09179e5c00 --- /dev/null +++ b/checkpoints_v2m_part2/visualizations/base_weights_heatmap.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ab2130f3257e6a8a366d01cea3f30821924a0233f01623fb67446b431e2427c +size 115263 diff --git a/checkpoints_v2m_part2/visualizations/fine_tuned_weights_heatmap.png b/checkpoints_v2m_part2/visualizations/fine_tuned_weights_heatmap.png new file mode 100644 index 0000000000000000000000000000000000000000..b14271de17995ffd4dd41e25fb2f54f268c3aaa7 --- /dev/null +++ b/checkpoints_v2m_part2/visualizations/fine_tuned_weights_heatmap.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf440603221049bd1e44c5e20f0696aedb0b8f07603c779bf8bc3bd0fdd9a421 +size 119112 diff --git a/checkpoints_v2m_part2/visualizations/mapping_error_distribution.png b/checkpoints_v2m_part2/visualizations/mapping_error_distribution.png new file mode 100644 index 0000000000000000000000000000000000000000..8aa29d21576ca4285b82e67a4168fcc932479b35 --- /dev/null +++ b/checkpoints_v2m_part2/visualizations/mapping_error_distribution.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ee7de98e832ee564a268f5ced850c67b1265c315a678e3d33150113e38af4fe +size 120943 diff --git a/checkpoints_v2m_part2/visualizations/weight_changes_heatmap.png b/checkpoints_v2m_part2/visualizations/weight_changes_heatmap.png new file mode 100644 index 0000000000000000000000000000000000000000..b4c68737105f9526c4a8a3381da50f0001627679 --- /dev/null +++ b/checkpoints_v2m_part2/visualizations/weight_changes_heatmap.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c38945c3a9654bfb16a53e66a1d169991d30c5167f79ab7631da6473fbcf365a +size 118141