Upload checkpoint-1035 with 0.9414 accuracy

Browse files

Files changed (7) hide show

.gitattributes +1 -0
README.md +45 -0
config.json +62 -0
model.safetensors +3 -0
model_card.json +15 -0
preprocessor_config.json +27 -0
trainer_state.json +354 -0

.gitattributes ADDED Viewed

	@@ -0,0 +1 @@


1	+ model.safetensors filter=lfs diff=lfs merge=lfs -text

README.md ADDED Viewed

	@@ -0,0 +1,45 @@

+---
+license: apache-2.0
+tags:
+- image-classification
+- dinov2
+- roadwork-detection
+- natix-subnet
+---
+# DINOv2-Large Roadwork Detector
+Fine-tuned DINOv2-large model for roadwork detection on Natix subnet.
+## Model Details
+- **Base Model**: facebook/dinov2-large
+- **Checkpoint**: checkpoint-1035
+- **Submitted By**: 5HT5XkhAg9jTzJLedi16L7uXnMTAy5CCvJd1YjaUy39gsETG
+- **Submission Time**: 2026-02-09 00:25:58
+## Performance Metrics
+- **Accuracy**: 0.9414
+- **Precision**: 0.9654
+- **Recall**: 0.9616
+- **F1 Score**: 0.9635
+## Usage
+```python
+from transformers import AutoImageProcessor, AutoModelForImageClassification
+from PIL import Image
+processor = AutoImageProcessor.from_pretrained("YOUR_USERNAME/YOUR_REPO")
+model = AutoModelForImageClassification.from_pretrained("YOUR_USERNAME/YOUR_REPO")
+image = Image.open("path/to/image.jpg")
+inputs = processor(images=image, return_tensors="pt")
+outputs = model(**inputs)
+predictions = outputs.logits.softmax(dim=-1)
+```
+## Training
+Trained on Natix roadwork dataset using fine-tuning approach.

config.json ADDED Viewed

	@@ -0,0 +1,62 @@

+{
+  "apply_layernorm": true,
+  "architectures": [
+    "Dinov2ForImageClassification"
+  ],
+  "attention_probs_dropout_prob": 0.0,
+  "drop_path_rate": 0.0,
+  "dtype": "float32",
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.0,
+  "hidden_size": 1024,
+  "image_size": 518,
+  "initializer_range": 0.02,
+  "layer_norm_eps": 1e-06,
+  "layerscale_value": 1.0,
+  "mlp_ratio": 4,
+  "model_type": "dinov2",
+  "num_attention_heads": 16,
+  "num_channels": 3,
+  "num_hidden_layers": 24,
+  "out_features": [
+    "stage24"
+  ],
+  "out_indices": [
+    24
+  ],
+  "patch_size": 14,
+  "problem_type": "single_label_classification",
+  "qkv_bias": true,
+  "reshape_hidden_states": true,
+  "stage_names": [
+    "stem",
+    "stage1",
+    "stage2",
+    "stage3",
+    "stage4",
+    "stage5",
+    "stage6",
+    "stage7",
+    "stage8",
+    "stage9",
+    "stage10",
+    "stage11",
+    "stage12",
+    "stage13",
+    "stage14",
+    "stage15",
+    "stage16",
+    "stage17",
+    "stage18",
+    "stage19",
+    "stage20",
+    "stage21",
+    "stage22",
+    "stage23",
+    "stage24"
+  ],
+  "transformers_version": "5.1.0",
+  "use_cache": false,
+  "use_mask_token": true,
+  "use_swiglu_ffn": false
+}

model.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:fb62eb68f403ea620881907f01fda20b4d88bfa4a815f1015fd4cf39ed5ace66
+size 1217542512

model_card.json ADDED Viewed

	@@ -0,0 +1,15 @@

+{
+  "model_name": "DINOv2-Large-Roadwork-Detector",
+  "description": "Fine-tuned DINOv2-large for roadwork detection",
+  "version": "1.0.0",
+  "checkpoint": "checkpoint-1035",
+  "submitted_by": "5HT5XkhAg9jTzJLedi16L7uXnMTAy5CCvJd1YjaUy39gsETG",
+  "submission_time": 1770596758,
+  "metrics": {
+    "accuracy": "0.9414",
+    "precision": "0.9654",
+    "recall": "0.9616",
+    "f1": "0.9635",
+    "epoch": 15.0
+  }
+}

preprocessor_config.json ADDED Viewed

	@@ -0,0 +1,27 @@

+{
+  "crop_size": {
+    "height": 224,
+    "width": 224
+  },
+  "do_center_crop": true,
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.485,
+    0.456,
+    0.406
+  ],
+  "image_processor_type": "BitImageProcessor",
+  "image_std": [
+    0.229,
+    0.224,
+    0.225
+  ],
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "shortest_edge": 256
+  }
+}

trainer_state.json ADDED Viewed

	@@ -0,0 +1,354 @@

+{
+  "best_global_step": 1035,
+  "best_metric": 0.9635036496350365,
+  "best_model_checkpoint": "/home/user/natix-mining-project/models/dinov2_roadwork/checkpoint-1035",
+  "epoch": 15.0,
+  "eval_steps": 500,
+  "global_step": 1035,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.7299270072992701,
+      "grad_norm": 16.808786392211914,
+      "learning_rate": 3.3333333333333335e-07,
+      "loss": 0.5872037124633789,
+      "step": 50
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8219616204690832,
+      "eval_f1": 0.8982327848872639,
+      "eval_loss": 0.4097755253314972,
+      "eval_precision": 0.8318284424379232,
+      "eval_recall": 0.976158940397351,
+      "eval_runtime": 87.0676,
+      "eval_samples_per_second": 10.773,
+      "eval_steps_per_second": 0.345,
+      "step": 69
+    },
+    {
+      "epoch": 1.4525547445255476,
+      "grad_norm": 26.81678581237793,
+      "learning_rate": 6.884057971014494e-07,
+      "loss": 0.420146598815918,
+      "step": 100
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.908315565031983,
+      "eval_f1": 0.944372574385511,
+      "eval_loss": 0.23657475411891937,
+      "eval_precision": 0.922882427307206,
+      "eval_recall": 0.9668874172185431,
+      "eval_runtime": 89.7516,
+      "eval_samples_per_second": 10.451,
+      "eval_steps_per_second": 0.334,
+      "step": 138
+    },
+    {
+      "epoch": 2.1751824817518246,
+      "grad_norm": 17.34227180480957,
+      "learning_rate": 1.0434782608695653e-06,
+      "loss": 0.27175182342529297,
+      "step": 150
+    },
+    {
+      "epoch": 2.905109489051095,
+      "grad_norm": 19.6271915435791,
+      "learning_rate": 1.4057971014492755e-06,
+      "loss": 0.23565080642700195,
+      "step": 200
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9168443496801706,
+      "eval_f1": 0.9483443708609272,
+      "eval_loss": 0.20701788365840912,
+      "eval_precision": 0.9483443708609272,
+      "eval_recall": 0.9483443708609272,
+      "eval_runtime": 83.376,
+      "eval_samples_per_second": 11.25,
+      "eval_steps_per_second": 0.36,
+      "step": 207
+    },
+    {
+      "epoch": 3.627737226277372,
+      "grad_norm": 19.950403213500977,
+      "learning_rate": 1.7681159420289855e-06,
+      "loss": 0.2077177619934082,
+      "step": 250
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.9275053304904051,
+      "eval_f1": 0.9554390563564875,
+      "eval_loss": 0.1865338534116745,
+      "eval_precision": 0.9455252918287937,
+      "eval_recall": 0.9655629139072848,
+      "eval_runtime": 95.1442,
+      "eval_samples_per_second": 9.859,
+      "eval_steps_per_second": 0.315,
+      "step": 276
+    },
+    {
+      "epoch": 4.350364963503649,
+      "grad_norm": 14.074359893798828,
+      "learning_rate": 2.130434782608696e-06,
+      "loss": 0.1941886329650879,
+      "step": 300
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.9392324093816631,
+      "eval_f1": 0.9623762376237623,
+      "eval_loss": 0.1691393107175827,
+      "eval_precision": 0.9592105263157895,
+      "eval_recall": 0.9655629139072848,
+      "eval_runtime": 94.5554,
+      "eval_samples_per_second": 9.92,
+      "eval_steps_per_second": 0.317,
+      "step": 345
+    },
+    {
+      "epoch": 5.072992700729927,
+      "grad_norm": 15.120301246643066,
+      "learning_rate": 2.492753623188406e-06,
+      "loss": 0.18418676376342774,
+      "step": 350
+    },
+    {
+      "epoch": 5.802919708029197,
+      "grad_norm": 20.06453514099121,
+      "learning_rate": 2.847826086956522e-06,
+      "loss": 0.16181354522705077,
+      "step": 400
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.9317697228144989,
+      "eval_f1": 0.9571619812583668,
+      "eval_loss": 0.17490684986114502,
+      "eval_precision": 0.9675236806495264,
+      "eval_recall": 0.9470198675496688,
+      "eval_runtime": 78.099,
+      "eval_samples_per_second": 12.01,
+      "eval_steps_per_second": 0.384,
+      "step": 414
+    },
+    {
+      "epoch": 6.525547445255475,
+      "grad_norm": 16.788759231567383,
+      "learning_rate": 3.2101449275362324e-06,
+      "loss": 0.15671510696411134,
+      "step": 450
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.9157782515991472,
+      "eval_f1": 0.9460013670539986,
+      "eval_loss": 0.22261112928390503,
+      "eval_precision": 0.9774011299435028,
+      "eval_recall": 0.9165562913907285,
+      "eval_runtime": 92.5911,
+      "eval_samples_per_second": 10.131,
+      "eval_steps_per_second": 0.324,
+      "step": 483
+    },
+    {
+      "epoch": 7.248175182481752,
+      "grad_norm": 36.52346420288086,
+      "learning_rate": 3.5724637681159423e-06,
+      "loss": 0.16417793273925782,
+      "step": 500
+    },
+    {
+      "epoch": 7.978102189781022,
+      "grad_norm": 146.48744201660156,
+      "learning_rate": 3.9347826086956525e-06,
+      "loss": 0.13912958145141602,
+      "step": 550
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.9381663113006397,
+      "eval_f1": 0.9617414248021108,
+      "eval_loss": 0.15613189339637756,
+      "eval_precision": 0.9579500657030223,
+      "eval_recall": 0.9655629139072848,
+      "eval_runtime": 125.8153,
+      "eval_samples_per_second": 7.455,
+      "eval_steps_per_second": 0.238,
+      "step": 552
+    },
+    {
+      "epoch": 8.700729927007298,
+      "grad_norm": 37.7802734375,
+      "learning_rate": 4.297101449275362e-06,
+      "loss": 0.12975292205810546,
+      "step": 600
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.9402985074626866,
+      "eval_f1": 0.9627659574468085,
+      "eval_loss": 0.1664065420627594,
+      "eval_precision": 0.9666221628838452,
+      "eval_recall": 0.9589403973509933,
+      "eval_runtime": 93.79,
+      "eval_samples_per_second": 10.001,
+      "eval_steps_per_second": 0.32,
+      "step": 621
+    },
+    {
+      "epoch": 9.423357664233576,
+      "grad_norm": 39.47712707519531,
+      "learning_rate": 4.659420289855073e-06,
+      "loss": 0.12847793579101563,
+      "step": 650
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.9402985074626866,
+      "eval_f1": 0.963302752293578,
+      "eval_loss": 0.16533519327640533,
+      "eval_precision": 0.953307392996109,
+      "eval_recall": 0.9735099337748344,
+      "eval_runtime": 92.125,
+      "eval_samples_per_second": 10.182,
+      "eval_steps_per_second": 0.326,
+      "step": 690
+    },
+    {
+      "epoch": 10.145985401459853,
+      "grad_norm": 13.382551193237305,
+      "learning_rate": 4.999997120818891e-06,
+      "loss": 0.12533601760864257,
+      "step": 700
+    },
+    {
+      "epoch": 10.875912408759124,
+      "grad_norm": 152.12037658691406,
+      "learning_rate": 4.999101429246326e-06,
+      "loss": 0.10754921913146973,
+      "step": 750
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.9402985074626866,
+      "eval_f1": 0.9628154050464808,
+      "eval_loss": 0.15896020829677582,
+      "eval_precision": 0.9653794940079894,
+      "eval_recall": 0.9602649006622517,
+      "eval_runtime": 92.6027,
+      "eval_samples_per_second": 10.129,
+      "eval_steps_per_second": 0.324,
+      "step": 759
+    },
+    {
+      "epoch": 11.598540145985401,
+      "grad_norm": 7.632987022399902,
+      "learning_rate": 4.996606852480508e-06,
+      "loss": 0.08507664680480957,
+      "step": 800
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.9275053304904051,
+      "eval_f1": 0.9562419562419563,
+      "eval_loss": 0.19671548902988434,
+      "eval_precision": 0.9299123904881101,
+      "eval_recall": 0.9841059602649007,
+      "eval_runtime": 79.1855,
+      "eval_samples_per_second": 11.846,
+      "eval_steps_per_second": 0.379,
+      "step": 828
+    },
+    {
+      "epoch": 12.321167883211679,
+      "grad_norm": 451.5773010253906,
+      "learning_rate": 4.992514986511825e-06,
+      "loss": 0.11187627792358398,
+      "step": 850
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.9339019189765458,
+      "eval_f1": 0.9602053915275995,
+      "eval_loss": 0.20483285188674927,
+      "eval_precision": 0.9315068493150684,
+      "eval_recall": 0.990728476821192,
+      "eval_runtime": 84.5285,
+      "eval_samples_per_second": 11.097,
+      "eval_steps_per_second": 0.355,
+      "step": 897
+    },
+    {
+      "epoch": 13.043795620437956,
+      "grad_norm": 27.037843704223633,
+      "learning_rate": 4.986828449250792e-06,
+      "loss": 0.08265207290649414,
+      "step": 900
+    },
+    {
+      "epoch": 13.773722627737227,
+      "grad_norm": 58.419273376464844,
+      "learning_rate": 4.979550878853154e-06,
+      "loss": 0.08042881965637207,
+      "step": 950
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.9349680170575693,
+      "eval_f1": 0.9605688429217841,
+      "eval_loss": 0.21270374953746796,
+      "eval_precision": 0.9381313131313131,
+      "eval_recall": 0.9841059602649007,
+      "eval_runtime": 92.7757,
+      "eval_samples_per_second": 10.11,
+      "eval_steps_per_second": 0.323,
+      "step": 966
+    },
+    {
+      "epoch": 14.496350364963504,
+      "grad_norm": 60.74350357055664,
+      "learning_rate": 4.97068693139226e-06,
+      "loss": 0.08053727149963379,
+      "step": 1000
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.94136460554371,
+      "eval_f1": 0.9635036496350365,
+      "eval_loss": 0.18071264028549194,
+      "eval_precision": 0.9654255319148937,
+      "eval_recall": 0.9615894039735099,
+      "eval_runtime": 87.3638,
+      "eval_samples_per_second": 10.737,
+      "eval_steps_per_second": 0.343,
+      "step": 1035
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 6900,
+  "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": 1.80402939303552e+19,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}