Added backgrount to README
Browse files
README.md
CHANGED
@@ -1,3 +1,88 @@
|
|
1 |
---
|
2 |
license: gpl-3.0
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: gpl-3.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
tags:
|
6 |
+
- archaeology
|
7 |
---
|
8 |
+
|
9 |
+
## ABOUT ARCHAEO-CLIP
|
10 |
+
|
11 |
+
|
12 |
+
This model is the result of "fine-tuning" the openai/clip-vit-base-patch32 model using captioned images of archaeological artifacts published by Open Context. This model is the latest of several iterations in experiments to improve the captions, debug the training pipeline, and try different fine-tuning parameters. It seems a model with a relatively low training rate helps add some "archaeological knowledge" while still retain much of the general knowledge of out-of-the-box CLIP.
|
13 |
+
|
14 |
+
We'll use this fine-tuned model in the future to do more experiments, including further fine-tuning with captioned images from open access museum collections of archaeological materials. Below we itemize the specific training parameters used in the fine tuning of this model
|
15 |
+
|
16 |
+
The training (45,256 captioned archaeological artifact images from Open Context) and test data, as well as the specific Python code used to run the fine-tuning run can be found here:
|
17 |
+
|
18 |
+
https://github.com/opencontext/archaeology-images-ai
|
19 |
+
|
20 |
+
|
21 |
+
|
22 |
+
|
23 |
+
|
24 |
+
|
25 |
+
|
26 |
+
Fine Tuning
|
27 |
+
-----------
|
28 |
+
Below is the invocation and specific parameters I used for fine tuning:
|
29 |
+
|
30 |
+
```bash
|
31 |
+
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_path" --overwrite_output_dir=True --max_seq_length=77 --num_train_epochs=25 --caption_column="caption" --overwrite_cache=True --remove_unused_columns=False --do_train=True --per_device_train_batch_size=64 --per_device_eval_batch_size=64 --learning_rate="2e-5" --warmup_steps="2" --weight_decay 0.2
|
32 |
+
|
33 |
+
12/10/2023 21:35:43 - WARNING - __main__ - Process rank: 0, device: cuda:0, n_gpu: 1distributed training: True, 16-bits training: False
|
34 |
+
Running tokenizer on train dataset: 100%|βββββββββββββββββββββββββββββββ| 45256/45256 [00:02<00:00, 21481.25 examples/s]Parameter 'transform'=<function main.<locals>.transform_images at 0x7fe53504d9e0> 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.
|
35 |
+
12/10/2023 21:35:47 - WARNING - datasets.fingerprint - Parameter 'transform'=<function main.<locals>.transform_images at 0x7fe53504d9e0> 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.
|
36 |
+
|
37 |
+
|
38 |
+
{'loss': 1.7174, 'learning_rate': 1.9437224545146346e-05, 'epoch': 0.71}
|
39 |
+
{'loss': 1.1706, 'learning_rate': 1.887218894790372e-05, 'epoch': 1.41}
|
40 |
+
{'loss': 0.9596, 'learning_rate': 1.8307153350661094e-05, 'epoch': 2.12}
|
41 |
+
{'loss': 0.7291, 'learning_rate': 1.7742117753418467e-05, 'epoch': 2.82}
|
42 |
+
{'loss': 0.5833, 'learning_rate': 1.717708215617584e-05, 'epoch': 3.53}
|
43 |
+
{'loss': 0.5094, 'learning_rate': 1.6612046558933215e-05, 'epoch': 4.24}
|
44 |
+
{'loss': 0.4368, 'learning_rate': 1.6047010961690588e-05, 'epoch': 4.94}
|
45 |
+
{'loss': 0.365, 'learning_rate': 1.548197536444796e-05, 'epoch': 5.65}
|
46 |
+
{'loss': 0.3394, 'learning_rate': 1.4916939767205336e-05, 'epoch': 6.36}
|
47 |
+
{'loss': 0.3159, 'learning_rate': 1.4351904169962709e-05, 'epoch': 7.06}
|
48 |
+
{'loss': 0.2776, 'learning_rate': 1.3786868572720083e-05, 'epoch': 7.77}
|
49 |
+
{'loss': 0.2584, 'learning_rate': 1.3221832975477456e-05, 'epoch': 8.47}
|
50 |
+
{'loss': 0.2464, 'learning_rate': 1.2656797378234832e-05, 'epoch': 9.18}
|
51 |
+
{'loss': 0.227, 'learning_rate': 1.2091761780992204e-05, 'epoch': 9.89}
|
52 |
+
{'loss': 0.2116, 'learning_rate': 1.1526726183749577e-05, 'epoch': 10.59}
|
53 |
+
{'loss': 0.2026, 'learning_rate': 1.0961690586506951e-05, 'epoch': 11.3}
|
54 |
+
{'loss': 0.1869, 'learning_rate': 1.0396654989264325e-05, 'epoch': 12.01}
|
55 |
+
{'loss': 0.1792, 'learning_rate': 9.831619392021698e-06, 'epoch': 12.71}
|
56 |
+
{'loss': 0.167, 'learning_rate': 9.266583794779072e-06, 'epoch': 13.42}
|
57 |
+
{'loss': 0.1671, 'learning_rate': 8.701548197536446e-06, 'epoch': 14.12}
|
58 |
+
{'loss': 0.154, 'learning_rate': 8.136512600293819e-06, 'epoch': 14.83}
|
59 |
+
{'loss': 0.1574, 'learning_rate': 7.571477003051193e-06, 'epoch': 15.54}
|
60 |
+
{'loss': 0.1496, 'learning_rate': 7.006441405808566e-06, 'epoch': 16.24}
|
61 |
+
{'loss': 0.1329, 'learning_rate': 5.876370211323313e-06, 'epoch': 17.66}
|
62 |
+
{'loss': 0.1316, 'learning_rate': 5.311334614080687e-06, 'epoch': 18.36}
|
63 |
+
{'loss': 0.1254, 'learning_rate': 4.746299016838062e-06, 'epoch': 19.07}
|
64 |
+
{'loss': 0.1266, 'learning_rate': 4.181263419595435e-06, 'epoch': 19.77}
|
65 |
+
{'loss': 0.1193, 'learning_rate': 3.6162278223528084e-06, 'epoch': 20.48}
|
66 |
+
{'loss': 0.1163, 'learning_rate': 3.0511922251101822e-06, 'epoch': 21.19}
|
67 |
+
{'loss': 0.1154, 'learning_rate': 2.486156627867556e-06, 'epoch': 21.89}
|
68 |
+
{'loss': 0.1125, 'learning_rate': 1.9211210306249294e-06, 'epoch': 22.6}
|
69 |
+
{'loss': 0.1063, 'learning_rate': 1.356085433382303e-06, 'epoch': 23.31}
|
70 |
+
{'loss': 0.1082, 'learning_rate': 7.91049836139677e-07, 'epoch': 24.01}
|
71 |
+
{'loss': 0.1032, 'learning_rate': 2.2601423889705053e-07, 'epoch': 24.72}
|
72 |
+
|
73 |
+
{'train_runtime': 78442.5601, 'train_samples_per_second': 14.423, 'train_steps_per_second': 0.226, 'train_loss': 0.31630637788503185, 'epoch': 25.0}
|
74 |
+
100%|ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 17700/17700 [21:47:22<00:00, 4.43s/it]
|
75 |
+
***** train metrics *****
|
76 |
+
epoch = 25.0
|
77 |
+
train_loss = 0.3163
|
78 |
+
train_runtime = 21:47:22.56
|
79 |
+
train_samples_per_second = 14.423
|
80 |
+
train_steps_per_second = 0.226
|
81 |
+
```
|
82 |
+
|
83 |
+
|
84 |
+
|
85 |
+
|
86 |
+
---
|
87 |
+
license: gpl-3.0
|
88 |
+
---
|