Eric Trained Model Nov 14

.gitattributes

@@ -0,0 +1 @@
+0_CLIPModel/model.safetensors filter=lfs diff=lfs merge=lfs -text

0_CLIPModel/config.json

@@ -14,7 +14,7 @@
     "model_type": "clip_text_model"
   },
   "torch_dtype": "float32",
-  "transformers_version": "4.35.0.dev0",
+  "transformers_version": "4.36.0.dev0",
   "vision_config": {
     "dropout": 0.0,
     "model_type": "clip_vision_model"

0_CLIPModel/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83bcf2d8c66a985c30c31f1887a1c7fdf0637e4ad4d8888d5ef0d8323c21b126
+size 605156676

0_CLIPModel/preprocessor_config.json

@@ -16,4 +16,4 @@
   ],
   "resample": 3,
   "size": 224
-}
+}

README.txt

@@ -0,0 +1,67 @@
+python -W ignore finetune-clip-huggingface/huggingface_finetune_clip.py     --output_dir /home/ekansa/github/archaeology-images-ai/results     --model_name_or_path openai/clip-vit-base-patch32     --train_file /home/ekansa/github/archaeology-images-ai/files/train.json     --validation_file /home/ekansa/github/archaeology-images-ai/files/test.json     --image_column image     --overwrite_output_dir=True     --max_seq_length=77     --num_train_epochs=20     --caption_column caption
+ --overwrite_cache=True     --remove_unused_columns=False     --do_train     --per_device_train_batch_size=64     --per_device_eval_batch_size=64     --learning_rate="5e-5" --warmup_steps="2" --weight_decay 0.2
+11/13/2023 19:25:55 - WARNING - __main__ - Process rank: 0, device: cuda:0, n_gpu: 1distributed training: True, 16-bits training: False
+Running tokenizer on train dataset: 100%|███████████████████████████████████| 39755/39755 [00:01<00:00, 22415.96 examples/s]
+Parameter 'transform'=<function main.<locals>.transform_images at 0x7fded1ed0cc0> of the transform datasets.arrow_dataset.Dataset.set_format couldn't be hashed properly, a random hash was used instead. Make sure your transforms and parameters are serializable with pickle or dill for the dataset fingerprinting and caching to work. If you reuse this transform, the caching mechanism will consider it to be different from the previous calls and recompute everything. This warning is only showed once. Subsequent hashing failures won't be showed.
+11/13/2023 19:26:00 - WARNING - datasets.fingerprint - Parameter 'transform'=<function main.<locals>.transform_images at 0x7fded1ed0cc0> of the transform datasets.arrow_dataset.Dataset.set_format couldn't be hashed properly, a random hash was used instead. Make sure your transforms and parameters are serializable with pickle or dill for the dataset fingerprinting and caching to work. If you reuse this transform, the caching mechanism will consider it to be different from the previous calls and recompute everything. This warning is only showed once. Subsequent hashing failures won't be showed.
+{'loss': 2.1774, 'learning_rate': 4.799807042932948e-05, 'epoch': 0.8}
+{'loss': 1.553, 'learning_rate': 4.5988100980865096e-05, 'epoch': 1.61}
+{'loss': 1.2706, 'learning_rate': 4.3978131532400705e-05, 'epoch': 2.41}
+{'loss': 1.0318, 'learning_rate': 4.196816208393632e-05, 'epoch': 3.22}
+{'loss': 0.8534, 'learning_rate': 3.9958192635471945e-05, 'epoch': 4.02}
+{'loss': 0.673, 'learning_rate': 3.794822318700756e-05, 'epoch': 4.82}
+{'loss': 0.564, 'learning_rate': 3.593825373854318e-05, 'epoch': 5.63}
+{'loss': 0.496, 'learning_rate': 3.3928284290078794e-05, 'epoch': 6.43}
+{'loss': 0.4287, 'learning_rate': 3.191831484161441e-05, 'epoch': 7.23}
+{'loss': 0.3796, 'learning_rate': 2.9908345393150027e-05, 'epoch': 8.04}
+{'loss': 0.3378, 'learning_rate': 2.789837594468564e-05, 'epoch': 8.84}
+{'loss': 0.3009, 'learning_rate': 2.5888406496221256e-05, 'epoch': 9.65}
+{'loss': 0.2707, 'learning_rate': 2.3878437047756876e-05, 'epoch': 10.45}
+{'loss': 0.2552, 'learning_rate': 2.1868467599292492e-05, 'epoch': 11.25}
+{'loss': 0.2293, 'learning_rate': 1.985849815082811e-05, 'epoch': 12.06}
+{'loss': 0.212, 'learning_rate': 1.7848528702363725e-05, 'epoch': 12.86}
+{'loss': 0.1879, 'learning_rate': 1.583855925389934e-05, 'epoch': 13.67}
+{'loss': 0.1782, 'learning_rate': 1.3828589805434958e-05, 'epoch': 14.47}
+{'loss': 0.1726, 'learning_rate': 1.1818620356970576e-05, 'epoch': 15.27}
+{'loss': 0.153, 'learning_rate': 9.80865090850619e-06, 'epoch': 16.08}
+{'loss': 0.1456, 'learning_rate': 7.798681460041808e-06, 'epoch': 16.88}
+{'loss': 0.1397, 'learning_rate': 5.788712011577424e-06, 'epoch': 17.68}
+{'loss': 0.1326, 'learning_rate': 3.7787425631130406e-06, 'epoch': 18.49}
+{'loss': 0.1228, 'learning_rate': 1.7687731146486576e-06, 'epoch': 19.29}
+{'train_runtime': 40077.3502, 'train_samples_per_second': 19.839, 'train_steps_per_second': 0.31, 'train_loss': 0.4972445463444259, 'epoch': 20.0}
+100%|██████████████████████████████████████████████████████████████████████████████| 12440/12440 [11:07:57<00:00,  3.22s/it]
+***** train metrics *****
+  epoch                    =        20.0
+  train_loss               =      0.4972
+  train_runtime            = 11:07:57.35
+  train_samples_per_second =      19.839
+  train_steps_per_second   =        0.31
+
+
+Then restarted at checkpoint:
+06:34 $ python -W ignore finetune-clip-huggingface/huggingface_finetune_clip.py     --output_dir /home/ekansa/github/archaeology-images-ai/results     --model_name_or_path openai/clip-vit-base-patch32     --train_file /home/ekansa/github/archaeology-images-ai/files/train.json     --validation_file /home/ekansa/github/archaeology-images-ai/files/test.json     --image_column image     --overwrite_output_dir=False     --max_seq_length=77     --num_train_epochs=30     --caption_column caption     --overwrite_cache=True     --remove_unused_columns=False     --do_train     --per_device_train_batch_size=64     --per_device_eval_batch_size=64     --learning_rate="5e-5" --warmup_steps="2" --weight_decay 0.2
+11/14/2023 08:43:58 - WARNING - __main__ - Process rank: 0, device: cuda:0, n_gpu: 1distributed training: True, 16-bits training: False
+Running tokenizer on train dataset: 100%|███████████████████████████████████| 39755/39755 [00:02<00:00, 13727.10 examples/s]
+Parameter 'transform'=<function main.<locals>.transform_images at 0x7f2655088cc0> of the transform datasets.arrow_dataset.Dataset.set_format couldn't be hashed properly, a random hash was used instead. Make sure your transforms and parameters are serializable with pickle or dill for the dataset fingerprinting and caching to work. If you reuse this transform, the caching mechanism will consider it to be different from the previous calls and recompute everything. This warning is only showed once. Subsequent hashing failures won't be showed.
+11/14/2023 08:44:08 - WARNING - datasets.fingerprint - Parameter 'transform'=<function main.<locals>.transform_images at 0x7f2655088cc0> of the transform datasets.arrow_dataset.Dataset.set_format couldn't be hashed properly, a random hash was used instead. Make sure your transforms and parameters are serializable with pickle or dill for the dataset fingerprinting and caching to work. If you reuse this transform, the caching mechanism will consider it to be different from the previous calls and recompute everything. This warning is only showed once. Subsequent hashing failures won't be showed.
+{'loss': 0.1433, 'learning_rate': 1.650766427269804e-05, 'epoch': 20.1}
+{'loss': 0.1539, 'learning_rate': 1.5167756458355668e-05, 'epoch': 20.9}
+{'loss': 0.1532, 'learning_rate': 1.3827848644013291e-05, 'epoch': 21.7}
+{'loss': 0.147, 'learning_rate': 1.2487940829670919e-05, 'epoch': 22.51}
+{'loss': 0.1423, 'learning_rate': 1.1148033015328546e-05, 'epoch': 23.31}
+{'loss': 0.1334, 'learning_rate': 9.808125200986172e-06, 'epoch': 24.12}
+{'loss': 0.1329, 'learning_rate': 8.468217386643799e-06, 'epoch': 24.92}
+{'loss': 0.1228, 'learning_rate': 7.1283095723014256e-06, 'epoch': 25.72}
+{'loss': 0.1234, 'learning_rate': 5.788401757959053e-06, 'epoch': 26.53}
+{'loss': 0.1166, 'learning_rate': 4.448493943616679e-06, 'epoch': 27.33}
+{'loss': 0.1131, 'learning_rate': 3.108586129274306e-06, 'epoch': 28.14}
+{'loss': 0.1118, 'learning_rate': 1.7686783149319325e-06, 'epoch': 28.94}
+{'loss': 0.11, 'learning_rate': 4.2877050058955945e-07, 'epoch': 29.74}
+{'train_runtime': 31058.1839, 'train_samples_per_second': 38.401, 'train_steps_per_second': 0.601, 'train_loss': 0.046548270668119736, 'epoch': 30.0}
+100%|███████████████████████████████████████████████████████████████████████████████| 18660/18660 [8:37:38<00:00,  1.66s/it]
+***** train metrics *****
+  epoch                    =       30.0
+  train_loss               =     0.0465
+  train_runtime            = 8:37:38.18
+  train_samples_per_second =     38.401
+  train_steps_per_second   =      0.601