diff --git a/src/open_clip/model_configs/ViT-bigG-14-CLIPA.json b/src/open_clip/model_configs/ViT-bigG-14-CLIPA.json new file mode 100644 index 000000000..83ec709f8 --- /dev/null +++ b/src/open_clip/model_configs/ViT-bigG-14-CLIPA.json @@ -0,0 +1,27 @@ +{ + "embed_dim": 1280, + "vision_cfg": { + "image_size": 224, + "layers": 48, + "width": 1664, + "head_width": 104, + "mlp_ratio": 4.9231, + "patch_size": 14, + "no_ln_pre": true, + "pool_type": "avg", + "final_ln_after_pool": true + }, + "text_cfg": { + "context_length": 32, + "vocab_size": 32000, + "hf_tokenizer_name": "bert-base-uncased", + "tokenizer_kwargs": { + "strip_sep_token": true + }, + "width": 1280, + "heads": 20, + "layers": 32, + "pool_type": "last", + "no_causal_mask": true + } +} \ No newline at end of file diff --git a/src/open_clip/pretrained.py b/src/open_clip/pretrained.py index 2454f5797..d1e8b58af 100644 --- a/src/open_clip/pretrained.py +++ b/src/open_clip/pretrained.py @@ -395,9 +395,9 @@ def _apcfg(url='', hf_hub='', **kwargs): laion2b=_apcfg(hf_hub='rwightman/ViT-H-14-CLIPA-336-laion2B/'), datacomp1b=_apcfg(hf_hub='rwightman/ViT-H-14-CLIPA-336-datacomp1B/'), ), - # "ViT-bigG-14-CLIPA": dict( - # datacomp1b=_apcfg(hf_hub='rwightman/ViT-bigG-14-CLIPA-datacomp1B/'), - # ), + "ViT-bigG-14-CLIPA": dict( + datacomp1b=_apcfg(hf_hub='rwightman/ViT-bigG-14-CLIPA-datacomp1B/'), + ), "ViT-bigG-14-CLIPA-336": dict( datacomp1b=_apcfg(hf_hub='rwightman/ViT-bigG-14-CLIPA-336-datacomp1B/'), ),