Upload folder using huggingface_hub
Browse files- checkpoint-3800/config.json +46 -0
- checkpoint-3800/model.safetensors +3 -0
- checkpoint-3800/optimizer.pt +3 -0
- checkpoint-3800/preprocessor_config.json +22 -0
- checkpoint-3800/rng_state.pth +3 -0
- checkpoint-3800/scheduler.pt +3 -0
- checkpoint-3800/trainer_state.json +961 -0
- checkpoint-3800/training_args.bin +3 -0
- config.json +46 -0
- model.safetensors +3 -0
- preprocessor_config.json +22 -0
- training_args.bin +3 -0
checkpoint-3800/config.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "google/vit-base-patch16-224-in21k",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"ViTForImageClassification"
|
| 5 |
+
],
|
| 6 |
+
"attention_probs_dropout_prob": 0.0,
|
| 7 |
+
"encoder_stride": 16,
|
| 8 |
+
"hidden_act": "gelu",
|
| 9 |
+
"hidden_dropout_prob": 0.0,
|
| 10 |
+
"hidden_size": 768,
|
| 11 |
+
"id2label": {
|
| 12 |
+
"0": "Bridge",
|
| 13 |
+
"1": "Child",
|
| 14 |
+
"2": "Cobra",
|
| 15 |
+
"3": "Downward-Dog",
|
| 16 |
+
"4": "Pigeon",
|
| 17 |
+
"5": "Standing-Mountain",
|
| 18 |
+
"6": "Tree",
|
| 19 |
+
"7": "Triangle",
|
| 20 |
+
"8": "Warrior"
|
| 21 |
+
},
|
| 22 |
+
"image_size": 224,
|
| 23 |
+
"initializer_range": 0.02,
|
| 24 |
+
"intermediate_size": 3072,
|
| 25 |
+
"label2id": {
|
| 26 |
+
"Bridge": 0,
|
| 27 |
+
"Child": 1,
|
| 28 |
+
"Cobra": 2,
|
| 29 |
+
"Downward-Dog": 3,
|
| 30 |
+
"Pigeon": 4,
|
| 31 |
+
"Standing-Mountain": 5,
|
| 32 |
+
"Tree": 6,
|
| 33 |
+
"Triangle": 7,
|
| 34 |
+
"Warrior": 8
|
| 35 |
+
},
|
| 36 |
+
"layer_norm_eps": 1e-12,
|
| 37 |
+
"model_type": "vit",
|
| 38 |
+
"num_attention_heads": 12,
|
| 39 |
+
"num_channels": 3,
|
| 40 |
+
"num_hidden_layers": 12,
|
| 41 |
+
"patch_size": 16,
|
| 42 |
+
"problem_type": "single_label_classification",
|
| 43 |
+
"qkv_bias": true,
|
| 44 |
+
"torch_dtype": "float32",
|
| 45 |
+
"transformers_version": "4.35.2"
|
| 46 |
+
}
|
checkpoint-3800/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:929d3e1f0a2285aa2165bbd131c14e00d8743cb7e422831240ac999593c1c4b1
|
| 3 |
+
size 343245508
|
checkpoint-3800/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68ea7e481c5de788b9e7ba32eb4be8bd18b63935b1b4b8b04fec4fec22c241e6
|
| 3 |
+
size 686611461
|
checkpoint-3800/preprocessor_config.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_normalize": true,
|
| 3 |
+
"do_rescale": true,
|
| 4 |
+
"do_resize": true,
|
| 5 |
+
"image_mean": [
|
| 6 |
+
0.5,
|
| 7 |
+
0.5,
|
| 8 |
+
0.5
|
| 9 |
+
],
|
| 10 |
+
"image_processor_type": "ViTImageProcessor",
|
| 11 |
+
"image_std": [
|
| 12 |
+
0.5,
|
| 13 |
+
0.5,
|
| 14 |
+
0.5
|
| 15 |
+
],
|
| 16 |
+
"resample": 2,
|
| 17 |
+
"rescale_factor": 0.00392156862745098,
|
| 18 |
+
"size": {
|
| 19 |
+
"height": 224,
|
| 20 |
+
"width": 224
|
| 21 |
+
}
|
| 22 |
+
}
|
checkpoint-3800/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:21cfc308185c2aa1404ebf27c700939afcb55a1d5c912dae633432f4887bfcad
|
| 3 |
+
size 14575
|
checkpoint-3800/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:359355029ea11ca364fc25e2c18435e5f944992678b2a3b4158d44e2f52212c8
|
| 3 |
+
size 627
|
checkpoint-3800/trainer_state.json
ADDED
|
@@ -0,0 +1,961 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.3843018710613251,
|
| 3 |
+
"best_model_checkpoint": "yoga_pose_image_classification/checkpoint-3800",
|
| 4 |
+
"epoch": 100.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 3800,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 1.0,
|
| 13 |
+
"eval_accuracy": 0.1212624584717608,
|
| 14 |
+
"eval_loss": 2.18632435798645,
|
| 15 |
+
"eval_runtime": 9.4012,
|
| 16 |
+
"eval_samples_per_second": 64.034,
|
| 17 |
+
"eval_steps_per_second": 2.021,
|
| 18 |
+
"step": 38
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"epoch": 2.0,
|
| 22 |
+
"eval_accuracy": 0.2009966777408638,
|
| 23 |
+
"eval_loss": 2.154275417327881,
|
| 24 |
+
"eval_runtime": 9.4031,
|
| 25 |
+
"eval_samples_per_second": 64.021,
|
| 26 |
+
"eval_steps_per_second": 2.021,
|
| 27 |
+
"step": 76
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 3.0,
|
| 31 |
+
"eval_accuracy": 0.3239202657807309,
|
| 32 |
+
"eval_loss": 2.1111578941345215,
|
| 33 |
+
"eval_runtime": 9.3981,
|
| 34 |
+
"eval_samples_per_second": 64.056,
|
| 35 |
+
"eval_steps_per_second": 2.022,
|
| 36 |
+
"step": 114
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"epoch": 4.0,
|
| 40 |
+
"eval_accuracy": 0.4833887043189369,
|
| 41 |
+
"eval_loss": 2.052860736846924,
|
| 42 |
+
"eval_runtime": 9.3515,
|
| 43 |
+
"eval_samples_per_second": 64.374,
|
| 44 |
+
"eval_steps_per_second": 2.032,
|
| 45 |
+
"step": 152
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"epoch": 5.0,
|
| 49 |
+
"eval_accuracy": 0.5863787375415282,
|
| 50 |
+
"eval_loss": 1.9836093187332153,
|
| 51 |
+
"eval_runtime": 9.2802,
|
| 52 |
+
"eval_samples_per_second": 64.869,
|
| 53 |
+
"eval_steps_per_second": 2.047,
|
| 54 |
+
"step": 190
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 6.0,
|
| 58 |
+
"eval_accuracy": 0.6677740863787376,
|
| 59 |
+
"eval_loss": 1.9060654640197754,
|
| 60 |
+
"eval_runtime": 9.3038,
|
| 61 |
+
"eval_samples_per_second": 64.704,
|
| 62 |
+
"eval_steps_per_second": 2.042,
|
| 63 |
+
"step": 228
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"epoch": 7.0,
|
| 67 |
+
"eval_accuracy": 0.7126245847176079,
|
| 68 |
+
"eval_loss": 1.8243703842163086,
|
| 69 |
+
"eval_runtime": 9.2471,
|
| 70 |
+
"eval_samples_per_second": 65.102,
|
| 71 |
+
"eval_steps_per_second": 2.055,
|
| 72 |
+
"step": 266
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 8.0,
|
| 76 |
+
"eval_accuracy": 0.7691029900332226,
|
| 77 |
+
"eval_loss": 1.738010287284851,
|
| 78 |
+
"eval_runtime": 9.2752,
|
| 79 |
+
"eval_samples_per_second": 64.905,
|
| 80 |
+
"eval_steps_per_second": 2.048,
|
| 81 |
+
"step": 304
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 9.0,
|
| 85 |
+
"eval_accuracy": 0.7956810631229236,
|
| 86 |
+
"eval_loss": 1.6476677656173706,
|
| 87 |
+
"eval_runtime": 9.1847,
|
| 88 |
+
"eval_samples_per_second": 65.544,
|
| 89 |
+
"eval_steps_per_second": 2.069,
|
| 90 |
+
"step": 342
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"epoch": 10.0,
|
| 94 |
+
"eval_accuracy": 0.8073089700996677,
|
| 95 |
+
"eval_loss": 1.5577423572540283,
|
| 96 |
+
"eval_runtime": 9.1847,
|
| 97 |
+
"eval_samples_per_second": 65.544,
|
| 98 |
+
"eval_steps_per_second": 2.069,
|
| 99 |
+
"step": 380
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 11.0,
|
| 103 |
+
"eval_accuracy": 0.8222591362126246,
|
| 104 |
+
"eval_loss": 1.4709256887435913,
|
| 105 |
+
"eval_runtime": 8.9043,
|
| 106 |
+
"eval_samples_per_second": 67.608,
|
| 107 |
+
"eval_steps_per_second": 2.134,
|
| 108 |
+
"step": 418
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"epoch": 12.0,
|
| 112 |
+
"eval_accuracy": 0.8338870431893688,
|
| 113 |
+
"eval_loss": 1.3882230520248413,
|
| 114 |
+
"eval_runtime": 9.1638,
|
| 115 |
+
"eval_samples_per_second": 65.693,
|
| 116 |
+
"eval_steps_per_second": 2.073,
|
| 117 |
+
"step": 456
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"epoch": 13.0,
|
| 121 |
+
"eval_accuracy": 0.845514950166113,
|
| 122 |
+
"eval_loss": 1.314037799835205,
|
| 123 |
+
"eval_runtime": 9.4856,
|
| 124 |
+
"eval_samples_per_second": 63.465,
|
| 125 |
+
"eval_steps_per_second": 2.003,
|
| 126 |
+
"step": 494
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"epoch": 13.16,
|
| 130 |
+
"learning_rate": 3.5199999999999998e-06,
|
| 131 |
+
"loss": 1.7637,
|
| 132 |
+
"step": 500
|
| 133 |
+
},
|
| 134 |
+
{
|
| 135 |
+
"epoch": 14.0,
|
| 136 |
+
"eval_accuracy": 0.8554817275747508,
|
| 137 |
+
"eval_loss": 1.2419167757034302,
|
| 138 |
+
"eval_runtime": 9.2365,
|
| 139 |
+
"eval_samples_per_second": 65.176,
|
| 140 |
+
"eval_steps_per_second": 2.057,
|
| 141 |
+
"step": 532
|
| 142 |
+
},
|
| 143 |
+
{
|
| 144 |
+
"epoch": 15.0,
|
| 145 |
+
"eval_accuracy": 0.8621262458471761,
|
| 146 |
+
"eval_loss": 1.175502896308899,
|
| 147 |
+
"eval_runtime": 9.1995,
|
| 148 |
+
"eval_samples_per_second": 65.439,
|
| 149 |
+
"eval_steps_per_second": 2.065,
|
| 150 |
+
"step": 570
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"epoch": 16.0,
|
| 154 |
+
"eval_accuracy": 0.867109634551495,
|
| 155 |
+
"eval_loss": 1.1211938858032227,
|
| 156 |
+
"eval_runtime": 9.2041,
|
| 157 |
+
"eval_samples_per_second": 65.406,
|
| 158 |
+
"eval_steps_per_second": 2.064,
|
| 159 |
+
"step": 608
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 17.0,
|
| 163 |
+
"eval_accuracy": 0.8704318936877077,
|
| 164 |
+
"eval_loss": 1.0678402185440063,
|
| 165 |
+
"eval_runtime": 9.1819,
|
| 166 |
+
"eval_samples_per_second": 65.564,
|
| 167 |
+
"eval_steps_per_second": 2.069,
|
| 168 |
+
"step": 646
|
| 169 |
+
},
|
| 170 |
+
{
|
| 171 |
+
"epoch": 18.0,
|
| 172 |
+
"eval_accuracy": 0.8770764119601329,
|
| 173 |
+
"eval_loss": 1.0240187644958496,
|
| 174 |
+
"eval_runtime": 8.9907,
|
| 175 |
+
"eval_samples_per_second": 66.958,
|
| 176 |
+
"eval_steps_per_second": 2.113,
|
| 177 |
+
"step": 684
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 19.0,
|
| 181 |
+
"eval_accuracy": 0.8803986710963455,
|
| 182 |
+
"eval_loss": 0.9842038750648499,
|
| 183 |
+
"eval_runtime": 9.2217,
|
| 184 |
+
"eval_samples_per_second": 65.281,
|
| 185 |
+
"eval_steps_per_second": 2.06,
|
| 186 |
+
"step": 722
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 20.0,
|
| 190 |
+
"eval_accuracy": 0.8870431893687708,
|
| 191 |
+
"eval_loss": 0.9439226984977722,
|
| 192 |
+
"eval_runtime": 9.1617,
|
| 193 |
+
"eval_samples_per_second": 65.708,
|
| 194 |
+
"eval_steps_per_second": 2.074,
|
| 195 |
+
"step": 760
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 21.0,
|
| 199 |
+
"eval_accuracy": 0.8920265780730897,
|
| 200 |
+
"eval_loss": 0.9107761383056641,
|
| 201 |
+
"eval_runtime": 9.3007,
|
| 202 |
+
"eval_samples_per_second": 64.726,
|
| 203 |
+
"eval_steps_per_second": 2.043,
|
| 204 |
+
"step": 798
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"epoch": 22.0,
|
| 208 |
+
"eval_accuracy": 0.893687707641196,
|
| 209 |
+
"eval_loss": 0.8776938319206238,
|
| 210 |
+
"eval_runtime": 8.8098,
|
| 211 |
+
"eval_samples_per_second": 68.333,
|
| 212 |
+
"eval_steps_per_second": 2.157,
|
| 213 |
+
"step": 836
|
| 214 |
+
},
|
| 215 |
+
{
|
| 216 |
+
"epoch": 23.0,
|
| 217 |
+
"eval_accuracy": 0.893687707641196,
|
| 218 |
+
"eval_loss": 0.8496754765510559,
|
| 219 |
+
"eval_runtime": 9.052,
|
| 220 |
+
"eval_samples_per_second": 66.504,
|
| 221 |
+
"eval_steps_per_second": 2.099,
|
| 222 |
+
"step": 874
|
| 223 |
+
},
|
| 224 |
+
{
|
| 225 |
+
"epoch": 24.0,
|
| 226 |
+
"eval_accuracy": 0.8986710963455149,
|
| 227 |
+
"eval_loss": 0.8216127753257751,
|
| 228 |
+
"eval_runtime": 9.1831,
|
| 229 |
+
"eval_samples_per_second": 65.555,
|
| 230 |
+
"eval_steps_per_second": 2.069,
|
| 231 |
+
"step": 912
|
| 232 |
+
},
|
| 233 |
+
{
|
| 234 |
+
"epoch": 25.0,
|
| 235 |
+
"eval_accuracy": 0.9053156146179402,
|
| 236 |
+
"eval_loss": 0.7941116094589233,
|
| 237 |
+
"eval_runtime": 8.8187,
|
| 238 |
+
"eval_samples_per_second": 68.264,
|
| 239 |
+
"eval_steps_per_second": 2.155,
|
| 240 |
+
"step": 950
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"epoch": 26.0,
|
| 244 |
+
"eval_accuracy": 0.9086378737541528,
|
| 245 |
+
"eval_loss": 0.7701788544654846,
|
| 246 |
+
"eval_runtime": 9.255,
|
| 247 |
+
"eval_samples_per_second": 65.046,
|
| 248 |
+
"eval_steps_per_second": 2.053,
|
| 249 |
+
"step": 988
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 26.32,
|
| 253 |
+
"learning_rate": 2.9866666666666667e-06,
|
| 254 |
+
"loss": 0.8811,
|
| 255 |
+
"step": 1000
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"epoch": 27.0,
|
| 259 |
+
"eval_accuracy": 0.9069767441860465,
|
| 260 |
+
"eval_loss": 0.7498743534088135,
|
| 261 |
+
"eval_runtime": 8.9384,
|
| 262 |
+
"eval_samples_per_second": 67.35,
|
| 263 |
+
"eval_steps_per_second": 2.126,
|
| 264 |
+
"step": 1026
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"epoch": 28.0,
|
| 268 |
+
"eval_accuracy": 0.9102990033222591,
|
| 269 |
+
"eval_loss": 0.7278289794921875,
|
| 270 |
+
"eval_runtime": 9.2157,
|
| 271 |
+
"eval_samples_per_second": 65.324,
|
| 272 |
+
"eval_steps_per_second": 2.062,
|
| 273 |
+
"step": 1064
|
| 274 |
+
},
|
| 275 |
+
{
|
| 276 |
+
"epoch": 29.0,
|
| 277 |
+
"eval_accuracy": 0.9119601328903655,
|
| 278 |
+
"eval_loss": 0.7089008092880249,
|
| 279 |
+
"eval_runtime": 9.5658,
|
| 280 |
+
"eval_samples_per_second": 62.933,
|
| 281 |
+
"eval_steps_per_second": 1.986,
|
| 282 |
+
"step": 1102
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 30.0,
|
| 286 |
+
"eval_accuracy": 0.9186046511627907,
|
| 287 |
+
"eval_loss": 0.6902615427970886,
|
| 288 |
+
"eval_runtime": 9.276,
|
| 289 |
+
"eval_samples_per_second": 64.899,
|
| 290 |
+
"eval_steps_per_second": 2.048,
|
| 291 |
+
"step": 1140
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 31.0,
|
| 295 |
+
"eval_accuracy": 0.920265780730897,
|
| 296 |
+
"eval_loss": 0.6722272038459778,
|
| 297 |
+
"eval_runtime": 9.2907,
|
| 298 |
+
"eval_samples_per_second": 64.796,
|
| 299 |
+
"eval_steps_per_second": 2.045,
|
| 300 |
+
"step": 1178
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 32.0,
|
| 304 |
+
"eval_accuracy": 0.9169435215946844,
|
| 305 |
+
"eval_loss": 0.6599347591400146,
|
| 306 |
+
"eval_runtime": 9.146,
|
| 307 |
+
"eval_samples_per_second": 65.821,
|
| 308 |
+
"eval_steps_per_second": 2.077,
|
| 309 |
+
"step": 1216
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"epoch": 33.0,
|
| 313 |
+
"eval_accuracy": 0.9169435215946844,
|
| 314 |
+
"eval_loss": 0.6435149312019348,
|
| 315 |
+
"eval_runtime": 9.2827,
|
| 316 |
+
"eval_samples_per_second": 64.852,
|
| 317 |
+
"eval_steps_per_second": 2.047,
|
| 318 |
+
"step": 1254
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"epoch": 34.0,
|
| 322 |
+
"eval_accuracy": 0.9186046511627907,
|
| 323 |
+
"eval_loss": 0.6296250820159912,
|
| 324 |
+
"eval_runtime": 8.9518,
|
| 325 |
+
"eval_samples_per_second": 67.249,
|
| 326 |
+
"eval_steps_per_second": 2.122,
|
| 327 |
+
"step": 1292
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 35.0,
|
| 331 |
+
"eval_accuracy": 0.920265780730897,
|
| 332 |
+
"eval_loss": 0.6152604222297668,
|
| 333 |
+
"eval_runtime": 9.2036,
|
| 334 |
+
"eval_samples_per_second": 65.409,
|
| 335 |
+
"eval_steps_per_second": 2.064,
|
| 336 |
+
"step": 1330
|
| 337 |
+
},
|
| 338 |
+
{
|
| 339 |
+
"epoch": 36.0,
|
| 340 |
+
"eval_accuracy": 0.9219269102990033,
|
| 341 |
+
"eval_loss": 0.6024987101554871,
|
| 342 |
+
"eval_runtime": 9.1807,
|
| 343 |
+
"eval_samples_per_second": 65.573,
|
| 344 |
+
"eval_steps_per_second": 2.07,
|
| 345 |
+
"step": 1368
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 37.0,
|
| 349 |
+
"eval_accuracy": 0.9235880398671097,
|
| 350 |
+
"eval_loss": 0.5926721692085266,
|
| 351 |
+
"eval_runtime": 9.3058,
|
| 352 |
+
"eval_samples_per_second": 64.691,
|
| 353 |
+
"eval_steps_per_second": 2.042,
|
| 354 |
+
"step": 1406
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"epoch": 38.0,
|
| 358 |
+
"eval_accuracy": 0.9302325581395349,
|
| 359 |
+
"eval_loss": 0.5806441903114319,
|
| 360 |
+
"eval_runtime": 9.2355,
|
| 361 |
+
"eval_samples_per_second": 65.183,
|
| 362 |
+
"eval_steps_per_second": 2.057,
|
| 363 |
+
"step": 1444
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 39.0,
|
| 367 |
+
"eval_accuracy": 0.925249169435216,
|
| 368 |
+
"eval_loss": 0.5694455504417419,
|
| 369 |
+
"eval_runtime": 9.1082,
|
| 370 |
+
"eval_samples_per_second": 66.094,
|
| 371 |
+
"eval_steps_per_second": 2.086,
|
| 372 |
+
"step": 1482
|
| 373 |
+
},
|
| 374 |
+
{
|
| 375 |
+
"epoch": 39.47,
|
| 376 |
+
"learning_rate": 2.453333333333333e-06,
|
| 377 |
+
"loss": 0.5341,
|
| 378 |
+
"step": 1500
|
| 379 |
+
},
|
| 380 |
+
{
|
| 381 |
+
"epoch": 40.0,
|
| 382 |
+
"eval_accuracy": 0.9235880398671097,
|
| 383 |
+
"eval_loss": 0.5654778480529785,
|
| 384 |
+
"eval_runtime": 9.1414,
|
| 385 |
+
"eval_samples_per_second": 65.854,
|
| 386 |
+
"eval_steps_per_second": 2.078,
|
| 387 |
+
"step": 1520
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 41.0,
|
| 391 |
+
"eval_accuracy": 0.9285714285714286,
|
| 392 |
+
"eval_loss": 0.5506969094276428,
|
| 393 |
+
"eval_runtime": 9.2079,
|
| 394 |
+
"eval_samples_per_second": 65.379,
|
| 395 |
+
"eval_steps_per_second": 2.063,
|
| 396 |
+
"step": 1558
|
| 397 |
+
},
|
| 398 |
+
{
|
| 399 |
+
"epoch": 42.0,
|
| 400 |
+
"eval_accuracy": 0.9302325581395349,
|
| 401 |
+
"eval_loss": 0.543114185333252,
|
| 402 |
+
"eval_runtime": 9.2502,
|
| 403 |
+
"eval_samples_per_second": 65.08,
|
| 404 |
+
"eval_steps_per_second": 2.054,
|
| 405 |
+
"step": 1596
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 43.0,
|
| 409 |
+
"eval_accuracy": 0.9318936877076412,
|
| 410 |
+
"eval_loss": 0.5349024534225464,
|
| 411 |
+
"eval_runtime": 9.0081,
|
| 412 |
+
"eval_samples_per_second": 66.829,
|
| 413 |
+
"eval_steps_per_second": 2.109,
|
| 414 |
+
"step": 1634
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 44.0,
|
| 418 |
+
"eval_accuracy": 0.9318936877076412,
|
| 419 |
+
"eval_loss": 0.5262035727500916,
|
| 420 |
+
"eval_runtime": 9.1978,
|
| 421 |
+
"eval_samples_per_second": 65.45,
|
| 422 |
+
"eval_steps_per_second": 2.066,
|
| 423 |
+
"step": 1672
|
| 424 |
+
},
|
| 425 |
+
{
|
| 426 |
+
"epoch": 45.0,
|
| 427 |
+
"eval_accuracy": 0.9269102990033222,
|
| 428 |
+
"eval_loss": 0.5220299363136292,
|
| 429 |
+
"eval_runtime": 9.148,
|
| 430 |
+
"eval_samples_per_second": 65.807,
|
| 431 |
+
"eval_steps_per_second": 2.077,
|
| 432 |
+
"step": 1710
|
| 433 |
+
},
|
| 434 |
+
{
|
| 435 |
+
"epoch": 46.0,
|
| 436 |
+
"eval_accuracy": 0.9269102990033222,
|
| 437 |
+
"eval_loss": 0.5145980715751648,
|
| 438 |
+
"eval_runtime": 9.2485,
|
| 439 |
+
"eval_samples_per_second": 65.092,
|
| 440 |
+
"eval_steps_per_second": 2.054,
|
| 441 |
+
"step": 1748
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"epoch": 47.0,
|
| 445 |
+
"eval_accuracy": 0.9302325581395349,
|
| 446 |
+
"eval_loss": 0.5084472894668579,
|
| 447 |
+
"eval_runtime": 8.9428,
|
| 448 |
+
"eval_samples_per_second": 67.317,
|
| 449 |
+
"eval_steps_per_second": 2.125,
|
| 450 |
+
"step": 1786
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"epoch": 48.0,
|
| 454 |
+
"eval_accuracy": 0.9302325581395349,
|
| 455 |
+
"eval_loss": 0.5007317066192627,
|
| 456 |
+
"eval_runtime": 9.1094,
|
| 457 |
+
"eval_samples_per_second": 66.086,
|
| 458 |
+
"eval_steps_per_second": 2.086,
|
| 459 |
+
"step": 1824
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 49.0,
|
| 463 |
+
"eval_accuracy": 0.9285714285714286,
|
| 464 |
+
"eval_loss": 0.4938865602016449,
|
| 465 |
+
"eval_runtime": 9.3331,
|
| 466 |
+
"eval_samples_per_second": 64.501,
|
| 467 |
+
"eval_steps_per_second": 2.036,
|
| 468 |
+
"step": 1862
|
| 469 |
+
},
|
| 470 |
+
{
|
| 471 |
+
"epoch": 50.0,
|
| 472 |
+
"eval_accuracy": 0.9302325581395349,
|
| 473 |
+
"eval_loss": 0.48970168828964233,
|
| 474 |
+
"eval_runtime": 9.2175,
|
| 475 |
+
"eval_samples_per_second": 65.31,
|
| 476 |
+
"eval_steps_per_second": 2.061,
|
| 477 |
+
"step": 1900
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"epoch": 51.0,
|
| 481 |
+
"eval_accuracy": 0.9302325581395349,
|
| 482 |
+
"eval_loss": 0.4844241738319397,
|
| 483 |
+
"eval_runtime": 9.3402,
|
| 484 |
+
"eval_samples_per_second": 64.453,
|
| 485 |
+
"eval_steps_per_second": 2.034,
|
| 486 |
+
"step": 1938
|
| 487 |
+
},
|
| 488 |
+
{
|
| 489 |
+
"epoch": 52.0,
|
| 490 |
+
"eval_accuracy": 0.9302325581395349,
|
| 491 |
+
"eval_loss": 0.4789960980415344,
|
| 492 |
+
"eval_runtime": 9.2399,
|
| 493 |
+
"eval_samples_per_second": 65.152,
|
| 494 |
+
"eval_steps_per_second": 2.056,
|
| 495 |
+
"step": 1976
|
| 496 |
+
},
|
| 497 |
+
{
|
| 498 |
+
"epoch": 52.63,
|
| 499 |
+
"learning_rate": 1.92e-06,
|
| 500 |
+
"loss": 0.3822,
|
| 501 |
+
"step": 2000
|
| 502 |
+
},
|
| 503 |
+
{
|
| 504 |
+
"epoch": 53.0,
|
| 505 |
+
"eval_accuracy": 0.9302325581395349,
|
| 506 |
+
"eval_loss": 0.4740654230117798,
|
| 507 |
+
"eval_runtime": 9.2394,
|
| 508 |
+
"eval_samples_per_second": 65.155,
|
| 509 |
+
"eval_steps_per_second": 2.056,
|
| 510 |
+
"step": 2014
|
| 511 |
+
},
|
| 512 |
+
{
|
| 513 |
+
"epoch": 54.0,
|
| 514 |
+
"eval_accuracy": 0.9318936877076412,
|
| 515 |
+
"eval_loss": 0.4689539968967438,
|
| 516 |
+
"eval_runtime": 9.192,
|
| 517 |
+
"eval_samples_per_second": 65.491,
|
| 518 |
+
"eval_steps_per_second": 2.067,
|
| 519 |
+
"step": 2052
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 55.0,
|
| 523 |
+
"eval_accuracy": 0.9302325581395349,
|
| 524 |
+
"eval_loss": 0.4651150107383728,
|
| 525 |
+
"eval_runtime": 9.0767,
|
| 526 |
+
"eval_samples_per_second": 66.324,
|
| 527 |
+
"eval_steps_per_second": 2.093,
|
| 528 |
+
"step": 2090
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"epoch": 56.0,
|
| 532 |
+
"eval_accuracy": 0.9318936877076412,
|
| 533 |
+
"eval_loss": 0.4605846703052521,
|
| 534 |
+
"eval_runtime": 9.2523,
|
| 535 |
+
"eval_samples_per_second": 65.065,
|
| 536 |
+
"eval_steps_per_second": 2.054,
|
| 537 |
+
"step": 2128
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 57.0,
|
| 541 |
+
"eval_accuracy": 0.9302325581395349,
|
| 542 |
+
"eval_loss": 0.4576462507247925,
|
| 543 |
+
"eval_runtime": 9.1536,
|
| 544 |
+
"eval_samples_per_second": 65.767,
|
| 545 |
+
"eval_steps_per_second": 2.076,
|
| 546 |
+
"step": 2166
|
| 547 |
+
},
|
| 548 |
+
{
|
| 549 |
+
"epoch": 58.0,
|
| 550 |
+
"eval_accuracy": 0.9318936877076412,
|
| 551 |
+
"eval_loss": 0.45208579301834106,
|
| 552 |
+
"eval_runtime": 9.1727,
|
| 553 |
+
"eval_samples_per_second": 65.63,
|
| 554 |
+
"eval_steps_per_second": 2.071,
|
| 555 |
+
"step": 2204
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 59.0,
|
| 559 |
+
"eval_accuracy": 0.9318936877076412,
|
| 560 |
+
"eval_loss": 0.4491443336009979,
|
| 561 |
+
"eval_runtime": 9.1974,
|
| 562 |
+
"eval_samples_per_second": 65.453,
|
| 563 |
+
"eval_steps_per_second": 2.066,
|
| 564 |
+
"step": 2242
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"epoch": 60.0,
|
| 568 |
+
"eval_accuracy": 0.9302325581395349,
|
| 569 |
+
"eval_loss": 0.44608035683631897,
|
| 570 |
+
"eval_runtime": 9.363,
|
| 571 |
+
"eval_samples_per_second": 64.295,
|
| 572 |
+
"eval_steps_per_second": 2.029,
|
| 573 |
+
"step": 2280
|
| 574 |
+
},
|
| 575 |
+
{
|
| 576 |
+
"epoch": 61.0,
|
| 577 |
+
"eval_accuracy": 0.9318936877076412,
|
| 578 |
+
"eval_loss": 0.4429391622543335,
|
| 579 |
+
"eval_runtime": 9.2675,
|
| 580 |
+
"eval_samples_per_second": 64.959,
|
| 581 |
+
"eval_steps_per_second": 2.05,
|
| 582 |
+
"step": 2318
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 62.0,
|
| 586 |
+
"eval_accuracy": 0.9318936877076412,
|
| 587 |
+
"eval_loss": 0.4398341476917267,
|
| 588 |
+
"eval_runtime": 9.2027,
|
| 589 |
+
"eval_samples_per_second": 65.416,
|
| 590 |
+
"eval_steps_per_second": 2.065,
|
| 591 |
+
"step": 2356
|
| 592 |
+
},
|
| 593 |
+
{
|
| 594 |
+
"epoch": 63.0,
|
| 595 |
+
"eval_accuracy": 0.9318936877076412,
|
| 596 |
+
"eval_loss": 0.43718549609184265,
|
| 597 |
+
"eval_runtime": 9.0863,
|
| 598 |
+
"eval_samples_per_second": 66.254,
|
| 599 |
+
"eval_steps_per_second": 2.091,
|
| 600 |
+
"step": 2394
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 64.0,
|
| 604 |
+
"eval_accuracy": 0.9302325581395349,
|
| 605 |
+
"eval_loss": 0.435174822807312,
|
| 606 |
+
"eval_runtime": 9.1649,
|
| 607 |
+
"eval_samples_per_second": 65.685,
|
| 608 |
+
"eval_steps_per_second": 2.073,
|
| 609 |
+
"step": 2432
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"epoch": 65.0,
|
| 613 |
+
"eval_accuracy": 0.9302325581395349,
|
| 614 |
+
"eval_loss": 0.4314589500427246,
|
| 615 |
+
"eval_runtime": 9.0429,
|
| 616 |
+
"eval_samples_per_second": 66.572,
|
| 617 |
+
"eval_steps_per_second": 2.101,
|
| 618 |
+
"step": 2470
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 65.79,
|
| 622 |
+
"learning_rate": 1.3866666666666666e-06,
|
| 623 |
+
"loss": 0.3089,
|
| 624 |
+
"step": 2500
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 66.0,
|
| 628 |
+
"eval_accuracy": 0.9302325581395349,
|
| 629 |
+
"eval_loss": 0.4279780983924866,
|
| 630 |
+
"eval_runtime": 8.8928,
|
| 631 |
+
"eval_samples_per_second": 67.695,
|
| 632 |
+
"eval_steps_per_second": 2.137,
|
| 633 |
+
"step": 2508
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 67.0,
|
| 637 |
+
"eval_accuracy": 0.9302325581395349,
|
| 638 |
+
"eval_loss": 0.4262917935848236,
|
| 639 |
+
"eval_runtime": 9.0012,
|
| 640 |
+
"eval_samples_per_second": 66.88,
|
| 641 |
+
"eval_steps_per_second": 2.111,
|
| 642 |
+
"step": 2546
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 68.0,
|
| 646 |
+
"eval_accuracy": 0.9302325581395349,
|
| 647 |
+
"eval_loss": 0.42236465215682983,
|
| 648 |
+
"eval_runtime": 8.9695,
|
| 649 |
+
"eval_samples_per_second": 67.116,
|
| 650 |
+
"eval_steps_per_second": 2.118,
|
| 651 |
+
"step": 2584
|
| 652 |
+
},
|
| 653 |
+
{
|
| 654 |
+
"epoch": 69.0,
|
| 655 |
+
"eval_accuracy": 0.9302325581395349,
|
| 656 |
+
"eval_loss": 0.420457124710083,
|
| 657 |
+
"eval_runtime": 9.2333,
|
| 658 |
+
"eval_samples_per_second": 65.199,
|
| 659 |
+
"eval_steps_per_second": 2.058,
|
| 660 |
+
"step": 2622
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 70.0,
|
| 664 |
+
"eval_accuracy": 0.9302325581395349,
|
| 665 |
+
"eval_loss": 0.4179752767086029,
|
| 666 |
+
"eval_runtime": 9.1119,
|
| 667 |
+
"eval_samples_per_second": 66.067,
|
| 668 |
+
"eval_steps_per_second": 2.085,
|
| 669 |
+
"step": 2660
|
| 670 |
+
},
|
| 671 |
+
{
|
| 672 |
+
"epoch": 71.0,
|
| 673 |
+
"eval_accuracy": 0.9285714285714286,
|
| 674 |
+
"eval_loss": 0.41662338376045227,
|
| 675 |
+
"eval_runtime": 9.2498,
|
| 676 |
+
"eval_samples_per_second": 65.082,
|
| 677 |
+
"eval_steps_per_second": 2.054,
|
| 678 |
+
"step": 2698
|
| 679 |
+
},
|
| 680 |
+
{
|
| 681 |
+
"epoch": 72.0,
|
| 682 |
+
"eval_accuracy": 0.9302325581395349,
|
| 683 |
+
"eval_loss": 0.4137563109397888,
|
| 684 |
+
"eval_runtime": 9.0837,
|
| 685 |
+
"eval_samples_per_second": 66.273,
|
| 686 |
+
"eval_steps_per_second": 2.092,
|
| 687 |
+
"step": 2736
|
| 688 |
+
},
|
| 689 |
+
{
|
| 690 |
+
"epoch": 73.0,
|
| 691 |
+
"eval_accuracy": 0.9302325581395349,
|
| 692 |
+
"eval_loss": 0.4119124710559845,
|
| 693 |
+
"eval_runtime": 9.123,
|
| 694 |
+
"eval_samples_per_second": 65.987,
|
| 695 |
+
"eval_steps_per_second": 2.083,
|
| 696 |
+
"step": 2774
|
| 697 |
+
},
|
| 698 |
+
{
|
| 699 |
+
"epoch": 74.0,
|
| 700 |
+
"eval_accuracy": 0.9352159468438538,
|
| 701 |
+
"eval_loss": 0.4097132682800293,
|
| 702 |
+
"eval_runtime": 9.2672,
|
| 703 |
+
"eval_samples_per_second": 64.96,
|
| 704 |
+
"eval_steps_per_second": 2.05,
|
| 705 |
+
"step": 2812
|
| 706 |
+
},
|
| 707 |
+
{
|
| 708 |
+
"epoch": 75.0,
|
| 709 |
+
"eval_accuracy": 0.9335548172757475,
|
| 710 |
+
"eval_loss": 0.40696021914482117,
|
| 711 |
+
"eval_runtime": 8.9895,
|
| 712 |
+
"eval_samples_per_second": 66.967,
|
| 713 |
+
"eval_steps_per_second": 2.114,
|
| 714 |
+
"step": 2850
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"epoch": 76.0,
|
| 718 |
+
"eval_accuracy": 0.9318936877076412,
|
| 719 |
+
"eval_loss": 0.40516966581344604,
|
| 720 |
+
"eval_runtime": 9.1036,
|
| 721 |
+
"eval_samples_per_second": 66.128,
|
| 722 |
+
"eval_steps_per_second": 2.087,
|
| 723 |
+
"step": 2888
|
| 724 |
+
},
|
| 725 |
+
{
|
| 726 |
+
"epoch": 77.0,
|
| 727 |
+
"eval_accuracy": 0.9335548172757475,
|
| 728 |
+
"eval_loss": 0.40282949805259705,
|
| 729 |
+
"eval_runtime": 9.1585,
|
| 730 |
+
"eval_samples_per_second": 65.731,
|
| 731 |
+
"eval_steps_per_second": 2.075,
|
| 732 |
+
"step": 2926
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"epoch": 78.0,
|
| 736 |
+
"eval_accuracy": 0.9335548172757475,
|
| 737 |
+
"eval_loss": 0.4002813994884491,
|
| 738 |
+
"eval_runtime": 9.2285,
|
| 739 |
+
"eval_samples_per_second": 65.233,
|
| 740 |
+
"eval_steps_per_second": 2.059,
|
| 741 |
+
"step": 2964
|
| 742 |
+
},
|
| 743 |
+
{
|
| 744 |
+
"epoch": 78.95,
|
| 745 |
+
"learning_rate": 8.533333333333334e-07,
|
| 746 |
+
"loss": 0.2708,
|
| 747 |
+
"step": 3000
|
| 748 |
+
},
|
| 749 |
+
{
|
| 750 |
+
"epoch": 79.0,
|
| 751 |
+
"eval_accuracy": 0.9352159468438538,
|
| 752 |
+
"eval_loss": 0.3989640772342682,
|
| 753 |
+
"eval_runtime": 9.2555,
|
| 754 |
+
"eval_samples_per_second": 65.042,
|
| 755 |
+
"eval_steps_per_second": 2.053,
|
| 756 |
+
"step": 3002
|
| 757 |
+
},
|
| 758 |
+
{
|
| 759 |
+
"epoch": 80.0,
|
| 760 |
+
"eval_accuracy": 0.9368770764119602,
|
| 761 |
+
"eval_loss": 0.3976523280143738,
|
| 762 |
+
"eval_runtime": 9.2496,
|
| 763 |
+
"eval_samples_per_second": 65.084,
|
| 764 |
+
"eval_steps_per_second": 2.054,
|
| 765 |
+
"step": 3040
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"epoch": 81.0,
|
| 769 |
+
"eval_accuracy": 0.9335548172757475,
|
| 770 |
+
"eval_loss": 0.39560216665267944,
|
| 771 |
+
"eval_runtime": 9.3445,
|
| 772 |
+
"eval_samples_per_second": 64.423,
|
| 773 |
+
"eval_steps_per_second": 2.033,
|
| 774 |
+
"step": 3078
|
| 775 |
+
},
|
| 776 |
+
{
|
| 777 |
+
"epoch": 82.0,
|
| 778 |
+
"eval_accuracy": 0.9368770764119602,
|
| 779 |
+
"eval_loss": 0.39483192563056946,
|
| 780 |
+
"eval_runtime": 9.0825,
|
| 781 |
+
"eval_samples_per_second": 66.282,
|
| 782 |
+
"eval_steps_per_second": 2.092,
|
| 783 |
+
"step": 3116
|
| 784 |
+
},
|
| 785 |
+
{
|
| 786 |
+
"epoch": 83.0,
|
| 787 |
+
"eval_accuracy": 0.9368770764119602,
|
| 788 |
+
"eval_loss": 0.3940925896167755,
|
| 789 |
+
"eval_runtime": 8.9097,
|
| 790 |
+
"eval_samples_per_second": 67.567,
|
| 791 |
+
"eval_steps_per_second": 2.133,
|
| 792 |
+
"step": 3154
|
| 793 |
+
},
|
| 794 |
+
{
|
| 795 |
+
"epoch": 84.0,
|
| 796 |
+
"eval_accuracy": 0.9368770764119602,
|
| 797 |
+
"eval_loss": 0.3927687406539917,
|
| 798 |
+
"eval_runtime": 9.2565,
|
| 799 |
+
"eval_samples_per_second": 65.035,
|
| 800 |
+
"eval_steps_per_second": 2.053,
|
| 801 |
+
"step": 3192
|
| 802 |
+
},
|
| 803 |
+
{
|
| 804 |
+
"epoch": 85.0,
|
| 805 |
+
"eval_accuracy": 0.9368770764119602,
|
| 806 |
+
"eval_loss": 0.39090487360954285,
|
| 807 |
+
"eval_runtime": 9.3794,
|
| 808 |
+
"eval_samples_per_second": 64.184,
|
| 809 |
+
"eval_steps_per_second": 2.026,
|
| 810 |
+
"step": 3230
|
| 811 |
+
},
|
| 812 |
+
{
|
| 813 |
+
"epoch": 86.0,
|
| 814 |
+
"eval_accuracy": 0.9368770764119602,
|
| 815 |
+
"eval_loss": 0.3901437222957611,
|
| 816 |
+
"eval_runtime": 9.1195,
|
| 817 |
+
"eval_samples_per_second": 66.013,
|
| 818 |
+
"eval_steps_per_second": 2.083,
|
| 819 |
+
"step": 3268
|
| 820 |
+
},
|
| 821 |
+
{
|
| 822 |
+
"epoch": 87.0,
|
| 823 |
+
"eval_accuracy": 0.9368770764119602,
|
| 824 |
+
"eval_loss": 0.3894195556640625,
|
| 825 |
+
"eval_runtime": 9.1199,
|
| 826 |
+
"eval_samples_per_second": 66.01,
|
| 827 |
+
"eval_steps_per_second": 2.083,
|
| 828 |
+
"step": 3306
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"epoch": 88.0,
|
| 832 |
+
"eval_accuracy": 0.9368770764119602,
|
| 833 |
+
"eval_loss": 0.38919058442115784,
|
| 834 |
+
"eval_runtime": 9.2628,
|
| 835 |
+
"eval_samples_per_second": 64.991,
|
| 836 |
+
"eval_steps_per_second": 2.051,
|
| 837 |
+
"step": 3344
|
| 838 |
+
},
|
| 839 |
+
{
|
| 840 |
+
"epoch": 89.0,
|
| 841 |
+
"eval_accuracy": 0.9368770764119602,
|
| 842 |
+
"eval_loss": 0.388712078332901,
|
| 843 |
+
"eval_runtime": 9.1513,
|
| 844 |
+
"eval_samples_per_second": 65.783,
|
| 845 |
+
"eval_steps_per_second": 2.076,
|
| 846 |
+
"step": 3382
|
| 847 |
+
},
|
| 848 |
+
{
|
| 849 |
+
"epoch": 90.0,
|
| 850 |
+
"eval_accuracy": 0.9368770764119602,
|
| 851 |
+
"eval_loss": 0.3878062665462494,
|
| 852 |
+
"eval_runtime": 8.9396,
|
| 853 |
+
"eval_samples_per_second": 67.341,
|
| 854 |
+
"eval_steps_per_second": 2.125,
|
| 855 |
+
"step": 3420
|
| 856 |
+
},
|
| 857 |
+
{
|
| 858 |
+
"epoch": 91.0,
|
| 859 |
+
"eval_accuracy": 0.9368770764119602,
|
| 860 |
+
"eval_loss": 0.38750186562538147,
|
| 861 |
+
"eval_runtime": 9.2575,
|
| 862 |
+
"eval_samples_per_second": 65.028,
|
| 863 |
+
"eval_steps_per_second": 2.052,
|
| 864 |
+
"step": 3458
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 92.0,
|
| 868 |
+
"eval_accuracy": 0.9385382059800664,
|
| 869 |
+
"eval_loss": 0.38650283217430115,
|
| 870 |
+
"eval_runtime": 9.1672,
|
| 871 |
+
"eval_samples_per_second": 65.669,
|
| 872 |
+
"eval_steps_per_second": 2.073,
|
| 873 |
+
"step": 3496
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 92.11,
|
| 877 |
+
"learning_rate": 3.2e-07,
|
| 878 |
+
"loss": 0.2496,
|
| 879 |
+
"step": 3500
|
| 880 |
+
},
|
| 881 |
+
{
|
| 882 |
+
"epoch": 93.0,
|
| 883 |
+
"eval_accuracy": 0.9368770764119602,
|
| 884 |
+
"eval_loss": 0.3866115212440491,
|
| 885 |
+
"eval_runtime": 9.0461,
|
| 886 |
+
"eval_samples_per_second": 66.548,
|
| 887 |
+
"eval_steps_per_second": 2.1,
|
| 888 |
+
"step": 3534
|
| 889 |
+
},
|
| 890 |
+
{
|
| 891 |
+
"epoch": 94.0,
|
| 892 |
+
"eval_accuracy": 0.9385382059800664,
|
| 893 |
+
"eval_loss": 0.3854145407676697,
|
| 894 |
+
"eval_runtime": 9.209,
|
| 895 |
+
"eval_samples_per_second": 65.371,
|
| 896 |
+
"eval_steps_per_second": 2.063,
|
| 897 |
+
"step": 3572
|
| 898 |
+
},
|
| 899 |
+
{
|
| 900 |
+
"epoch": 95.0,
|
| 901 |
+
"eval_accuracy": 0.9385382059800664,
|
| 902 |
+
"eval_loss": 0.38511377573013306,
|
| 903 |
+
"eval_runtime": 9.1643,
|
| 904 |
+
"eval_samples_per_second": 65.69,
|
| 905 |
+
"eval_steps_per_second": 2.073,
|
| 906 |
+
"step": 3610
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"epoch": 96.0,
|
| 910 |
+
"eval_accuracy": 0.9385382059800664,
|
| 911 |
+
"eval_loss": 0.38524767756462097,
|
| 912 |
+
"eval_runtime": 9.2315,
|
| 913 |
+
"eval_samples_per_second": 65.211,
|
| 914 |
+
"eval_steps_per_second": 2.058,
|
| 915 |
+
"step": 3648
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 97.0,
|
| 919 |
+
"eval_accuracy": 0.9385382059800664,
|
| 920 |
+
"eval_loss": 0.38468137383461,
|
| 921 |
+
"eval_runtime": 9.3077,
|
| 922 |
+
"eval_samples_per_second": 64.678,
|
| 923 |
+
"eval_steps_per_second": 2.041,
|
| 924 |
+
"step": 3686
|
| 925 |
+
},
|
| 926 |
+
{
|
| 927 |
+
"epoch": 98.0,
|
| 928 |
+
"eval_accuracy": 0.9385382059800664,
|
| 929 |
+
"eval_loss": 0.38459545373916626,
|
| 930 |
+
"eval_runtime": 9.2928,
|
| 931 |
+
"eval_samples_per_second": 64.781,
|
| 932 |
+
"eval_steps_per_second": 2.045,
|
| 933 |
+
"step": 3724
|
| 934 |
+
},
|
| 935 |
+
{
|
| 936 |
+
"epoch": 99.0,
|
| 937 |
+
"eval_accuracy": 0.9385382059800664,
|
| 938 |
+
"eval_loss": 0.38430511951446533,
|
| 939 |
+
"eval_runtime": 9.1498,
|
| 940 |
+
"eval_samples_per_second": 65.794,
|
| 941 |
+
"eval_steps_per_second": 2.077,
|
| 942 |
+
"step": 3762
|
| 943 |
+
},
|
| 944 |
+
{
|
| 945 |
+
"epoch": 100.0,
|
| 946 |
+
"eval_accuracy": 0.9385382059800664,
|
| 947 |
+
"eval_loss": 0.3843018710613251,
|
| 948 |
+
"eval_runtime": 9.1442,
|
| 949 |
+
"eval_samples_per_second": 65.834,
|
| 950 |
+
"eval_steps_per_second": 2.078,
|
| 951 |
+
"step": 3800
|
| 952 |
+
}
|
| 953 |
+
],
|
| 954 |
+
"logging_steps": 500,
|
| 955 |
+
"max_steps": 3800,
|
| 956 |
+
"num_train_epochs": 100,
|
| 957 |
+
"save_steps": 500,
|
| 958 |
+
"total_flos": 1.8630243068851814e+19,
|
| 959 |
+
"trial_name": null,
|
| 960 |
+
"trial_params": null
|
| 961 |
+
}
|
checkpoint-3800/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:acbec0f283828ec8d6f0c8e71a6119dbc3158d8801b0d50b9dfbeb0a01d5771c
|
| 3 |
+
size 4091
|
config.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "google/vit-base-patch16-224-in21k",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"ViTForImageClassification"
|
| 5 |
+
],
|
| 6 |
+
"attention_probs_dropout_prob": 0.0,
|
| 7 |
+
"encoder_stride": 16,
|
| 8 |
+
"hidden_act": "gelu",
|
| 9 |
+
"hidden_dropout_prob": 0.0,
|
| 10 |
+
"hidden_size": 768,
|
| 11 |
+
"id2label": {
|
| 12 |
+
"0": "Bridge",
|
| 13 |
+
"1": "Child",
|
| 14 |
+
"2": "Cobra",
|
| 15 |
+
"3": "Downward-Dog",
|
| 16 |
+
"4": "Pigeon",
|
| 17 |
+
"5": "Standing-Mountain",
|
| 18 |
+
"6": "Tree",
|
| 19 |
+
"7": "Triangle",
|
| 20 |
+
"8": "Warrior"
|
| 21 |
+
},
|
| 22 |
+
"image_size": 224,
|
| 23 |
+
"initializer_range": 0.02,
|
| 24 |
+
"intermediate_size": 3072,
|
| 25 |
+
"label2id": {
|
| 26 |
+
"Bridge": 0,
|
| 27 |
+
"Child": 1,
|
| 28 |
+
"Cobra": 2,
|
| 29 |
+
"Downward-Dog": 3,
|
| 30 |
+
"Pigeon": 4,
|
| 31 |
+
"Standing-Mountain": 5,
|
| 32 |
+
"Tree": 6,
|
| 33 |
+
"Triangle": 7,
|
| 34 |
+
"Warrior": 8
|
| 35 |
+
},
|
| 36 |
+
"layer_norm_eps": 1e-12,
|
| 37 |
+
"model_type": "vit",
|
| 38 |
+
"num_attention_heads": 12,
|
| 39 |
+
"num_channels": 3,
|
| 40 |
+
"num_hidden_layers": 12,
|
| 41 |
+
"patch_size": 16,
|
| 42 |
+
"problem_type": "single_label_classification",
|
| 43 |
+
"qkv_bias": true,
|
| 44 |
+
"torch_dtype": "float32",
|
| 45 |
+
"transformers_version": "4.35.2"
|
| 46 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:929d3e1f0a2285aa2165bbd131c14e00d8743cb7e422831240ac999593c1c4b1
|
| 3 |
+
size 343245508
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_normalize": true,
|
| 3 |
+
"do_rescale": true,
|
| 4 |
+
"do_resize": true,
|
| 5 |
+
"image_mean": [
|
| 6 |
+
0.5,
|
| 7 |
+
0.5,
|
| 8 |
+
0.5
|
| 9 |
+
],
|
| 10 |
+
"image_processor_type": "ViTImageProcessor",
|
| 11 |
+
"image_std": [
|
| 12 |
+
0.5,
|
| 13 |
+
0.5,
|
| 14 |
+
0.5
|
| 15 |
+
],
|
| 16 |
+
"resample": 2,
|
| 17 |
+
"rescale_factor": 0.00392156862745098,
|
| 18 |
+
"size": {
|
| 19 |
+
"height": 224,
|
| 20 |
+
"width": 224
|
| 21 |
+
}
|
| 22 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:acbec0f283828ec8d6f0c8e71a6119dbc3158d8801b0d50b9dfbeb0a01d5771c
|
| 3 |
+
size 4091
|