End of training

README.md

@@ -2,6 +2,8 @@
 license: apache-2.0
 base_model: IDEA-Research/grounding-dino-tiny
 tags:
+- object-detection
+- vision
 - generated_from_trainer
 model-index:
 - name: grounding-dino-tiny-finetuned-cppe-5-10k-steps
@@ -13,7 +15,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # grounding-dino-tiny-finetuned-cppe-5-10k-steps
 
-This model is a fine-tuned version of [IDEA-Research/grounding-dino-tiny](https://huggingface.co/IDEA-Research/grounding-dino-tiny) on an unknown dataset.
+This model is a fine-tuned version of [IDEA-Research/grounding-dino-tiny](https://huggingface.co/IDEA-Research/grounding-dino-tiny) on the cppe-5 dataset.
 It achieves the following results on the evaluation set:
 - Loss: 5.5317
 - Map: 0.0151

all_results.json

@@ -0,0 +1,34 @@
+{
+    "epoch": 10.0,
+    "test_loss": 4.769590377807617,
+    "test_map": 0.0362,
+    "test_map_50": 0.0515,
+    "test_map_75": 0.0389,
+    "test_map_Coverall": 0.181,
+    "test_map_Face_Shield": 0.0,
+    "test_map_Gloves": 0.0,
+    "test_map_Goggles": 0.0,
+    "test_map_Mask": 0.0,
+    "test_map_large": 0.0462,
+    "test_map_medium": 0.0376,
+    "test_map_small": 0.1143,
+    "test_mar_1": 0.0221,
+    "test_mar_10": 0.1204,
+    "test_mar_100": 0.1421,
+    "test_mar_100_Coverall": 0.7106,
+    "test_mar_100_Face_Shield": 0.0,
+    "test_mar_100_Gloves": 0.0,
+    "test_mar_100_Goggles": 0.0,
+    "test_mar_100_Mask": 0.0,
+    "test_mar_large": 0.1588,
+    "test_mar_medium": 0.1544,
+    "test_mar_small": 0.1667,
+    "test_runtime": 9.2109,
+    "test_samples_per_second": 3.148,
+    "test_steps_per_second": 3.148,
+    "total_flos": 1.946309821216161e+19,
+    "train_loss": 841.0750103975183,
+    "train_runtime": 4302.0466,
+    "train_samples_per_second": 1.976,
+    "train_steps_per_second": 1.976
+}

runs/Aug09_04-42-08_a44b5742ff80/events.out.tfevents.1723182963.a44b5742ff80.6692.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24f7f7a2c73ccfafca67527ccc3d8535bd844684d4fae4a40fdffa87e7e35f1e
+size 1548

test_results.json

@@ -0,0 +1,29 @@
+{
+    "epoch": 10.0,
+    "test_loss": 4.769590377807617,
+    "test_map": 0.0362,
+    "test_map_50": 0.0515,
+    "test_map_75": 0.0389,
+    "test_map_Coverall": 0.181,
+    "test_map_Face_Shield": 0.0,
+    "test_map_Gloves": 0.0,
+    "test_map_Goggles": 0.0,
+    "test_map_Mask": 0.0,
+    "test_map_large": 0.0462,
+    "test_map_medium": 0.0376,
+    "test_map_small": 0.1143,
+    "test_mar_1": 0.0221,
+    "test_mar_10": 0.1204,
+    "test_mar_100": 0.1421,
+    "test_mar_100_Coverall": 0.7106,
+    "test_mar_100_Face_Shield": 0.0,
+    "test_mar_100_Gloves": 0.0,
+    "test_mar_100_Goggles": 0.0,
+    "test_mar_100_Mask": 0.0,
+    "test_mar_large": 0.1588,
+    "test_mar_medium": 0.1544,
+    "test_mar_small": 0.1667,
+    "test_runtime": 9.2109,
+    "test_samples_per_second": 3.148,
+    "test_steps_per_second": 3.148
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 10.0,
+    "total_flos": 1.946309821216161e+19,
+    "train_loss": 841.0750103975183,
+    "train_runtime": 4302.0466,
+    "train_samples_per_second": 1.976,
+    "train_steps_per_second": 1.976
+}

trainer_state.json

@@ -0,0 +1,412 @@
+{
+  "best_metric": 0.0182,
+  "best_model_checkpoint": "grounding-dino-tiny-finetuned-cppe-5-10k-steps/checkpoint-7650",
+  "epoch": 10.0,
+  "eval_steps": 500,
+  "global_step": 8500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0,
+      "grad_norm": 64.9486312866211,
+      "learning_rate": 4.5e-05,
+      "loss": 8355.9482,
+      "step": 850
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 6.613674163818359,
+      "eval_map": 0.014,
+      "eval_map_50": 0.0272,
+      "eval_map_75": 0.0134,
+      "eval_map_Coverall": 0.0702,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0243,
+      "eval_map_medium": 0.0111,
+      "eval_map_small": 0.0048,
+      "eval_mar_1": 0.0149,
+      "eval_mar_10": 0.0893,
+      "eval_mar_100": 0.1073,
+      "eval_mar_100_Coverall": 0.5366,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1328,
+      "eval_mar_medium": 0.0889,
+      "eval_mar_small": 0.0523,
+      "eval_runtime": 39.8427,
+      "eval_samples_per_second": 3.765,
+      "eval_steps_per_second": 3.765,
+      "step": 850
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 49.31262969970703,
+      "learning_rate": 4e-05,
+      "loss": 6.7523,
+      "step": 1700
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 6.2356672286987305,
+      "eval_map": 0.0162,
+      "eval_map_50": 0.0302,
+      "eval_map_75": 0.0148,
+      "eval_map_Coverall": 0.0809,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0192,
+      "eval_map_medium": 0.0189,
+      "eval_map_small": 0.0106,
+      "eval_mar_1": 0.0247,
+      "eval_mar_10": 0.0894,
+      "eval_mar_100": 0.107,
+      "eval_mar_100_Coverall": 0.535,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1258,
+      "eval_mar_medium": 0.0968,
+      "eval_mar_small": 0.0643,
+      "eval_runtime": 38.5396,
+      "eval_samples_per_second": 3.892,
+      "eval_steps_per_second": 3.892,
+      "step": 1700
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 112.90940856933594,
+      "learning_rate": 3.5e-05,
+      "loss": 6.5566,
+      "step": 2550
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 6.08903169631958,
+      "eval_map": 0.0158,
+      "eval_map_50": 0.0294,
+      "eval_map_75": 0.0134,
+      "eval_map_Coverall": 0.0791,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0215,
+      "eval_map_medium": 0.0199,
+      "eval_map_small": 0.01,
+      "eval_mar_1": 0.0222,
+      "eval_mar_10": 0.0876,
+      "eval_mar_100": 0.1065,
+      "eval_mar_100_Coverall": 0.5323,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1324,
+      "eval_mar_medium": 0.0846,
+      "eval_mar_small": 0.0671,
+      "eval_runtime": 37.4163,
+      "eval_samples_per_second": 4.009,
+      "eval_steps_per_second": 4.009,
+      "step": 2550
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 24.23554039001465,
+      "learning_rate": 3e-05,
+      "loss": 6.2217,
+      "step": 3400
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 5.902782440185547,
+      "eval_map": 0.0144,
+      "eval_map_50": 0.0271,
+      "eval_map_75": 0.0134,
+      "eval_map_Coverall": 0.0721,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0225,
+      "eval_map_medium": 0.0096,
+      "eval_map_small": 0.0066,
+      "eval_mar_1": 0.0232,
+      "eval_mar_10": 0.0857,
+      "eval_mar_100": 0.107,
+      "eval_mar_100_Coverall": 0.5348,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1397,
+      "eval_mar_medium": 0.0823,
+      "eval_mar_small": 0.06,
+      "eval_runtime": 38.3987,
+      "eval_samples_per_second": 3.906,
+      "eval_steps_per_second": 3.906,
+      "step": 3400
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 103.00000762939453,
+      "learning_rate": 2.5e-05,
+      "loss": 6.0963,
+      "step": 4250
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 5.841100692749023,
+      "eval_map": 0.0126,
+      "eval_map_50": 0.0215,
+      "eval_map_75": 0.014,
+      "eval_map_Coverall": 0.0631,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0178,
+      "eval_map_medium": 0.0138,
+      "eval_map_small": 0.0055,
+      "eval_mar_1": 0.0201,
+      "eval_mar_10": 0.0811,
+      "eval_mar_100": 0.1052,
+      "eval_mar_100_Coverall": 0.5258,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1377,
+      "eval_mar_medium": 0.0942,
+      "eval_mar_small": 0.044,
+      "eval_runtime": 37.9192,
+      "eval_samples_per_second": 3.956,
+      "eval_steps_per_second": 3.956,
+      "step": 4250
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 100.59229278564453,
+      "learning_rate": 2e-05,
+      "loss": 5.996,
+      "step": 5100
+    },
+    {
+      "epoch": 6.0,
+      "eval_loss": 5.724414825439453,
+      "eval_map": 0.0162,
+      "eval_map_50": 0.0311,
+      "eval_map_75": 0.0166,
+      "eval_map_Coverall": 0.0812,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0221,
+      "eval_map_medium": 0.0145,
+      "eval_map_small": 0.0059,
+      "eval_mar_1": 0.0223,
+      "eval_mar_10": 0.0869,
+      "eval_mar_100": 0.1088,
+      "eval_mar_100_Coverall": 0.5437,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1328,
+      "eval_mar_medium": 0.0919,
+      "eval_mar_small": 0.0667,
+      "eval_runtime": 37.3147,
+      "eval_samples_per_second": 4.02,
+      "eval_steps_per_second": 4.02,
+      "step": 5100
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 121.10063934326172,
+      "learning_rate": 1.5e-05,
+      "loss": 5.8971,
+      "step": 5950
+    },
+    {
+      "epoch": 7.0,
+      "eval_loss": 5.547349452972412,
+      "eval_map": 0.0154,
+      "eval_map_50": 0.027,
+      "eval_map_75": 0.016,
+      "eval_map_Coverall": 0.0769,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0208,
+      "eval_map_medium": 0.014,
+      "eval_map_small": 0.01,
+      "eval_mar_1": 0.0244,
+      "eval_mar_10": 0.0946,
+      "eval_mar_100": 0.1154,
+      "eval_mar_100_Coverall": 0.5772,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1311,
+      "eval_mar_medium": 0.106,
+      "eval_mar_small": 0.084,
+      "eval_runtime": 37.3478,
+      "eval_samples_per_second": 4.016,
+      "eval_steps_per_second": 4.016,
+      "step": 5950
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 67.22107696533203,
+      "learning_rate": 1e-05,
+      "loss": 5.7451,
+      "step": 6800
+    },
+    {
+      "epoch": 8.0,
+      "eval_loss": 5.523053169250488,
+      "eval_map": 0.0146,
+      "eval_map_50": 0.0267,
+      "eval_map_75": 0.0148,
+      "eval_map_Coverall": 0.0731,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0183,
+      "eval_map_medium": 0.0161,
+      "eval_map_small": 0.0021,
+      "eval_mar_1": 0.0256,
+      "eval_mar_10": 0.0905,
+      "eval_mar_100": 0.1125,
+      "eval_mar_100_Coverall": 0.5624,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.128,
+      "eval_mar_medium": 0.1062,
+      "eval_mar_small": 0.0325,
+      "eval_runtime": 37.1808,
+      "eval_samples_per_second": 4.034,
+      "eval_steps_per_second": 4.034,
+      "step": 6800
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 40.62353515625,
+      "learning_rate": 5e-06,
+      "loss": 5.7931,
+      "step": 7650
+    },
+    {
+      "epoch": 9.0,
+      "eval_loss": 5.518974781036377,
+      "eval_map": 0.0182,
+      "eval_map_50": 0.032,
+      "eval_map_75": 0.0195,
+      "eval_map_Coverall": 0.091,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0249,
+      "eval_map_medium": 0.0147,
+      "eval_map_small": 0.0175,
+      "eval_mar_1": 0.0299,
+      "eval_mar_10": 0.1048,
+      "eval_mar_100": 0.1138,
+      "eval_mar_100_Coverall": 0.5688,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1309,
+      "eval_mar_medium": 0.0945,
+      "eval_mar_small": 0.08,
+      "eval_runtime": 36.2829,
+      "eval_samples_per_second": 4.134,
+      "eval_steps_per_second": 4.134,
+      "step": 7650
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 38.14755630493164,
+      "learning_rate": 0.0,
+      "loss": 5.7435,
+      "step": 8500
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 5.531744956970215,
+      "eval_map": 0.0151,
+      "eval_map_50": 0.0275,
+      "eval_map_75": 0.0157,
+      "eval_map_Coverall": 0.0755,
+      "eval_map_Face_Shield": 0.0,
+      "eval_map_Gloves": 0.0,
+      "eval_map_Goggles": 0.0,
+      "eval_map_Mask": 0.0,
+      "eval_map_large": 0.0236,
+      "eval_map_medium": 0.0149,
+      "eval_map_small": 0.0125,
+      "eval_mar_1": 0.0202,
+      "eval_mar_10": 0.0902,
+      "eval_mar_100": 0.1127,
+      "eval_mar_100_Coverall": 0.5636,
+      "eval_mar_100_Face_Shield": 0.0,
+      "eval_mar_100_Gloves": 0.0,
+      "eval_mar_100_Goggles": 0.0,
+      "eval_mar_100_Mask": 0.0,
+      "eval_mar_large": 0.1461,
+      "eval_mar_medium": 0.0975,
+      "eval_mar_small": 0.0815,
+      "eval_runtime": 36.9405,
+      "eval_samples_per_second": 4.061,
+      "eval_steps_per_second": 4.061,
+      "step": 8500
+    },
+    {
+      "epoch": 10.0,
+      "step": 8500,
+      "total_flos": 1.946309821216161e+19,
+      "train_loss": 841.0750103975183,
+      "train_runtime": 4302.0466,
+      "train_samples_per_second": 1.976,
+      "train_steps_per_second": 1.976
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 8500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.946309821216161e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}