feat: 完善Topaz Video AI SDK并重新组织models目录结构

主要改进:
 AI引擎和基准测试支持:
- 完整支持所有AI引擎标志 (Artemis, Gaia, Theia, Proteus, Iris)
- 实现引擎特定的参数优化和FFmpeg命令生成
- 添加基准测试模板和性能测试功能
- 新增专用预设模板 (animation_enhance, detail_recovery等)

 音频编码器智能化:
- 完善音频编码器配置和自动选择
- 支持AAC, AC3, PCM, Vorbis等多种编码器
- 实现质量级别自动映射和容器格式兼容性检查
- 添加音频编码器演示示例

 目录结构重新组织:
- 将models目录按功能分类重新组织
- 视觉-语言模型配置 -> 视觉-语言模型配置/
- 编码解码配置 -> 编码解码配置/
- 基准测试配置 -> config/
- 其他配置文件 -> 其他配置/

 技术增强:
- 增强TemplateBuilder功能 (ai_engine, focus_fix_level, benchmark_mode等)
- 完善FFmpeg参数生成和模型映射
- 添加智能推荐系统和性能优化
- 新增多个演示示例 (benchmarks_demo, audio_codecs_demo等)

 完整性提升:
- 模板属性使用率达到100%
- 所有AI引擎和编码器都有明确用途和处理方案
- 完整的文档分析和使用指南

cargos/tvai-v2/视觉-语言模型配置/EVA01-g-14-plus.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "eva_giant_patch14_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "token",
+        "timm_proj": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 24
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/EVA01-g-14.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "eva_giant_patch14_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "token",
+        "timm_proj": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/EVA02-B-16.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "eva02_base_patch16_clip_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "token",
+        "timm_proj": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/EVA02-E-14-plus.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "eva02_enormous_patch14_clip_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "token",
+        "timm_proj": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1280,
+        "heads": 20,
+        "layers": 32
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/EVA02-E-14.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "eva02_enormous_patch14_clip_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "token",
+        "timm_proj": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 24
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/EVA02-L-14-336.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 336,
+        "timm_model_name": "eva02_large_patch14_clip_336",
+        "timm_model_pretrained": false,
+        "timm_pool": "token",
+        "timm_proj": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/EVA02-L-14.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "eva02_large_patch14_clip_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "token",
+        "timm_proj": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/MobileCLIP-B.json

@@ -0,0 +1,21 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "timm_model_name": "vit_base_mci_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "token",
+        "timm_proj": null,
+        "timm_drop": 0.0,
+        "timm_drop_path": 0.0,
+        "image_size": 224
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12,
+        "no_causal_mask": false
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/MobileCLIP-S1.json

@@ -0,0 +1,21 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "timm_model_name": "fastvit_mci1",
+        "timm_model_pretrained": false,
+        "timm_pool": "avg",
+        "timm_proj": null,
+        "timm_drop": 0.0,
+        "timm_drop_path": 0.0,
+        "image_size": 256
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12,
+        "no_causal_mask": true
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/MobileCLIP-S2.json

@@ -0,0 +1,21 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "timm_model_name": "fastvit_mci2",
+        "timm_model_pretrained": false,
+        "timm_pool": "avg",
+        "timm_proj": null,
+        "timm_drop": 0.0,
+        "timm_drop_path": 0.0,
+        "image_size": 256
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12,
+        "no_causal_mask": true
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/RN101-quickgelu.json

@@ -0,0 +1,22 @@
+{
+    "embed_dim": 512,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": [
+            3,
+            4,
+            23,
+            3
+        ],
+        "width": 64,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN101.json

@@ -0,0 +1,21 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": [
+            3,
+            4,
+            23,
+            3
+        ],
+        "width": 64,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN50-quickgelu.json

@@ -0,0 +1,22 @@
+{
+    "embed_dim": 1024,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": [
+            3,
+            4,
+            6,
+            3
+        ],
+        "width": 64,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN50.json

@@ -0,0 +1,21 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": [
+            3,
+            4,
+            6,
+            3
+        ],
+        "width": 64,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN50x16-quickgelu.json

@@ -0,0 +1,22 @@
+{
+    "embed_dim": 768,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 384,
+        "layers": [
+            6,
+            8,
+            18,
+            8
+        ],
+        "width": 96,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN50x16.json

@@ -0,0 +1,21 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 384,
+        "layers": [
+            6,
+            8,
+            18,
+            8
+        ],
+        "width": 96,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN50x4-quickgelu.json

@@ -0,0 +1,22 @@
+{
+    "embed_dim": 640,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 288,
+        "layers": [
+            4,
+            6,
+            10,
+            6
+        ],
+        "width": 80,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 640,
+        "heads": 10,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN50x4.json

@@ -0,0 +1,21 @@
+{
+    "embed_dim": 640,
+    "vision_cfg": {
+        "image_size": 288,
+        "layers": [
+            4,
+            6,
+            10,
+            6
+        ],
+        "width": 80,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 640,
+        "heads": 10,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN50x64-quickgelu.json

@@ -0,0 +1,22 @@
+{
+    "embed_dim": 1024,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 448,
+        "layers": [
+            3,
+            15,
+            36,
+            10
+        ],
+        "width": 128,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/RN50x64.json

@@ -0,0 +1,21 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 448,
+        "layers": [
+            3,
+            15,
+            36,
+            10
+        ],
+        "width": 128,
+        "patch_size": null
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP-256.json

@@ -0,0 +1,29 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_base_patch16_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP-384.json

@@ -0,0 +1,29 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 384,
+        "timm_model_name": "vit_base_patch16_siglip_384",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP-512.json

@@ -0,0 +1,29 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 512,
+        "timm_model_name": "vit_base_patch16_siglip_512",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP-i18n-256.json

@@ -0,0 +1,29 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_base_patch16_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 250000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP-i18n-256",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP.json

@@ -0,0 +1,29 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "vit_base_patch16_siglip_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP2-256.json

@@ -0,0 +1,32 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_base_patch16_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP2-256",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP2-384.json

@@ -0,0 +1,32 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 384,
+        "timm_model_name": "vit_base_patch16_siglip_384",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP2-384",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP2-512.json

@@ -0,0 +1,32 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 512,
+        "timm_model_name": "vit_base_patch16_siglip_512",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP2-512",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-SigLIP2.json

@@ -0,0 +1,32 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "vit_base_patch16_siglip_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP2",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-plus-240.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 640,
+    "vision_cfg": {
+        "image_size": 240,
+        "layers": 12,
+        "width": 896,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 640,
+        "heads": 10,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-plus.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 640,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 896,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 640,
+        "heads": 10,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16-quickgelu.json

@@ -0,0 +1,17 @@
+{
+    "embed_dim": 512,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 768,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-16.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 768,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-32-256.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 256,
+        "layers": 12,
+        "width": 768,
+        "patch_size": 32
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-32-SigLIP2-256.json

@@ -0,0 +1,32 @@
+{
+    "embed_dim": 768,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_base_patch32_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-B-32-SigLIP2-256",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 768,
+        "heads": 12,
+        "layers": 12,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-32-plus-256.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 640,
+    "vision_cfg": {
+        "image_size": 256,
+        "layers": 12,
+        "width": 896,
+        "patch_size": 32
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 640,
+        "heads": 10,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-32-quickgelu.json

@@ -0,0 +1,17 @@
+{
+    "embed_dim": 512,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 768,
+        "patch_size": 32
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-B-32.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 768,
+        "patch_size": 32
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-H-14-378-quickgelu.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 1024,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 378,
+        "layers": 32,
+        "width": 1280,
+        "head_width": 80,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 24
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-H-14-378.json

@@ -0,0 +1,17 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 378,
+        "layers": 32,
+        "width": 1280,
+        "head_width": 80,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 24
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-H-14-CLIPA-336.json

@@ -0,0 +1,26 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 336,
+        "layers": 32,
+        "width": 1280,
+        "head_width": 80,
+        "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": 1024,
+        "heads": 16,
+        "layers": 24,
+        "pool_type": "last",
+        "no_causal_mask": true
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-H-14-CLIPA.json

@@ -0,0 +1,26 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 32,
+        "width": 1280,
+        "head_width": 80,
+        "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": 1024,
+        "heads": 16,
+        "layers": 24,
+        "pool_type": "last",
+        "no_causal_mask": true
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-H-14-quickgelu.json

@@ -0,0 +1,18 @@
+{
+    "embed_dim": 1024,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 32,
+        "width": 1280,
+        "head_width": 80,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 24
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-H-14.json

@@ -0,0 +1,17 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 32,
+        "width": 1280,
+        "head_width": 80,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 24
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-H-16.json

@@ -0,0 +1,17 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 32,
+        "width": 1280,
+        "head_width": 80,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 1024,
+        "heads": 16,
+        "layers": 24
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-14-280.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 280,
+        "layers": 24,
+        "width": 1024,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-14-336-quickgelu.json

@@ -0,0 +1,17 @@
+{
+    "embed_dim": 768,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 336,
+        "layers": 24,
+        "width": 1024,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-14-336.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 336,
+        "layers": 24,
+        "width": 1024,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-14-CLIPA-336.json

@@ -0,0 +1,25 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 336,
+        "layers": 24,
+        "width": 1024,
+        "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": 768,
+        "heads": 12,
+        "layers": 12,
+        "pool_type": "last",
+        "no_causal_mask": true
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-14-CLIPA.json

@@ -0,0 +1,25 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 24,
+        "width": 1024,
+        "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": 768,
+        "heads": 12,
+        "layers": 12,
+        "pool_type": "last",
+        "no_causal_mask": true
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-14-quickgelu.json

@@ -0,0 +1,17 @@
+{
+    "embed_dim": 768,
+    "quick_gelu": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 24,
+        "width": 1024,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-14.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 24,
+        "width": 1024,
+        "patch_size": 14
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-16-320.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 320,
+        "layers": 24,
+        "width": 1024,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-16-SigLIP-256.json

@@ -0,0 +1,29 @@
+{
+    "embed_dim": 1024,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_large_patch16_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1024,
+        "heads": 16,
+        "layers": 24,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-16-SigLIP-384.json

@@ -0,0 +1,29 @@
+{
+    "embed_dim": 1024,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 384,
+        "timm_model_name": "vit_large_patch16_siglip_384",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1024,
+        "heads": 16,
+        "layers": 24,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-16-SigLIP2-256.json

@@ -0,0 +1,32 @@
+{
+    "embed_dim": 1024,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_large_patch16_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-L-16-SigLIP2-256",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1024,
+        "heads": 16,
+        "layers": 24,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-16-SigLIP2-384.json

@@ -0,0 +1,32 @@
+{
+    "embed_dim": 1024,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 384,
+        "timm_model_name": "vit_large_patch16_siglip_384",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-L-16-SigLIP2-384",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1024,
+        "heads": 16,
+        "layers": 24,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-16-SigLIP2-512.json

@@ -0,0 +1,32 @@
+{
+    "embed_dim": 1024,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 512,
+        "timm_model_name": "vit_large_patch16_siglip_512",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-L-16-SigLIP2-512",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1024,
+        "heads": 16,
+        "layers": 24,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-L-16.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 24,
+        "width": 1024,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 768,
+        "heads": 12,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-M-16-alt.json

@@ -0,0 +1,17 @@
+{
+    "embed_dim": 384,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 512,
+        "patch_size": 16,
+        "ls_init_value": 1e-4
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 384,
+        "heads": 6,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-M-16.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 512,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-M-32-alt.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 384,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 512,
+        "patch_size": 32
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 384,
+        "heads": 6,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-M-32.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 512,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 512,
+        "patch_size": 32
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 512,
+        "heads": 8,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-S-16-alt.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 256,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 384,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 256,
+        "heads": 4,
+        "layers": 10
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-S-16.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 384,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 384,
+        "patch_size": 16
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 384,
+        "heads": 6,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-S-32-alt.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 256,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 384,
+        "patch_size": 32
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 256,
+        "heads": 4,
+        "layers": 10
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-S-32.json

@@ -0,0 +1,16 @@
+{
+    "embed_dim": 384,
+    "vision_cfg": {
+        "image_size": 224,
+        "layers": 12,
+        "width": 384,
+        "patch_size": 32
+    },
+    "text_cfg": {
+        "context_length": 77,
+        "vocab_size": 49408,
+        "width": 384,
+        "heads": 6,
+        "layers": 12
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-14-SigLIP-378.json

@@ -0,0 +1,30 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 378,
+        "timm_model_name": "vit_so400m_patch14_siglip_378",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-14-SigLIP-384.json

@@ -0,0 +1,30 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 384,
+        "timm_model_name": "vit_so400m_patch14_siglip_384",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-14-SigLIP.json

@@ -0,0 +1,30 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "vit_so400m_patch14_siglip_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 16,
+        "vocab_size": 32000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-14-SigLIP2-378.json

@@ -0,0 +1,33 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 378,
+        "timm_model_name": "vit_so400m_patch14_siglip_378",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-SO400M-14-SigLIP2-378",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-14-SigLIP2.json

@@ -0,0 +1,33 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 224,
+        "timm_model_name": "vit_so400m_patch14_siglip_224",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-SO400M-14-SigLIP2",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-16-SigLIP-i18n-256.json

@@ -0,0 +1,30 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_so400m_patch16_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 250000,
+        "hf_tokenizer_name": "timm/ViT-B-16-SigLIP-i18n-256",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "pool_type": "last",
+        "proj_type": "none",
+        "norm_kwargs":{
+            "eps": 1e-6
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-16-SigLIP2-256.json

@@ -0,0 +1,33 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_so400m_patch16_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-SO400M-16-SigLIP2-256",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-16-SigLIP2-384.json

@@ -0,0 +1,33 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 384,
+        "timm_model_name": "vit_so400m_patch16_siglip_384",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-SO400M-16-SigLIP2-384",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-SO400M-16-SigLIP2-512.json

@@ -0,0 +1,33 @@
+{
+    "embed_dim": 1152,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 512,
+        "timm_model_name": "vit_so400m_patch16_siglip_512",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-SO400M-16-SigLIP2-512",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-bigG-14-CLIPA-336.json

@@ -0,0 +1,27 @@
+{
+    "embed_dim": 1280,
+    "vision_cfg": {
+        "image_size": 336,
+        "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
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/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
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-bigG-14-quickgelu.json

@@ -0,0 +1,19 @@
+{
+    "embed_dim": 1280,
+    "quick_gelu": true,
+    "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
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-bigG-14.json

@@ -0,0 +1,18 @@
+{
+    "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
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-e-14.json

@@ -0,0 +1,18 @@
+{
+    "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
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-g-14.json

@@ -0,0 +1,18 @@
+{
+    "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
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-gopt-16-SigLIP2-256.json

@@ -0,0 +1,33 @@
+{
+    "embed_dim": 1536,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 256,
+        "timm_model_name": "vit_giantopt_patch16_siglip_256",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-gopt-16-SigLIP2-256",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViT-gopt-16-SigLIP2-384.json

@@ -0,0 +1,33 @@
+{
+    "embed_dim": 1536,
+    "init_logit_bias": -10,
+    "custom_text": true,
+    "vision_cfg": {
+        "image_size": 384,
+        "timm_model_name": "vit_giantopt_patch16_siglip_384",
+        "timm_model_pretrained": false,
+        "timm_pool": "map",
+        "timm_proj": "none"
+    },
+    "text_cfg": {
+        "context_length": 64,
+        "vocab_size": 256000,
+        "hf_tokenizer_name": "timm/ViT-gopt-16-SigLIP2-384",
+        "tokenizer_kwargs": {
+            "clean": "canonicalize"
+        },
+        "width": 1152,
+        "heads": 16,
+        "layers": 27,
+        "mlp_ratio": 3.7362,
+        "no_causal_mask": true,
+        "proj_bias": true,
+        "pool_type": "last",
+        "norm_kwargs":{
+            "eps": 1e-6
+        },
+        "act_kwargs": {
+            "approximate": "tanh"
+        }
+    }
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-B-LTT.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-B.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-L-256.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-L-336.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-L-384.json

@@ -0,0 +1,20 @@
+{
+    "embed_dim": 768,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_large_384",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 384
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 768,
+      "heads": 12,
+      "layers": 12
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-L.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-L2-256.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-L2-336.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-L2-384.json

@@ -0,0 +1,20 @@
+{
+    "embed_dim": 1024,
+    "vision_cfg": {
+      "timm_model_name": "vitamin_large2_384",
+      "timm_model_pretrained": false,
+      "timm_pool": "",
+      "timm_proj": "linear",
+      "timm_drop": 0.0,
+      "timm_drop_path": 0.1,
+      "image_size": 384
+    },
+    "text_cfg": {
+      "context_length": 77,
+      "vocab_size": 49408,
+      "width": 1024,
+      "heads": 16,
+      "layers": 24
+    },
+    "custom_text": true
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-L2.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-S-LTT.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-S.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-XL-256.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-XL-336.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/ViTamin-XL-384.json

@@ -0,0 +1,20 @@
+{
+    "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
+}

cargos/tvai-v2/视觉-语言模型配置/coca_ViT-B-32.json

@@ -0,0 +1,30 @@
+{
+    "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
+}