Upload folder using huggingface_hub
Browse files- checkpoint-1407/config.json +142 -0
- checkpoint-1407/model.safetensors +3 -0
- checkpoint-1407/optimizer.pt +3 -0
- checkpoint-1407/preprocessor_config.json +24 -0
- checkpoint-1407/rng_state.pth +3 -0
- checkpoint-1407/scheduler.pt +3 -0
- checkpoint-1407/trainer_state.json +58 -0
- checkpoint-1407/training_args.bin +3 -0
- checkpoint-2814/config.json +142 -0
- checkpoint-2814/model.safetensors +3 -0
- checkpoint-2814/optimizer.pt +3 -0
- checkpoint-2814/preprocessor_config.json +24 -0
- checkpoint-2814/rng_state.pth +3 -0
- checkpoint-2814/scheduler.pt +3 -0
- checkpoint-2814/trainer_state.json +89 -0
- checkpoint-2814/training_args.bin +3 -0
- checkpoint-4221/config.json +142 -0
- checkpoint-4221/model.safetensors +3 -0
- checkpoint-4221/optimizer.pt +3 -0
- checkpoint-4221/preprocessor_config.json +24 -0
- checkpoint-4221/rng_state.pth +3 -0
- checkpoint-4221/scheduler.pt +3 -0
- checkpoint-4221/trainer_state.json +120 -0
- checkpoint-4221/training_args.bin +3 -0
- config.json +142 -0
- model.safetensors +3 -0
- preprocessor_config.json +24 -0
- training_args.bin +3 -0
checkpoint-1407/config.json
ADDED
|
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "tench",
|
| 7 |
+
"1": "goldfish",
|
| 8 |
+
"2": "great white shark",
|
| 9 |
+
"3": "tiger shark",
|
| 10 |
+
"4": "hammerhead",
|
| 11 |
+
"5": "electric ray",
|
| 12 |
+
"6": "stingray",
|
| 13 |
+
"7": "cock",
|
| 14 |
+
"8": "hen",
|
| 15 |
+
"9": "ostrich",
|
| 16 |
+
"10": "brambling",
|
| 17 |
+
"11": "goldfinch",
|
| 18 |
+
"12": "house finch",
|
| 19 |
+
"13": "junco",
|
| 20 |
+
"14": "indigo bunting",
|
| 21 |
+
"15": "robin",
|
| 22 |
+
"16": "bulbul",
|
| 23 |
+
"17": "jay",
|
| 24 |
+
"18": "magpie",
|
| 25 |
+
"19": "chickadee",
|
| 26 |
+
"20": "water ouzel",
|
| 27 |
+
"21": "kite",
|
| 28 |
+
"22": "bald eagle",
|
| 29 |
+
"23": "vulture",
|
| 30 |
+
"24": "great grey owl",
|
| 31 |
+
"25": "european fire salamander",
|
| 32 |
+
"26": "common newt",
|
| 33 |
+
"27": "eft",
|
| 34 |
+
"28": "spotted salamander",
|
| 35 |
+
"29": "axolotl",
|
| 36 |
+
"30": "bullfrog",
|
| 37 |
+
"31": "tree frog",
|
| 38 |
+
"32": "tailed frog",
|
| 39 |
+
"33": "loggerhead",
|
| 40 |
+
"34": "leatherback turtle",
|
| 41 |
+
"35": "mud turtle",
|
| 42 |
+
"36": "terrapin",
|
| 43 |
+
"37": "box turtle",
|
| 44 |
+
"38": "banded gecko",
|
| 45 |
+
"39": "common iguana",
|
| 46 |
+
"40": "american chameleon",
|
| 47 |
+
"41": "whiptail",
|
| 48 |
+
"42": "agama",
|
| 49 |
+
"43": "frilled lizard",
|
| 50 |
+
"44": "alligator lizard",
|
| 51 |
+
"45": "gila monster",
|
| 52 |
+
"46": "green lizard",
|
| 53 |
+
"47": "african chameleon",
|
| 54 |
+
"48": "komodo dragon",
|
| 55 |
+
"49": "african crocodile"
|
| 56 |
+
},
|
| 57 |
+
"initializer_factor": 1.0,
|
| 58 |
+
"label2id": {
|
| 59 |
+
"african chameleon": 47,
|
| 60 |
+
"african crocodile": 49,
|
| 61 |
+
"agama": 42,
|
| 62 |
+
"alligator lizard": 44,
|
| 63 |
+
"american chameleon": 40,
|
| 64 |
+
"axolotl": 29,
|
| 65 |
+
"bald eagle": 22,
|
| 66 |
+
"banded gecko": 38,
|
| 67 |
+
"box turtle": 37,
|
| 68 |
+
"brambling": 10,
|
| 69 |
+
"bulbul": 16,
|
| 70 |
+
"bullfrog": 30,
|
| 71 |
+
"chickadee": 19,
|
| 72 |
+
"cock": 7,
|
| 73 |
+
"common iguana": 39,
|
| 74 |
+
"common newt": 26,
|
| 75 |
+
"eft": 27,
|
| 76 |
+
"electric ray": 5,
|
| 77 |
+
"european fire salamander": 25,
|
| 78 |
+
"frilled lizard": 43,
|
| 79 |
+
"gila monster": 45,
|
| 80 |
+
"goldfinch": 11,
|
| 81 |
+
"goldfish": 1,
|
| 82 |
+
"great grey owl": 24,
|
| 83 |
+
"great white shark": 2,
|
| 84 |
+
"green lizard": 46,
|
| 85 |
+
"hammerhead": 4,
|
| 86 |
+
"hen": 8,
|
| 87 |
+
"house finch": 12,
|
| 88 |
+
"indigo bunting": 14,
|
| 89 |
+
"jay": 17,
|
| 90 |
+
"junco": 13,
|
| 91 |
+
"kite": 21,
|
| 92 |
+
"komodo dragon": 48,
|
| 93 |
+
"leatherback turtle": 34,
|
| 94 |
+
"loggerhead": 33,
|
| 95 |
+
"magpie": 18,
|
| 96 |
+
"mud turtle": 35,
|
| 97 |
+
"ostrich": 9,
|
| 98 |
+
"robin": 15,
|
| 99 |
+
"spotted salamander": 28,
|
| 100 |
+
"stingray": 6,
|
| 101 |
+
"tailed frog": 32,
|
| 102 |
+
"tench": 0,
|
| 103 |
+
"terrapin": 36,
|
| 104 |
+
"tiger shark": 3,
|
| 105 |
+
"tree frog": 31,
|
| 106 |
+
"vulture": 23,
|
| 107 |
+
"water ouzel": 20,
|
| 108 |
+
"whiptail": 41
|
| 109 |
+
},
|
| 110 |
+
"model_type": "siglip",
|
| 111 |
+
"problem_type": "single_label_classification",
|
| 112 |
+
"text_config": {
|
| 113 |
+
"attention_dropout": 0.0,
|
| 114 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 115 |
+
"hidden_size": 768,
|
| 116 |
+
"intermediate_size": 3072,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"max_position_embeddings": 64,
|
| 119 |
+
"model_type": "siglip_text_model",
|
| 120 |
+
"num_attention_heads": 12,
|
| 121 |
+
"num_hidden_layers": 12,
|
| 122 |
+
"projection_size": 768,
|
| 123 |
+
"torch_dtype": "float32",
|
| 124 |
+
"vocab_size": 256000
|
| 125 |
+
},
|
| 126 |
+
"torch_dtype": "float32",
|
| 127 |
+
"transformers_version": "4.50.0",
|
| 128 |
+
"vision_config": {
|
| 129 |
+
"attention_dropout": 0.0,
|
| 130 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 131 |
+
"hidden_size": 768,
|
| 132 |
+
"image_size": 224,
|
| 133 |
+
"intermediate_size": 3072,
|
| 134 |
+
"layer_norm_eps": 1e-06,
|
| 135 |
+
"model_type": "siglip_vision_model",
|
| 136 |
+
"num_attention_heads": 12,
|
| 137 |
+
"num_channels": 3,
|
| 138 |
+
"num_hidden_layers": 12,
|
| 139 |
+
"patch_size": 16,
|
| 140 |
+
"torch_dtype": "float32"
|
| 141 |
+
}
|
| 142 |
+
}
|
checkpoint-1407/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b14f24be37b81d0006d71ab0359d7e68ac432e92472e3fdc508d85ddf0a47e6e
|
| 3 |
+
size 371715648
|
checkpoint-1407/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:28a396bb52fcb662e6361d1463185cf07d12ab2026497708a727e7566dc801cd
|
| 3 |
+
size 686851066
|
checkpoint-1407/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
checkpoint-1407/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b74b26745296cb9bb5c1b1141c11bc7e676b763a253302db10a4eb4b6db1f536
|
| 3 |
+
size 14244
|
checkpoint-1407/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:28bca237ff83fd098f3b467970fe6030c8b0da7604e1cdfe97c9b03b92b9d96d
|
| 3 |
+
size 1064
|
checkpoint-1407/trainer_state.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 1407,
|
| 3 |
+
"best_metric": 0.37436172366142273,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-1407",
|
| 5 |
+
"epoch": 1.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 1407,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.35536602700781805,
|
| 14 |
+
"grad_norm": 55.63911437988281,
|
| 15 |
+
"learning_rate": 4.460561016542796e-06,
|
| 16 |
+
"loss": 1.6819,
|
| 17 |
+
"step": 500
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.7107320540156361,
|
| 21 |
+
"grad_norm": 49.583003997802734,
|
| 22 |
+
"learning_rate": 3.861184368257013e-06,
|
| 23 |
+
"loss": 0.5948,
|
| 24 |
+
"step": 1000
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 1.0,
|
| 28 |
+
"eval_accuracy": 0.8836888888888889,
|
| 29 |
+
"eval_loss": 0.37436172366142273,
|
| 30 |
+
"eval_model_preparation_time": 0.0024,
|
| 31 |
+
"eval_runtime": 700.9364,
|
| 32 |
+
"eval_samples_per_second": 64.2,
|
| 33 |
+
"eval_steps_per_second": 8.025,
|
| 34 |
+
"step": 1407
|
| 35 |
+
}
|
| 36 |
+
],
|
| 37 |
+
"logging_steps": 500,
|
| 38 |
+
"max_steps": 4221,
|
| 39 |
+
"num_input_tokens_seen": 0,
|
| 40 |
+
"num_train_epochs": 3,
|
| 41 |
+
"save_steps": 500,
|
| 42 |
+
"stateful_callbacks": {
|
| 43 |
+
"TrainerControl": {
|
| 44 |
+
"args": {
|
| 45 |
+
"should_epoch_stop": false,
|
| 46 |
+
"should_evaluate": false,
|
| 47 |
+
"should_log": false,
|
| 48 |
+
"should_save": true,
|
| 49 |
+
"should_training_stop": false
|
| 50 |
+
},
|
| 51 |
+
"attributes": {}
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"total_flos": 3.77049751013376e+18,
|
| 55 |
+
"train_batch_size": 32,
|
| 56 |
+
"trial_name": null,
|
| 57 |
+
"trial_params": null
|
| 58 |
+
}
|
checkpoint-1407/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8d23e4dd21d70dc3decbc150e22ae936ad50be95a366a34502fabd29ae06140
|
| 3 |
+
size 5304
|
checkpoint-2814/config.json
ADDED
|
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "tench",
|
| 7 |
+
"1": "goldfish",
|
| 8 |
+
"2": "great white shark",
|
| 9 |
+
"3": "tiger shark",
|
| 10 |
+
"4": "hammerhead",
|
| 11 |
+
"5": "electric ray",
|
| 12 |
+
"6": "stingray",
|
| 13 |
+
"7": "cock",
|
| 14 |
+
"8": "hen",
|
| 15 |
+
"9": "ostrich",
|
| 16 |
+
"10": "brambling",
|
| 17 |
+
"11": "goldfinch",
|
| 18 |
+
"12": "house finch",
|
| 19 |
+
"13": "junco",
|
| 20 |
+
"14": "indigo bunting",
|
| 21 |
+
"15": "robin",
|
| 22 |
+
"16": "bulbul",
|
| 23 |
+
"17": "jay",
|
| 24 |
+
"18": "magpie",
|
| 25 |
+
"19": "chickadee",
|
| 26 |
+
"20": "water ouzel",
|
| 27 |
+
"21": "kite",
|
| 28 |
+
"22": "bald eagle",
|
| 29 |
+
"23": "vulture",
|
| 30 |
+
"24": "great grey owl",
|
| 31 |
+
"25": "european fire salamander",
|
| 32 |
+
"26": "common newt",
|
| 33 |
+
"27": "eft",
|
| 34 |
+
"28": "spotted salamander",
|
| 35 |
+
"29": "axolotl",
|
| 36 |
+
"30": "bullfrog",
|
| 37 |
+
"31": "tree frog",
|
| 38 |
+
"32": "tailed frog",
|
| 39 |
+
"33": "loggerhead",
|
| 40 |
+
"34": "leatherback turtle",
|
| 41 |
+
"35": "mud turtle",
|
| 42 |
+
"36": "terrapin",
|
| 43 |
+
"37": "box turtle",
|
| 44 |
+
"38": "banded gecko",
|
| 45 |
+
"39": "common iguana",
|
| 46 |
+
"40": "american chameleon",
|
| 47 |
+
"41": "whiptail",
|
| 48 |
+
"42": "agama",
|
| 49 |
+
"43": "frilled lizard",
|
| 50 |
+
"44": "alligator lizard",
|
| 51 |
+
"45": "gila monster",
|
| 52 |
+
"46": "green lizard",
|
| 53 |
+
"47": "african chameleon",
|
| 54 |
+
"48": "komodo dragon",
|
| 55 |
+
"49": "african crocodile"
|
| 56 |
+
},
|
| 57 |
+
"initializer_factor": 1.0,
|
| 58 |
+
"label2id": {
|
| 59 |
+
"african chameleon": 47,
|
| 60 |
+
"african crocodile": 49,
|
| 61 |
+
"agama": 42,
|
| 62 |
+
"alligator lizard": 44,
|
| 63 |
+
"american chameleon": 40,
|
| 64 |
+
"axolotl": 29,
|
| 65 |
+
"bald eagle": 22,
|
| 66 |
+
"banded gecko": 38,
|
| 67 |
+
"box turtle": 37,
|
| 68 |
+
"brambling": 10,
|
| 69 |
+
"bulbul": 16,
|
| 70 |
+
"bullfrog": 30,
|
| 71 |
+
"chickadee": 19,
|
| 72 |
+
"cock": 7,
|
| 73 |
+
"common iguana": 39,
|
| 74 |
+
"common newt": 26,
|
| 75 |
+
"eft": 27,
|
| 76 |
+
"electric ray": 5,
|
| 77 |
+
"european fire salamander": 25,
|
| 78 |
+
"frilled lizard": 43,
|
| 79 |
+
"gila monster": 45,
|
| 80 |
+
"goldfinch": 11,
|
| 81 |
+
"goldfish": 1,
|
| 82 |
+
"great grey owl": 24,
|
| 83 |
+
"great white shark": 2,
|
| 84 |
+
"green lizard": 46,
|
| 85 |
+
"hammerhead": 4,
|
| 86 |
+
"hen": 8,
|
| 87 |
+
"house finch": 12,
|
| 88 |
+
"indigo bunting": 14,
|
| 89 |
+
"jay": 17,
|
| 90 |
+
"junco": 13,
|
| 91 |
+
"kite": 21,
|
| 92 |
+
"komodo dragon": 48,
|
| 93 |
+
"leatherback turtle": 34,
|
| 94 |
+
"loggerhead": 33,
|
| 95 |
+
"magpie": 18,
|
| 96 |
+
"mud turtle": 35,
|
| 97 |
+
"ostrich": 9,
|
| 98 |
+
"robin": 15,
|
| 99 |
+
"spotted salamander": 28,
|
| 100 |
+
"stingray": 6,
|
| 101 |
+
"tailed frog": 32,
|
| 102 |
+
"tench": 0,
|
| 103 |
+
"terrapin": 36,
|
| 104 |
+
"tiger shark": 3,
|
| 105 |
+
"tree frog": 31,
|
| 106 |
+
"vulture": 23,
|
| 107 |
+
"water ouzel": 20,
|
| 108 |
+
"whiptail": 41
|
| 109 |
+
},
|
| 110 |
+
"model_type": "siglip",
|
| 111 |
+
"problem_type": "single_label_classification",
|
| 112 |
+
"text_config": {
|
| 113 |
+
"attention_dropout": 0.0,
|
| 114 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 115 |
+
"hidden_size": 768,
|
| 116 |
+
"intermediate_size": 3072,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"max_position_embeddings": 64,
|
| 119 |
+
"model_type": "siglip_text_model",
|
| 120 |
+
"num_attention_heads": 12,
|
| 121 |
+
"num_hidden_layers": 12,
|
| 122 |
+
"projection_size": 768,
|
| 123 |
+
"torch_dtype": "float32",
|
| 124 |
+
"vocab_size": 256000
|
| 125 |
+
},
|
| 126 |
+
"torch_dtype": "float32",
|
| 127 |
+
"transformers_version": "4.50.0",
|
| 128 |
+
"vision_config": {
|
| 129 |
+
"attention_dropout": 0.0,
|
| 130 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 131 |
+
"hidden_size": 768,
|
| 132 |
+
"image_size": 224,
|
| 133 |
+
"intermediate_size": 3072,
|
| 134 |
+
"layer_norm_eps": 1e-06,
|
| 135 |
+
"model_type": "siglip_vision_model",
|
| 136 |
+
"num_attention_heads": 12,
|
| 137 |
+
"num_channels": 3,
|
| 138 |
+
"num_hidden_layers": 12,
|
| 139 |
+
"patch_size": 16,
|
| 140 |
+
"torch_dtype": "float32"
|
| 141 |
+
}
|
| 142 |
+
}
|
checkpoint-2814/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:75b86b367aa40b3080b1321586e8d66f2ee997b46c440663ab064534345cd7e4
|
| 3 |
+
size 371715648
|
checkpoint-2814/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb120407e2089eef577da13cd77ef821adaf576184766a69e45025b49d6fd305
|
| 3 |
+
size 686851066
|
checkpoint-2814/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
checkpoint-2814/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e50f194b7053263dfc0231a7bc774563a66466226c5fb2c7d3ba52c2e81382fa
|
| 3 |
+
size 14244
|
checkpoint-2814/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b24f30bec061c3b66513dca2ba910b57cd3a3885186f8e076fccc60cd3bd3c79
|
| 3 |
+
size 1064
|
checkpoint-2814/trainer_state.json
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 2814,
|
| 3 |
+
"best_metric": 0.2967633605003357,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-2814",
|
| 5 |
+
"epoch": 2.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 2814,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.35536602700781805,
|
| 14 |
+
"grad_norm": 55.63911437988281,
|
| 15 |
+
"learning_rate": 4.460561016542796e-06,
|
| 16 |
+
"loss": 1.6819,
|
| 17 |
+
"step": 500
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.7107320540156361,
|
| 21 |
+
"grad_norm": 49.583003997802734,
|
| 22 |
+
"learning_rate": 3.861184368257013e-06,
|
| 23 |
+
"loss": 0.5948,
|
| 24 |
+
"step": 1000
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 1.0,
|
| 28 |
+
"eval_accuracy": 0.8836888888888889,
|
| 29 |
+
"eval_loss": 0.37436172366142273,
|
| 30 |
+
"eval_model_preparation_time": 0.0024,
|
| 31 |
+
"eval_runtime": 700.9364,
|
| 32 |
+
"eval_samples_per_second": 64.2,
|
| 33 |
+
"eval_steps_per_second": 8.025,
|
| 34 |
+
"step": 1407
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 1.0660980810234542,
|
| 38 |
+
"grad_norm": 53.438941955566406,
|
| 39 |
+
"learning_rate": 3.26180771997123e-06,
|
| 40 |
+
"loss": 0.4896,
|
| 41 |
+
"step": 1500
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 1.4214641080312722,
|
| 45 |
+
"grad_norm": 36.83859634399414,
|
| 46 |
+
"learning_rate": 2.6624310716854473e-06,
|
| 47 |
+
"loss": 0.4047,
|
| 48 |
+
"step": 2000
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"epoch": 1.7768301350390903,
|
| 52 |
+
"grad_norm": 36.33766174316406,
|
| 53 |
+
"learning_rate": 2.063054423399664e-06,
|
| 54 |
+
"loss": 0.3934,
|
| 55 |
+
"step": 2500
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 2.0,
|
| 59 |
+
"eval_accuracy": 0.9048,
|
| 60 |
+
"eval_loss": 0.2967633605003357,
|
| 61 |
+
"eval_model_preparation_time": 0.0024,
|
| 62 |
+
"eval_runtime": 693.599,
|
| 63 |
+
"eval_samples_per_second": 64.879,
|
| 64 |
+
"eval_steps_per_second": 8.11,
|
| 65 |
+
"step": 2814
|
| 66 |
+
}
|
| 67 |
+
],
|
| 68 |
+
"logging_steps": 500,
|
| 69 |
+
"max_steps": 4221,
|
| 70 |
+
"num_input_tokens_seen": 0,
|
| 71 |
+
"num_train_epochs": 3,
|
| 72 |
+
"save_steps": 500,
|
| 73 |
+
"stateful_callbacks": {
|
| 74 |
+
"TrainerControl": {
|
| 75 |
+
"args": {
|
| 76 |
+
"should_epoch_stop": false,
|
| 77 |
+
"should_evaluate": false,
|
| 78 |
+
"should_log": false,
|
| 79 |
+
"should_save": true,
|
| 80 |
+
"should_training_stop": false
|
| 81 |
+
},
|
| 82 |
+
"attributes": {}
|
| 83 |
+
}
|
| 84 |
+
},
|
| 85 |
+
"total_flos": 7.54099502026752e+18,
|
| 86 |
+
"train_batch_size": 32,
|
| 87 |
+
"trial_name": null,
|
| 88 |
+
"trial_params": null
|
| 89 |
+
}
|
checkpoint-2814/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8d23e4dd21d70dc3decbc150e22ae936ad50be95a366a34502fabd29ae06140
|
| 3 |
+
size 5304
|
checkpoint-4221/config.json
ADDED
|
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "tench",
|
| 7 |
+
"1": "goldfish",
|
| 8 |
+
"2": "great white shark",
|
| 9 |
+
"3": "tiger shark",
|
| 10 |
+
"4": "hammerhead",
|
| 11 |
+
"5": "electric ray",
|
| 12 |
+
"6": "stingray",
|
| 13 |
+
"7": "cock",
|
| 14 |
+
"8": "hen",
|
| 15 |
+
"9": "ostrich",
|
| 16 |
+
"10": "brambling",
|
| 17 |
+
"11": "goldfinch",
|
| 18 |
+
"12": "house finch",
|
| 19 |
+
"13": "junco",
|
| 20 |
+
"14": "indigo bunting",
|
| 21 |
+
"15": "robin",
|
| 22 |
+
"16": "bulbul",
|
| 23 |
+
"17": "jay",
|
| 24 |
+
"18": "magpie",
|
| 25 |
+
"19": "chickadee",
|
| 26 |
+
"20": "water ouzel",
|
| 27 |
+
"21": "kite",
|
| 28 |
+
"22": "bald eagle",
|
| 29 |
+
"23": "vulture",
|
| 30 |
+
"24": "great grey owl",
|
| 31 |
+
"25": "european fire salamander",
|
| 32 |
+
"26": "common newt",
|
| 33 |
+
"27": "eft",
|
| 34 |
+
"28": "spotted salamander",
|
| 35 |
+
"29": "axolotl",
|
| 36 |
+
"30": "bullfrog",
|
| 37 |
+
"31": "tree frog",
|
| 38 |
+
"32": "tailed frog",
|
| 39 |
+
"33": "loggerhead",
|
| 40 |
+
"34": "leatherback turtle",
|
| 41 |
+
"35": "mud turtle",
|
| 42 |
+
"36": "terrapin",
|
| 43 |
+
"37": "box turtle",
|
| 44 |
+
"38": "banded gecko",
|
| 45 |
+
"39": "common iguana",
|
| 46 |
+
"40": "american chameleon",
|
| 47 |
+
"41": "whiptail",
|
| 48 |
+
"42": "agama",
|
| 49 |
+
"43": "frilled lizard",
|
| 50 |
+
"44": "alligator lizard",
|
| 51 |
+
"45": "gila monster",
|
| 52 |
+
"46": "green lizard",
|
| 53 |
+
"47": "african chameleon",
|
| 54 |
+
"48": "komodo dragon",
|
| 55 |
+
"49": "african crocodile"
|
| 56 |
+
},
|
| 57 |
+
"initializer_factor": 1.0,
|
| 58 |
+
"label2id": {
|
| 59 |
+
"african chameleon": 47,
|
| 60 |
+
"african crocodile": 49,
|
| 61 |
+
"agama": 42,
|
| 62 |
+
"alligator lizard": 44,
|
| 63 |
+
"american chameleon": 40,
|
| 64 |
+
"axolotl": 29,
|
| 65 |
+
"bald eagle": 22,
|
| 66 |
+
"banded gecko": 38,
|
| 67 |
+
"box turtle": 37,
|
| 68 |
+
"brambling": 10,
|
| 69 |
+
"bulbul": 16,
|
| 70 |
+
"bullfrog": 30,
|
| 71 |
+
"chickadee": 19,
|
| 72 |
+
"cock": 7,
|
| 73 |
+
"common iguana": 39,
|
| 74 |
+
"common newt": 26,
|
| 75 |
+
"eft": 27,
|
| 76 |
+
"electric ray": 5,
|
| 77 |
+
"european fire salamander": 25,
|
| 78 |
+
"frilled lizard": 43,
|
| 79 |
+
"gila monster": 45,
|
| 80 |
+
"goldfinch": 11,
|
| 81 |
+
"goldfish": 1,
|
| 82 |
+
"great grey owl": 24,
|
| 83 |
+
"great white shark": 2,
|
| 84 |
+
"green lizard": 46,
|
| 85 |
+
"hammerhead": 4,
|
| 86 |
+
"hen": 8,
|
| 87 |
+
"house finch": 12,
|
| 88 |
+
"indigo bunting": 14,
|
| 89 |
+
"jay": 17,
|
| 90 |
+
"junco": 13,
|
| 91 |
+
"kite": 21,
|
| 92 |
+
"komodo dragon": 48,
|
| 93 |
+
"leatherback turtle": 34,
|
| 94 |
+
"loggerhead": 33,
|
| 95 |
+
"magpie": 18,
|
| 96 |
+
"mud turtle": 35,
|
| 97 |
+
"ostrich": 9,
|
| 98 |
+
"robin": 15,
|
| 99 |
+
"spotted salamander": 28,
|
| 100 |
+
"stingray": 6,
|
| 101 |
+
"tailed frog": 32,
|
| 102 |
+
"tench": 0,
|
| 103 |
+
"terrapin": 36,
|
| 104 |
+
"tiger shark": 3,
|
| 105 |
+
"tree frog": 31,
|
| 106 |
+
"vulture": 23,
|
| 107 |
+
"water ouzel": 20,
|
| 108 |
+
"whiptail": 41
|
| 109 |
+
},
|
| 110 |
+
"model_type": "siglip",
|
| 111 |
+
"problem_type": "single_label_classification",
|
| 112 |
+
"text_config": {
|
| 113 |
+
"attention_dropout": 0.0,
|
| 114 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 115 |
+
"hidden_size": 768,
|
| 116 |
+
"intermediate_size": 3072,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"max_position_embeddings": 64,
|
| 119 |
+
"model_type": "siglip_text_model",
|
| 120 |
+
"num_attention_heads": 12,
|
| 121 |
+
"num_hidden_layers": 12,
|
| 122 |
+
"projection_size": 768,
|
| 123 |
+
"torch_dtype": "float32",
|
| 124 |
+
"vocab_size": 256000
|
| 125 |
+
},
|
| 126 |
+
"torch_dtype": "float32",
|
| 127 |
+
"transformers_version": "4.50.0",
|
| 128 |
+
"vision_config": {
|
| 129 |
+
"attention_dropout": 0.0,
|
| 130 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 131 |
+
"hidden_size": 768,
|
| 132 |
+
"image_size": 224,
|
| 133 |
+
"intermediate_size": 3072,
|
| 134 |
+
"layer_norm_eps": 1e-06,
|
| 135 |
+
"model_type": "siglip_vision_model",
|
| 136 |
+
"num_attention_heads": 12,
|
| 137 |
+
"num_channels": 3,
|
| 138 |
+
"num_hidden_layers": 12,
|
| 139 |
+
"patch_size": 16,
|
| 140 |
+
"torch_dtype": "float32"
|
| 141 |
+
}
|
| 142 |
+
}
|
checkpoint-4221/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:edd0974741a9d968692d2932133f6324506d2ba3d702002c7cd320e332bc39cf
|
| 3 |
+
size 371715648
|
checkpoint-4221/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c75ccf3011478e3428bd5395d95b615a5aaedbabd2d5ca539b73946c0a7d115d
|
| 3 |
+
size 686851066
|
checkpoint-4221/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
checkpoint-4221/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7a24d1253a051df5f81ee82b622f907ca1a151743c2afa8ac16fbc6bd7d371f4
|
| 3 |
+
size 14244
|
checkpoint-4221/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d72dff37292982bb8f7bc7b3253be747a4cc03055cc68c0bfdc7e0fdd3685682
|
| 3 |
+
size 1064
|
checkpoint-4221/trainer_state.json
ADDED
|
@@ -0,0 +1,120 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 4221,
|
| 3 |
+
"best_metric": 0.2564261853694916,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-4221",
|
| 5 |
+
"epoch": 3.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 4221,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.35536602700781805,
|
| 14 |
+
"grad_norm": 55.63911437988281,
|
| 15 |
+
"learning_rate": 4.460561016542796e-06,
|
| 16 |
+
"loss": 1.6819,
|
| 17 |
+
"step": 500
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.7107320540156361,
|
| 21 |
+
"grad_norm": 49.583003997802734,
|
| 22 |
+
"learning_rate": 3.861184368257013e-06,
|
| 23 |
+
"loss": 0.5948,
|
| 24 |
+
"step": 1000
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 1.0,
|
| 28 |
+
"eval_accuracy": 0.8836888888888889,
|
| 29 |
+
"eval_loss": 0.37436172366142273,
|
| 30 |
+
"eval_model_preparation_time": 0.0024,
|
| 31 |
+
"eval_runtime": 700.9364,
|
| 32 |
+
"eval_samples_per_second": 64.2,
|
| 33 |
+
"eval_steps_per_second": 8.025,
|
| 34 |
+
"step": 1407
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 1.0660980810234542,
|
| 38 |
+
"grad_norm": 53.438941955566406,
|
| 39 |
+
"learning_rate": 3.26180771997123e-06,
|
| 40 |
+
"loss": 0.4896,
|
| 41 |
+
"step": 1500
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 1.4214641080312722,
|
| 45 |
+
"grad_norm": 36.83859634399414,
|
| 46 |
+
"learning_rate": 2.6624310716854473e-06,
|
| 47 |
+
"loss": 0.4047,
|
| 48 |
+
"step": 2000
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"epoch": 1.7768301350390903,
|
| 52 |
+
"grad_norm": 36.33766174316406,
|
| 53 |
+
"learning_rate": 2.063054423399664e-06,
|
| 54 |
+
"loss": 0.3934,
|
| 55 |
+
"step": 2500
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 2.0,
|
| 59 |
+
"eval_accuracy": 0.9048,
|
| 60 |
+
"eval_loss": 0.2967633605003357,
|
| 61 |
+
"eval_model_preparation_time": 0.0024,
|
| 62 |
+
"eval_runtime": 693.599,
|
| 63 |
+
"eval_samples_per_second": 64.879,
|
| 64 |
+
"eval_steps_per_second": 8.11,
|
| 65 |
+
"step": 2814
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 2.1321961620469083,
|
| 69 |
+
"grad_norm": 19.034664154052734,
|
| 70 |
+
"learning_rate": 1.4636777751138817e-06,
|
| 71 |
+
"loss": 0.3444,
|
| 72 |
+
"step": 3000
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 2.487562189054726,
|
| 76 |
+
"grad_norm": 17.526777267456055,
|
| 77 |
+
"learning_rate": 8.643011268280989e-07,
|
| 78 |
+
"loss": 0.3002,
|
| 79 |
+
"step": 3500
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 2.8429282160625444,
|
| 83 |
+
"grad_norm": 35.142356872558594,
|
| 84 |
+
"learning_rate": 2.64924478542316e-07,
|
| 85 |
+
"loss": 0.3094,
|
| 86 |
+
"step": 4000
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 3.0,
|
| 90 |
+
"eval_accuracy": 0.9181333333333334,
|
| 91 |
+
"eval_loss": 0.2564261853694916,
|
| 92 |
+
"eval_model_preparation_time": 0.0024,
|
| 93 |
+
"eval_runtime": 694.172,
|
| 94 |
+
"eval_samples_per_second": 64.825,
|
| 95 |
+
"eval_steps_per_second": 8.103,
|
| 96 |
+
"step": 4221
|
| 97 |
+
}
|
| 98 |
+
],
|
| 99 |
+
"logging_steps": 500,
|
| 100 |
+
"max_steps": 4221,
|
| 101 |
+
"num_input_tokens_seen": 0,
|
| 102 |
+
"num_train_epochs": 3,
|
| 103 |
+
"save_steps": 500,
|
| 104 |
+
"stateful_callbacks": {
|
| 105 |
+
"TrainerControl": {
|
| 106 |
+
"args": {
|
| 107 |
+
"should_epoch_stop": false,
|
| 108 |
+
"should_evaluate": false,
|
| 109 |
+
"should_log": false,
|
| 110 |
+
"should_save": true,
|
| 111 |
+
"should_training_stop": true
|
| 112 |
+
},
|
| 113 |
+
"attributes": {}
|
| 114 |
+
}
|
| 115 |
+
},
|
| 116 |
+
"total_flos": 1.131149253040128e+19,
|
| 117 |
+
"train_batch_size": 32,
|
| 118 |
+
"trial_name": null,
|
| 119 |
+
"trial_params": null
|
| 120 |
+
}
|
checkpoint-4221/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8d23e4dd21d70dc3decbc150e22ae936ad50be95a366a34502fabd29ae06140
|
| 3 |
+
size 5304
|
config.json
ADDED
|
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "tench",
|
| 7 |
+
"1": "goldfish",
|
| 8 |
+
"2": "great white shark",
|
| 9 |
+
"3": "tiger shark",
|
| 10 |
+
"4": "hammerhead",
|
| 11 |
+
"5": "electric ray",
|
| 12 |
+
"6": "stingray",
|
| 13 |
+
"7": "cock",
|
| 14 |
+
"8": "hen",
|
| 15 |
+
"9": "ostrich",
|
| 16 |
+
"10": "brambling",
|
| 17 |
+
"11": "goldfinch",
|
| 18 |
+
"12": "house finch",
|
| 19 |
+
"13": "junco",
|
| 20 |
+
"14": "indigo bunting",
|
| 21 |
+
"15": "robin",
|
| 22 |
+
"16": "bulbul",
|
| 23 |
+
"17": "jay",
|
| 24 |
+
"18": "magpie",
|
| 25 |
+
"19": "chickadee",
|
| 26 |
+
"20": "water ouzel",
|
| 27 |
+
"21": "kite",
|
| 28 |
+
"22": "bald eagle",
|
| 29 |
+
"23": "vulture",
|
| 30 |
+
"24": "great grey owl",
|
| 31 |
+
"25": "european fire salamander",
|
| 32 |
+
"26": "common newt",
|
| 33 |
+
"27": "eft",
|
| 34 |
+
"28": "spotted salamander",
|
| 35 |
+
"29": "axolotl",
|
| 36 |
+
"30": "bullfrog",
|
| 37 |
+
"31": "tree frog",
|
| 38 |
+
"32": "tailed frog",
|
| 39 |
+
"33": "loggerhead",
|
| 40 |
+
"34": "leatherback turtle",
|
| 41 |
+
"35": "mud turtle",
|
| 42 |
+
"36": "terrapin",
|
| 43 |
+
"37": "box turtle",
|
| 44 |
+
"38": "banded gecko",
|
| 45 |
+
"39": "common iguana",
|
| 46 |
+
"40": "american chameleon",
|
| 47 |
+
"41": "whiptail",
|
| 48 |
+
"42": "agama",
|
| 49 |
+
"43": "frilled lizard",
|
| 50 |
+
"44": "alligator lizard",
|
| 51 |
+
"45": "gila monster",
|
| 52 |
+
"46": "green lizard",
|
| 53 |
+
"47": "african chameleon",
|
| 54 |
+
"48": "komodo dragon",
|
| 55 |
+
"49": "african crocodile"
|
| 56 |
+
},
|
| 57 |
+
"initializer_factor": 1.0,
|
| 58 |
+
"label2id": {
|
| 59 |
+
"african chameleon": 47,
|
| 60 |
+
"african crocodile": 49,
|
| 61 |
+
"agama": 42,
|
| 62 |
+
"alligator lizard": 44,
|
| 63 |
+
"american chameleon": 40,
|
| 64 |
+
"axolotl": 29,
|
| 65 |
+
"bald eagle": 22,
|
| 66 |
+
"banded gecko": 38,
|
| 67 |
+
"box turtle": 37,
|
| 68 |
+
"brambling": 10,
|
| 69 |
+
"bulbul": 16,
|
| 70 |
+
"bullfrog": 30,
|
| 71 |
+
"chickadee": 19,
|
| 72 |
+
"cock": 7,
|
| 73 |
+
"common iguana": 39,
|
| 74 |
+
"common newt": 26,
|
| 75 |
+
"eft": 27,
|
| 76 |
+
"electric ray": 5,
|
| 77 |
+
"european fire salamander": 25,
|
| 78 |
+
"frilled lizard": 43,
|
| 79 |
+
"gila monster": 45,
|
| 80 |
+
"goldfinch": 11,
|
| 81 |
+
"goldfish": 1,
|
| 82 |
+
"great grey owl": 24,
|
| 83 |
+
"great white shark": 2,
|
| 84 |
+
"green lizard": 46,
|
| 85 |
+
"hammerhead": 4,
|
| 86 |
+
"hen": 8,
|
| 87 |
+
"house finch": 12,
|
| 88 |
+
"indigo bunting": 14,
|
| 89 |
+
"jay": 17,
|
| 90 |
+
"junco": 13,
|
| 91 |
+
"kite": 21,
|
| 92 |
+
"komodo dragon": 48,
|
| 93 |
+
"leatherback turtle": 34,
|
| 94 |
+
"loggerhead": 33,
|
| 95 |
+
"magpie": 18,
|
| 96 |
+
"mud turtle": 35,
|
| 97 |
+
"ostrich": 9,
|
| 98 |
+
"robin": 15,
|
| 99 |
+
"spotted salamander": 28,
|
| 100 |
+
"stingray": 6,
|
| 101 |
+
"tailed frog": 32,
|
| 102 |
+
"tench": 0,
|
| 103 |
+
"terrapin": 36,
|
| 104 |
+
"tiger shark": 3,
|
| 105 |
+
"tree frog": 31,
|
| 106 |
+
"vulture": 23,
|
| 107 |
+
"water ouzel": 20,
|
| 108 |
+
"whiptail": 41
|
| 109 |
+
},
|
| 110 |
+
"model_type": "siglip",
|
| 111 |
+
"problem_type": "single_label_classification",
|
| 112 |
+
"text_config": {
|
| 113 |
+
"attention_dropout": 0.0,
|
| 114 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 115 |
+
"hidden_size": 768,
|
| 116 |
+
"intermediate_size": 3072,
|
| 117 |
+
"layer_norm_eps": 1e-06,
|
| 118 |
+
"max_position_embeddings": 64,
|
| 119 |
+
"model_type": "siglip_text_model",
|
| 120 |
+
"num_attention_heads": 12,
|
| 121 |
+
"num_hidden_layers": 12,
|
| 122 |
+
"projection_size": 768,
|
| 123 |
+
"torch_dtype": "float32",
|
| 124 |
+
"vocab_size": 256000
|
| 125 |
+
},
|
| 126 |
+
"torch_dtype": "float32",
|
| 127 |
+
"transformers_version": "4.50.0",
|
| 128 |
+
"vision_config": {
|
| 129 |
+
"attention_dropout": 0.0,
|
| 130 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 131 |
+
"hidden_size": 768,
|
| 132 |
+
"image_size": 224,
|
| 133 |
+
"intermediate_size": 3072,
|
| 134 |
+
"layer_norm_eps": 1e-06,
|
| 135 |
+
"model_type": "siglip_vision_model",
|
| 136 |
+
"num_attention_heads": 12,
|
| 137 |
+
"num_channels": 3,
|
| 138 |
+
"num_hidden_layers": 12,
|
| 139 |
+
"patch_size": 16,
|
| 140 |
+
"torch_dtype": "float32"
|
| 141 |
+
}
|
| 142 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:edd0974741a9d968692d2932133f6324506d2ba3d702002c7cd320e332bc39cf
|
| 3 |
+
size 371715648
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8d23e4dd21d70dc3decbc150e22ae936ad50be95a366a34502fabd29ae06140
|
| 3 |
+
size 5304
|