open_clip

src/open_clip/model_configs/ViT-bigG-14.json

+{
+    "embed_dim": 1280,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 48,
+        "width": 1664,
+        "head_width": 104,
+        "mlp_ratio": 4.9231,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1280,
+        "heads": 20,
+        "layers": 32
+    }
+}
\ No newline at end of file

src/open_clip/model_configs/ViT-e-14.json

+{
+    "embed_dim": 1280,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 56,
+        "width": 1792,
+        "head_width": 112,
+        "mlp_ratio": 8.5715,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1280,
+        "heads": 20,
+        "layers": 36
+    }
+}
\ No newline at end of file

src/open_clip/model_configs/ViT-g-14.json

+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 40,
+        "width": 1408,
+        "head_width": 88,
+        "mlp_ratio": 4.3637,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 24
+    }
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-B-LTT.json

+{
+    "embed_dim": 768,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_base_224",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 224
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 768,
+      "heads": 12,
+      "layers": 12
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-B.json

+{
+    "embed_dim": 512,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_base_224",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 224
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 512,
+      "heads": 8,
+      "layers": 12
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-L-256.json

+{
+    "embed_dim": 768,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_large_256",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 256
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 768,
+      "heads": 12,
+      "layers": 12
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-L-336.json

+{
+    "embed_dim": 768,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_large_336",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 336
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 768,
+      "heads": 12,
+      "layers": 12
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-L.json

+{
+    "embed_dim": 768,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_large_224",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 224
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 768,
+      "heads": 12,
+      "layers": 12
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-L2-256.json

+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_large2_256",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 256
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 1024,
+      "heads": 16,
+      "layers": 24
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-L2-336.json

+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_large2_336",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 336
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 1024,
+      "heads": 16,
+      "layers": 24
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-L2.json

+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_large2_224",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 224
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 1024,
+      "heads": 16,
+      "layers": 24
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-S-LTT.json

+{
+    "embed_dim": 768,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_small_224",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 224
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 768,
+      "heads": 12,
+      "layers": 12
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-S.json

+{
+    "embed_dim": 384,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_small_224",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 224
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 384,
+      "heads": 6,
+      "layers": 12
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-XL-256.json

+{
+    "embed_dim": 1152,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_xlarge_256",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 256
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 1152,
+      "heads": 16,
+      "layers": 27
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-XL-336.json

+{
+    "embed_dim": 1152,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_xlarge_336",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 336
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 1152,
+      "heads": 16,
+      "layers": 27
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/ViTamin-XL-384.json

+{
+    "embed_dim": 1152,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_xlarge_384",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 256
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 1152,
+      "heads": 16,
+      "layers": 27
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/coca_ViT-B-32.json

+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 768,
+        "patch_size": 32,
+        "attentional_pool": true,
+        "attn_pooler_heads": 8,
+        "output_tokens": true
+    },
+    "text_cfg": {
+        "context_length": 76,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12,
+        "embed_cls": true,
+        "output_tokens": true
+    },
+    "multimodal_cfg": {
+        "context_length": 76,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12,
+        "attn_pooler_heads": 8
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/coca_ViT-L-14.json

+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 24,
+        "width": 1024,
+        "patch_size": 14,
+        "attentional_pool": true,
+        "attn_pooler_heads": 8,
+        "output_tokens": true
+    },
+    "text_cfg": {
+        "context_length": 76,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "embed_cls": true,
+        "output_tokens": true
+    },
+    "multimodal_cfg": {
+        "context_length": 76,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "attn_pooler_heads": 12
+    },
+    "custom_text": true
+}

src/open_clip/model_configs/coca_base.json

+{
+    "embed_dim": 512,
+    "multimodal_cfg": {
+        "width": 768,
+        "context_length": 76,
+        "vocab_size": 64000,
+        "mlp_ratio": 4,
+        "layers": 12,
+        "dim_head": 64,
+        "heads": 12,
+        "n_queries": 256,
+        "attn_pooler_heads": 8
+    },
+    "vision_cfg": {
+        "image_size": 288,
+        "layers": 12,
+        "width": 768,
+        "patch_size": 18,
+        "output_tokens": true
+    },
+    "text_cfg": {
+        "context_length": 76,
+        "vocab_size": 64000,
+        "layers": 12,
+        "heads": 12,
+        "width": 768,
+        "embed_cls": true,
+        "output_tokens": true
+    },
+    "custom_text": true
+}
\ No newline at end of file

src/open_clip/model_configs/coca_roberta-ViT-B-32.json

+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 768,
+        "patch_size": 32,
+        "output_tokens": true
+    },
+    "text_cfg": {
+        "hf_model_name": "roberta-base",
+        "hf_tokenizer_name": "roberta-base",
+        "hf_proj_type": "linear",
+        "width": 768,
+        "output_tokens": true
+    },
+    "multimodal_cfg": {
+        "context_length": 76,
+        "width": 768,
+        "heads": 8,
+        "layers": 12
+    },
+    "custom_text": true
+}