Training complete: ConvNeXtV2 Tiny with 2.5x weighted recall bias

README.md

@@ -0,0 +1,81 @@
+---
+library_name: transformers
+license: apache-2.0
+base_model: facebook/convnextv2-tiny-1k-224
+tags:
+- generated_from_trainer
+metrics:
+- accuracy
+- precision
+- recall
+- f1
+model-index:
+- name: ConvNeXtV2_Tiny_v5
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# ConvNeXtV2_Tiny_v5
+
+This model is a fine-tuned version of [facebook/convnextv2-tiny-1k-224](https://huggingface.co/facebook/convnextv2-tiny-1k-224) on an unknown dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.0440
+- Accuracy: 0.9915
+- Precision: 0.9957
+- Recall: 0.9860
+- F1: 0.9908
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0001
+- train_batch_size: 128
+- eval_batch_size: 128
+- seed: 42
+- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 143
+- num_epochs: 13
+- mixed_precision_training: Native AMP
+
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1     |
+|:-------------:|:-----:|:----:|:---------------:|:--------:|:---------:|:------:|:------:|
+| 0.1871        | 1.0   | 111  | 0.2640          | 0.7658   | 0.6656    | 0.9902 | 0.7961 |
+| 0.1994        | 2.0   | 222  | 0.0950          | 0.9729   | 0.9667    | 0.9750 | 0.9708 |
+| 0.2047        | 3.0   | 333  | 0.0681          | 0.9862   | 0.9956    | 0.9744 | 0.9849 |
+| 0.1691        | 4.0   | 444  | 0.0697          | 0.9848   | 0.9901    | 0.9768 | 0.9834 |
+| 0.1664        | 5.0   | 555  | 0.0673          | 0.9887   | 0.9950    | 0.9805 | 0.9877 |
+| 0.1318        | 6.0   | 666  | 0.0747          | 0.9772   | 0.9642    | 0.9872 | 0.9756 |
+| 0.1367        | 7.0   | 777  | 0.0461          | 0.9896   | 0.9938    | 0.9835 | 0.9886 |
+| 0.1134        | 8.0   | 888  | 0.0431          | 0.9910   | 0.9951    | 0.9853 | 0.9902 |
+| 0.1782        | 9.0   | 999  | 0.0526          | 0.9896   | 0.9951    | 0.9823 | 0.9886 |
+| 0.1513        | 10.0  | 1110 | 0.0520          | 0.9901   | 0.9957    | 0.9829 | 0.9892 |
+| 0.1352        | 11.0  | 1221 | 0.0469          | 0.9901   | 0.9938    | 0.9847 | 0.9893 |
+| 0.1377        | 12.0  | 1332 | 0.0448          | 0.9907   | 0.9957    | 0.9841 | 0.9899 |
+| 0.1107        | 13.0  | 1443 | 0.0440          | 0.9915   | 0.9957    | 0.9860 | 0.9908 |
+
+
+### Framework versions
+
+- Transformers 5.2.0
+- Pytorch 2.9.0+cu126
+- Datasets 4.0.0
+- Tokenizers 0.22.2

config.json

@@ -0,0 +1,51 @@
+{
+  "architectures": [
+    "ConvNextV2ForImageClassification"
+  ],
+  "depths": [
+    3,
+    3,
+    9,
+    3
+  ],
+  "drop_path_rate": 0.0,
+  "dtype": "float32",
+  "hidden_act": "gelu",
+  "hidden_sizes": [
+    96,
+    192,
+    384,
+    768
+  ],
+  "id2label": {
+    "0": "0",
+    "1": "1"
+  },
+  "image_size": 224,
+  "initializer_range": 0.02,
+  "label2id": {
+    "0": 0,
+    "1": 1
+  },
+  "layer_norm_eps": 1e-12,
+  "model_type": "convnextv2",
+  "num_channels": 3,
+  "num_stages": 4,
+  "out_features": [
+    "stage4"
+  ],
+  "out_indices": [
+    4
+  ],
+  "patch_size": 4,
+  "problem_type": "single_label_classification",
+  "stage_names": [
+    "stem",
+    "stage1",
+    "stage2",
+    "stage3",
+    "stage4"
+  ],
+  "transformers_version": "5.2.0",
+  "use_cache": false
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b6e8b82762663acd7b9528dd911408c97c06dd1634f92c0b060ca1f3e1e3ed1
+size 111495808

preprocessor_config.json

@@ -0,0 +1,23 @@
+{
+  "crop_pct": 0.875,
+  "data_format": "channels_first",
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.485,
+    0.456,
+    0.406
+  ],
+  "image_processor_type": "ConvNextImageProcessorFast",
+  "image_std": [
+    0.229,
+    0.224,
+    0.225
+  ],
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "shortest_edge": 224
+  }
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ebb41ece1275fce017f7d955220837691eeaa2dbd315c4b0deee4600e83f357
+size 5201