Training in progress, epoch 3, checkpoint

checkpoint-321/config.json

@@ -1,69 +1,69 @@
-{
-  "_name_or_path": "microsoft/conditional-detr-resnet-50",
-  "activation_dropout": 0.0,
-  "activation_function": "relu",
-  "architectures": [
-    "ConditionalDetrForObjectDetection"
-  ],
-  "attention_dropout": 0.0,
-  "auxiliary_loss": false,
-  "backbone": "resnet50",
-  "backbone_config": null,
-  "backbone_kwargs": {
-    "in_chans": 3,
-    "out_indices": [
-      1,
-      2,
-      3,
-      4
-    ]
-  },
-  "bbox_cost": 5,
-  "bbox_loss_coefficient": 5,
-  "class_cost": 2,
-  "cls_loss_coefficient": 2,
-  "d_model": 256,
-  "decoder_attention_heads": 8,
-  "decoder_ffn_dim": 2048,
-  "decoder_layerdrop": 0.0,
-  "decoder_layers": 6,
-  "dice_loss_coefficient": 1,
-  "dilation": false,
-  "dropout": 0.1,
-  "encoder_attention_heads": 8,
-  "encoder_ffn_dim": 2048,
-  "encoder_layerdrop": 0.0,
-  "encoder_layers": 6,
-  "focal_alpha": 0.25,
-  "giou_cost": 2,
-  "giou_loss_coefficient": 2,
-  "id2label": {
-    "0": "Coverall",
-    "1": "Face_Shield",
-    "2": "Gloves",
-    "3": "Goggles",
-    "4": "Mask"
-  },
-  "init_std": 0.02,
-  "init_xavier_std": 1.0,
-  "is_encoder_decoder": true,
-  "label2id": {
-    "Coverall": 0,
-    "Face_Shield": 1,
-    "Gloves": 2,
-    "Goggles": 3,
-    "Mask": 4
-  },
-  "mask_loss_coefficient": 1,
-  "max_position_embeddings": 1024,
-  "model_type": "conditional_detr",
-  "num_channels": 3,
-  "num_hidden_layers": 6,
-  "num_queries": 300,
-  "position_embedding_type": "sine",
-  "scale_embedding": false,
-  "torch_dtype": "float32",
-  "transformers_version": "4.42.0.dev0",
-  "use_pretrained_backbone": true,
-  "use_timm_backbone": true
-}
+{
+  "_name_or_path": "microsoft/conditional-detr-resnet-50",
+  "activation_dropout": 0.0,
+  "activation_function": "relu",
+  "architectures": [
+    "ConditionalDetrForObjectDetection"
+  ],
+  "attention_dropout": 0.0,
+  "auxiliary_loss": false,
+  "backbone": "resnet50",
+  "backbone_config": null,
+  "backbone_kwargs": {
+    "in_chans": 3,
+    "out_indices": [
+      1,
+      2,
+      3,
+      4
+    ]
+  },
+  "bbox_cost": 5,
+  "bbox_loss_coefficient": 5,
+  "class_cost": 2,
+  "cls_loss_coefficient": 2,
+  "d_model": 256,
+  "decoder_attention_heads": 8,
+  "decoder_ffn_dim": 2048,
+  "decoder_layerdrop": 0.0,
+  "decoder_layers": 6,
+  "dice_loss_coefficient": 1,
+  "dilation": false,
+  "dropout": 0.1,
+  "encoder_attention_heads": 8,
+  "encoder_ffn_dim": 2048,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 6,
+  "focal_alpha": 0.25,
+  "giou_cost": 2,
+  "giou_loss_coefficient": 2,
+  "id2label": {
+    "0": "Coverall",
+    "1": "Face_Shield",
+    "2": "Gloves",
+    "3": "Goggles",
+    "4": "Mask"
+  },
+  "init_std": 0.02,
+  "init_xavier_std": 1.0,
+  "is_encoder_decoder": true,
+  "label2id": {
+    "Coverall": 0,
+    "Face_Shield": 1,
+    "Gloves": 2,
+    "Goggles": 3,
+    "Mask": 4
+  },
+  "mask_loss_coefficient": 1,
+  "max_position_embeddings": 1024,
+  "model_type": "conditional_detr",
+  "num_channels": 3,
+  "num_hidden_layers": 6,
+  "num_queries": 300,
+  "position_embedding_type": "sine",
+  "scale_embedding": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.42.4",
+  "use_pretrained_backbone": true,
+  "use_timm_backbone": true
+}

checkpoint-321/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:76d0ff4b1ea76eb3ec3e0411fd01609809cb81aa851c57d52a9714e4fb64db2d
+oid sha256:583ece8d2bfbe4fd13b1551a2a9fe45fb1b4990040abbc8f75b4dd8c791d20d2
 size 174079796

checkpoint-321/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a2536dab90bd0bfbd1c0f9995306c715beab6be8ed62600ebf8e8ffce892dfdb
-size 345687577
+oid sha256:1c225e7161717287996b39e83c403a512ab161420c33cd2cabc2c0e9eb3eaa12
+size 345697817

checkpoint-321/preprocessor_config.json

@@ -1,47 +1,26 @@
-{
-  "_valid_processor_keys": [
-    "images",
-    "annotations",
-    "return_segmentation_masks",
-    "masks_path",
-    "do_resize",
-    "size",
-    "resample",
-    "do_rescale",
-    "rescale_factor",
-    "do_normalize",
-    "do_convert_annotations",
-    "image_mean",
-    "image_std",
-    "do_pad",
-    "pad_size",
-    "format",
-    "return_tensors",
-    "data_format",
-    "input_data_format"
-  ],
-  "do_convert_annotations": true,
-  "do_normalize": true,
-  "do_pad": false,
-  "do_rescale": true,
-  "do_resize": false,
-  "format": "coco_detection",
-  "image_mean": [
-    0.485,
-    0.456,
-    0.406
-  ],
-  "image_processor_type": "ConditionalDetrImageProcessor",
-  "image_std": [
-    0.229,
-    0.224,
-    0.225
-  ],
-  "pad_size": null,
-  "resample": 2,
-  "rescale_factor": 0.00392156862745098,
-  "size": {
-    "longest_edge": 1333,
-    "shortest_edge": 800
-  }
-}
+{
+  "do_convert_annotations": true,
+  "do_normalize": true,
+  "do_pad": false,
+  "do_rescale": true,
+  "do_resize": false,
+  "format": "coco_detection",
+  "image_mean": [
+    0.485,
+    0.456,
+    0.406
+  ],
+  "image_processor_type": "ConditionalDetrImageProcessor",
+  "image_std": [
+    0.229,
+    0.224,
+    0.225
+  ],
+  "pad_size": null,
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "longest_edge": 1333,
+    "shortest_edge": 800
+  }
+}

checkpoint-321/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48e70d0dd8b84db3bee13caed6af307e1b89afc4eaaa119639b18d898ef62d7b
-size 13990
+oid sha256:991d7fcdff7714d622ac25c68137510df78ee30b6289b34c2f970a9d61d76ad8
+size 14244

checkpoint-321/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a3db8d80a8d54189a5635393973d53fc0522a4fdba8e55972fed72a60864cb39
+oid sha256:8d588360be02d945ff9ca59d4ae5a51f89ec2a61c99732097ff4d6f0b59eb795
 size 1064

checkpoint-321/trainer_state.json

@@ -1,123 +1,213 @@
-{
-  "best_metric": 0.0008,
-  "best_model_checkpoint": "test_detr_finetuned_cppe5/checkpoint-214",
-  "epoch": 3.0,
-  "eval_steps": 500,
-  "global_step": 321,
-  "is_hyper_param_search": false,
-  "is_local_process_zero": true,
-  "is_world_process_zero": true,
-  "log_history": [
-    {
-      "epoch": 1.0,
-      "eval_loss": 3.742539167404175,
-      "eval_map": 0.0,
-      "eval_map_50": 0.0001,
-      "eval_map_75": 0.0,
-      "eval_map_Coverall": 0.0,
-      "eval_map_Face_Shield": 0.0,
-      "eval_map_Gloves": 0.0,
-      "eval_map_Goggles": 0.0,
-      "eval_map_Mask": 0.0,
-      "eval_map_large": 0.0,
-      "eval_map_medium": 0.0,
-      "eval_map_small": 0.001,
-      "eval_mar_1": 0.0001,
-      "eval_mar_10": 0.0011,
-      "eval_mar_100": 0.0053,
-      "eval_mar_100_Coverall": 0.0,
-      "eval_mar_100_Face_Shield": 0.0,
-      "eval_mar_100_Gloves": 0.0031,
-      "eval_mar_100_Goggles": 0.0046,
-      "eval_mar_100_Mask": 0.0187,
-      "eval_mar_large": 0.0021,
-      "eval_mar_medium": 0.0121,
-      "eval_mar_small": 0.0007,
-      "eval_runtime": 41.4133,
-      "eval_samples_per_second": 3.622,
-      "eval_steps_per_second": 0.459,
-      "step": 107
-    },
-    {
-      "epoch": 2.0,
-      "eval_loss": 3.383169174194336,
-      "eval_map": 0.0008,
-      "eval_map_50": 0.003,
-      "eval_map_75": 0.0001,
-      "eval_map_Coverall": 0.0039,
-      "eval_map_Face_Shield": 0.0,
-      "eval_map_Gloves": 0.0,
-      "eval_map_Goggles": 0.0,
-      "eval_map_Mask": 0.0,
-      "eval_map_large": 0.0009,
-      "eval_map_medium": 0.0,
-      "eval_map_small": 0.0,
-      "eval_mar_1": 0.0039,
-      "eval_mar_10": 0.0119,
-      "eval_mar_100": 0.0161,
-      "eval_mar_100_Coverall": 0.0797,
-      "eval_mar_100_Face_Shield": 0.0,
-      "eval_mar_100_Gloves": 0.0009,
-      "eval_mar_100_Goggles": 0.0,
-      "eval_mar_100_Mask": 0.0,
-      "eval_mar_large": 0.0216,
-      "eval_mar_medium": 0.0,
-      "eval_mar_small": 0.0,
-      "eval_runtime": 36.6404,
-      "eval_samples_per_second": 4.094,
-      "eval_steps_per_second": 0.519,
-      "step": 214
-    },
-    {
-      "epoch": 3.0,
-      "eval_loss": 3.414217472076416,
-      "eval_map": 0.0008,
-      "eval_map_50": 0.0038,
-      "eval_map_75": 0.0,
-      "eval_map_Coverall": 0.0041,
-      "eval_map_Face_Shield": 0.0,
-      "eval_map_Gloves": 0.0,
-      "eval_map_Goggles": 0.0,
-      "eval_map_Mask": 0.0,
-      "eval_map_large": 0.001,
-      "eval_map_medium": 0.0,
-      "eval_map_small": 0.0,
-      "eval_mar_1": 0.0047,
-      "eval_mar_10": 0.0119,
-      "eval_mar_100": 0.0246,
-      "eval_mar_100_Coverall": 0.1077,
-      "eval_mar_100_Face_Shield": 0.0,
-      "eval_mar_100_Gloves": 0.0049,
-      "eval_mar_100_Goggles": 0.0,
-      "eval_mar_100_Mask": 0.0102,
-      "eval_mar_large": 0.0246,
-      "eval_mar_medium": 0.0069,
-      "eval_mar_small": 0.0,
-      "eval_runtime": 37.1515,
-      "eval_samples_per_second": 4.038,
-      "eval_steps_per_second": 0.511,
-      "step": 321
-    }
-  ],
-  "logging_steps": 500,
-  "max_steps": 10700,
-  "num_input_tokens_seen": 0,
-  "num_train_epochs": 100,
-  "save_steps": 500,
-  "stateful_callbacks": {
-    "TrainerControl": {
-      "args": {
-        "should_epoch_stop": false,
-        "should_evaluate": false,
-        "should_log": false,
-        "should_save": true,
-        "should_training_stop": false
-      },
-      "attributes": {}
-    }
-  },
-  "total_flos": 4.5811929694464e+17,
-  "train_batch_size": 8,
-  "trial_name": null,
-  "trial_params": null
-}
+{
+  "best_metric": 0.0539,
+  "best_model_checkpoint": "test_detr_finetuned_cppe5\\checkpoint-321",
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 321,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0,
+      "step": 107,
+      "train_loss": 2.531707286834717,
+      "train_map": 0.0033,
+      "train_map_50": 0.0117,
+      "train_map_75": 0.0011,
+      "train_map_Coverall": 0.0032,
+      "train_map_Face_Shield": 0.0001,
+      "train_map_Gloves": 0.0049,
+      "train_map_Goggles": 0.0,
+      "train_map_Mask": 0.0081,
+      "train_map_large": 0.0063,
+      "train_map_medium": 0.0037,
+      "train_map_small": 0.0017,
+      "train_mar_1": 0.0094,
+      "train_mar_10": 0.0727,
+      "train_mar_100": 0.1382,
+      "train_mar_100_Coverall": 0.2272,
+      "train_mar_100_Face_Shield": 0.0201,
+      "train_mar_100_Gloves": 0.199,
+      "train_mar_100_Goggles": 0.0,
+      "train_mar_100_Mask": 0.2445,
+      "train_mar_large": 0.0983,
+      "train_mar_medium": 0.1363,
+      "train_mar_small": 0.0346,
+      "train_runtime": 29.266,
+      "train_samples_per_second": 29.044,
+      "train_steps_per_second": 3.656
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 2.618133068084717,
+      "eval_map": 0.0023,
+      "eval_map_50": 0.0084,
+      "eval_map_75": 0.0006,
+      "eval_map_Coverall": 0.0025,
+      "eval_map_Face_Shield": 0.0002,
+      "eval_map_Gloves": 0.0026,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0062,
+      "eval_map_large": 0.0075,
+      "eval_map_medium": 0.0029,
+      "eval_map_small": 0.0019,
+      "eval_mar_1": 0.0082,
+      "eval_mar_10": 0.0496,
+      "eval_mar_100": 0.1009,
+      "eval_mar_100_Coverall": 0.1667,
+      "eval_mar_100_Face_Shield": 0.0152,
+      "eval_mar_100_Gloves": 0.1442,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.1782,
+      "eval_mar_large": 0.1089,
+      "eval_mar_medium": 0.1196,
+      "eval_mar_small": 0.0338,
+      "eval_runtime": 10.5496,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 1.801,
+      "step": 107
+    },
+    {
+      "epoch": 2.0,
+      "step": 214,
+      "train_loss": 2.1007583141326904,
+      "train_map": 0.0405,
+      "train_map_50": 0.0973,
+      "train_map_75": 0.0262,
+      "train_map_Coverall": 0.0862,
+      "train_map_Face_Shield": 0.0034,
+      "train_map_Gloves": 0.0161,
+      "train_map_Goggles": 0.0014,
+      "train_map_Mask": 0.0954,
+      "train_map_large": 0.0499,
+      "train_map_medium": 0.0351,
+      "train_map_small": 0.0156,
+      "train_mar_1": 0.1017,
+      "train_mar_10": 0.2491,
+      "train_mar_100": 0.2902,
+      "train_mar_100_Coverall": 0.5555,
+      "train_mar_100_Face_Shield": 0.1718,
+      "train_mar_100_Gloves": 0.2798,
+      "train_mar_100_Goggles": 0.0571,
+      "train_mar_100_Mask": 0.3865,
+      "train_mar_large": 0.2549,
+      "train_mar_medium": 0.2432,
+      "train_mar_small": 0.1161,
+      "train_runtime": 26.5597,
+      "train_samples_per_second": 32.003,
+      "train_steps_per_second": 4.029
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 2.251506805419922,
+      "eval_map": 0.0386,
+      "eval_map_50": 0.0946,
+      "eval_map_75": 0.0261,
+      "eval_map_Coverall": 0.1234,
+      "eval_map_Face_Shield": 0.005,
+      "eval_map_Gloves": 0.0102,
+      "eval_map_Goggles": 0.0011,
+      "eval_map_Mask": 0.0531,
+      "eval_map_large": 0.0645,
+      "eval_map_medium": 0.0251,
+      "eval_map_small": 0.0109,
+      "eval_mar_1": 0.0744,
+      "eval_mar_10": 0.1938,
+      "eval_mar_100": 0.2324,
+      "eval_mar_100_Coverall": 0.4653,
+      "eval_mar_100_Face_Shield": 0.1722,
+      "eval_mar_100_Gloves": 0.2165,
+      "eval_mar_100_Goggles": 0.0277,
+      "eval_mar_100_Mask": 0.2804,
+      "eval_mar_large": 0.3155,
+      "eval_mar_medium": 0.2153,
+      "eval_mar_small": 0.0878,
+      "eval_runtime": 9.8776,
+      "eval_samples_per_second": 15.186,
+      "eval_steps_per_second": 1.924,
+      "step": 214
+    },
+    {
+      "epoch": 3.0,
+      "step": 321,
+      "train_loss": 1.8027496337890625,
+      "train_map": 0.0689,
+      "train_map_50": 0.1515,
+      "train_map_75": 0.053,
+      "train_map_Coverall": 0.1899,
+      "train_map_Face_Shield": 0.0098,
+      "train_map_Gloves": 0.0287,
+      "train_map_Goggles": 0.0123,
+      "train_map_Mask": 0.104,
+      "train_map_large": 0.0958,
+      "train_map_medium": 0.0492,
+      "train_map_small": 0.0201,
+      "train_mar_1": 0.1643,
+      "train_mar_10": 0.3427,
+      "train_mar_100": 0.3901,
+      "train_mar_100_Coverall": 0.6673,
+      "train_mar_100_Face_Shield": 0.3124,
+      "train_mar_100_Gloves": 0.3556,
+      "train_mar_100_Goggles": 0.2114,
+      "train_mar_100_Mask": 0.4038,
+      "train_mar_large": 0.4421,
+      "train_mar_medium": 0.3085,
+      "train_mar_small": 0.1399,
+      "train_runtime": 25.8607,
+      "train_samples_per_second": 32.868,
+      "train_steps_per_second": 4.138
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 1.9757874011993408,
+      "eval_map": 0.0539,
+      "eval_map_50": 0.1302,
+      "eval_map_75": 0.0411,
+      "eval_map_Coverall": 0.1643,
+      "eval_map_Face_Shield": 0.0092,
+      "eval_map_Gloves": 0.0261,
+      "eval_map_Goggles": 0.0098,
+      "eval_map_Mask": 0.0601,
+      "eval_map_large": 0.1194,
+      "eval_map_medium": 0.0401,
+      "eval_map_small": 0.0231,
+      "eval_mar_1": 0.1144,
+      "eval_mar_10": 0.2597,
+      "eval_mar_100": 0.304,
+      "eval_mar_100_Coverall": 0.5284,
+      "eval_mar_100_Face_Shield": 0.243,
+      "eval_mar_100_Gloves": 0.2942,
+      "eval_mar_100_Goggles": 0.1446,
+      "eval_mar_100_Mask": 0.3098,
+      "eval_mar_large": 0.4992,
+      "eval_mar_medium": 0.259,
+      "eval_mar_small": 0.1355,
+      "eval_runtime": 9.6967,
+      "eval_samples_per_second": 15.469,
+      "eval_steps_per_second": 1.959,
+      "step": 321
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 10700,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 100,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.5811929694464e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-321/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f0dd550d512ad8d899ef002c48d14a56a1062c99a3d88fb14a426d30f565c622
+oid sha256:aed3eb8d6db83a01a4da94998232d2b334cd1e141aeaa94f3c5559fdda758c8e
 size 5176