Instructions to use UCSC-VLAA/ViT-bigG-14-CLIPA-336-datacomp1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- OpenCLIP
How to use UCSC-VLAA/ViT-bigG-14-CLIPA-336-datacomp1B with OpenCLIP:
import open_clip model, preprocess_train, preprocess_val = open_clip.create_model_and_transforms('hf-hub:UCSC-VLAA/ViT-bigG-14-CLIPA-336-datacomp1B') tokenizer = open_clip.get_tokenizer('hf-hub:UCSC-VLAA/ViT-bigG-14-CLIPA-336-datacomp1B') - Notebooks
- Google Colab
- Kaggle
Update open_clip_config.json
Browse files- open_clip_config.json +5 -1
open_clip_config.json
CHANGED
|
@@ -9,12 +9,16 @@
|
|
| 9 |
"mlp_ratio": 4.9231,
|
| 10 |
"patch_size": 14,
|
| 11 |
"no_ln_pre": true,
|
| 12 |
-
"pool_type": "avg"
|
|
|
|
| 13 |
},
|
| 14 |
"text_cfg": {
|
| 15 |
"context_length": 32,
|
| 16 |
"vocab_size": 32000,
|
| 17 |
"hf_tokenizer_name": "bert-base-uncased",
|
|
|
|
|
|
|
|
|
|
| 18 |
"width": 1280,
|
| 19 |
"heads": 20,
|
| 20 |
"layers": 32,
|
|
|
|
| 9 |
"mlp_ratio": 4.9231,
|
| 10 |
"patch_size": 14,
|
| 11 |
"no_ln_pre": true,
|
| 12 |
+
"pool_type": "avg",
|
| 13 |
+
"final_ln_after_pool": true
|
| 14 |
},
|
| 15 |
"text_cfg": {
|
| 16 |
"context_length": 32,
|
| 17 |
"vocab_size": 32000,
|
| 18 |
"hf_tokenizer_name": "bert-base-uncased",
|
| 19 |
+
"tokenizer_kwargs": {
|
| 20 |
+
"strip_sep_token": true
|
| 21 |
+
},
|
| 22 |
"width": 1280,
|
| 23 |
"heads": 20,
|
| 24 |
"layers": 32,
|