imeepos/mixvideo-v2
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat: add template json

Browse Source
This commit is contained in:
imeepos
2025-08-15 13:41:17 +08:00
parent c37fff0d09
commit 247eebef9e
261 changed files with 89699 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
cargos/tvai-v2/models/EVA01-g-14-plus.json Normal file
View File

@@ -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
}

18
cargos/tvai-v2/models/EVA01-g-14.json Normal file
View File

@@ -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
}

18
cargos/tvai-v2/models/EVA02-B-16.json Normal file
View File

@@ -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
}

18
cargos/tvai-v2/models/EVA02-E-14-plus.json Normal file
View File

@@ -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
}

18
cargos/tvai-v2/models/EVA02-E-14.json Normal file
View File

@@ -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
}

18
cargos/tvai-v2/models/EVA02-L-14-336.json Normal file
View File

@@ -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
}

18
cargos/tvai-v2/models/EVA02-L-14.json Normal file
View File

@@ -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
}

21
cargos/tvai-v2/models/MobileCLIP-B.json Normal file
View File

@@ -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
}

21
cargos/tvai-v2/models/MobileCLIP-S1.json Normal file
View File

@@ -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
}

21
cargos/tvai-v2/models/MobileCLIP-S2.json Normal file
View File

@@ -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
}

22
cargos/tvai-v2/models/RN101-quickgelu.json Normal file
View File

@@ -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
}
}

21
cargos/tvai-v2/models/RN101.json Normal file
View File

@@ -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
}
}

22
cargos/tvai-v2/models/RN50-quickgelu.json Normal file
View File

@@ -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
}
}

21
cargos/tvai-v2/models/RN50.json Normal file
View File

@@ -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
}
}

22
cargos/tvai-v2/models/RN50x16-quickgelu.json Normal file
View File

@@ -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
}
}

21
cargos/tvai-v2/models/RN50x16.json Normal file
View File

@@ -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
}
}

22
cargos/tvai-v2/models/RN50x4-quickgelu.json Normal file
View File

@@ -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
}
}

21
cargos/tvai-v2/models/RN50x4.json Normal file
View File

@@ -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
}
}

22
cargos/tvai-v2/models/RN50x64-quickgelu.json Normal file
View File

@@ -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
}
}

21
cargos/tvai-v2/models/RN50x64.json Normal file
View File

@@ -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
}
}

29
cargos/tvai-v2/models/ViT-B-16-SigLIP-256.json Normal file
View File

@@ -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
}
}
}

29
cargos/tvai-v2/models/ViT-B-16-SigLIP-384.json Normal file
View File

@@ -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
}
}
}

29
cargos/tvai-v2/models/ViT-B-16-SigLIP-512.json Normal file
View File

@@ -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
}
}
}

29
cargos/tvai-v2/models/ViT-B-16-SigLIP-i18n-256.json Normal file
View File

@@ -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
}
}
}

29
cargos/tvai-v2/models/ViT-B-16-SigLIP.json Normal file
View File

@@ -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
}
}
}

32
cargos/tvai-v2/models/ViT-B-16-SigLIP2-256.json Normal file
View File

@@ -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"
}
}
}

32
cargos/tvai-v2/models/ViT-B-16-SigLIP2-384.json Normal file
View File

@@ -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"
}
}
}

32
cargos/tvai-v2/models/ViT-B-16-SigLIP2-512.json Normal file
View File

@@ -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"
}
}
}

32
cargos/tvai-v2/models/ViT-B-16-SigLIP2.json Normal file
View File

@@ -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"
}
}
}

16
cargos/tvai-v2/models/ViT-B-16-plus-240.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-B-16-plus.json Normal file
View File

@@ -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
}
}

17
cargos/tvai-v2/models/ViT-B-16-quickgelu.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-B-16.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-B-32-256.json Normal file
View File

@@ -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
}
}

32
cargos/tvai-v2/models/ViT-B-32-SigLIP2-256.json Normal file
View File

@@ -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"
}
}
}

16
cargos/tvai-v2/models/ViT-B-32-plus-256.json Normal file
View File

@@ -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
}
}

17
cargos/tvai-v2/models/ViT-B-32-quickgelu.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-B-32.json Normal file
View File

@@ -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
}
}

18
cargos/tvai-v2/models/ViT-H-14-378-quickgelu.json Normal file
View File

@@ -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
}
}

17
cargos/tvai-v2/models/ViT-H-14-378.json Normal file
View File

@@ -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
}
}

26
cargos/tvai-v2/models/ViT-H-14-CLIPA-336.json Normal file
View File

@@ -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
}
}

26
cargos/tvai-v2/models/ViT-H-14-CLIPA.json Normal file
View File

@@ -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
}
}

18
cargos/tvai-v2/models/ViT-H-14-quickgelu.json Normal file
View File

@@ -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
}
}

17
cargos/tvai-v2/models/ViT-H-14.json Normal file
View File

@@ -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
}
}

17
cargos/tvai-v2/models/ViT-H-16.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-L-14-280.json Normal file
View File

@@ -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
}
}

17
cargos/tvai-v2/models/ViT-L-14-336-quickgelu.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-L-14-336.json Normal file
View File

@@ -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
}
}

25
cargos/tvai-v2/models/ViT-L-14-CLIPA-336.json Normal file
View File

@@ -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
}
}

25
cargos/tvai-v2/models/ViT-L-14-CLIPA.json Normal file
View File

@@ -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
}
}

17
cargos/tvai-v2/models/ViT-L-14-quickgelu.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-L-14.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-L-16-320.json Normal file
View File

@@ -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
}
}

29
cargos/tvai-v2/models/ViT-L-16-SigLIP-256.json Normal file
View File

@@ -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
}
}
}

29
cargos/tvai-v2/models/ViT-L-16-SigLIP-384.json Normal file
View File

@@ -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
}
}
}

32
cargos/tvai-v2/models/ViT-L-16-SigLIP2-256.json Normal file
View File

@@ -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"
}
}
}

32
cargos/tvai-v2/models/ViT-L-16-SigLIP2-384.json Normal file
View File

@@ -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"
}
}
}

32
cargos/tvai-v2/models/ViT-L-16-SigLIP2-512.json Normal file
View File

@@ -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"
}
}
}

16
cargos/tvai-v2/models/ViT-L-16.json Normal file
View File

@@ -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
}
}

17
cargos/tvai-v2/models/ViT-M-16-alt.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-M-16.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-M-32-alt.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-M-32.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-S-16-alt.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-S-16.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-S-32-alt.json Normal file
View File

@@ -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
}
}

16
cargos/tvai-v2/models/ViT-S-32.json Normal file
View File

@@ -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
}
}

30
cargos/tvai-v2/models/ViT-SO400M-14-SigLIP-378.json Normal file
View File

@@ -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
}
}
}

30
cargos/tvai-v2/models/ViT-SO400M-14-SigLIP-384.json Normal file
View File

@@ -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
}
}
}

30
cargos/tvai-v2/models/ViT-SO400M-14-SigLIP.json Normal file
View File

@@ -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
}
}
}

33
cargos/tvai-v2/models/ViT-SO400M-14-SigLIP2-378.json Normal file
View File

@@ -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"
}
}
}

33
cargos/tvai-v2/models/ViT-SO400M-14-SigLIP2.json Normal file
View File

@@ -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"
}
}
}

30
cargos/tvai-v2/models/ViT-SO400M-16-SigLIP-i18n-256.json Normal file
View File

@@ -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
}
}
}

33
cargos/tvai-v2/models/ViT-SO400M-16-SigLIP2-256.json Normal file
View File

@@ -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"
}
}
}

33
cargos/tvai-v2/models/ViT-SO400M-16-SigLIP2-384.json Normal file
View File

@@ -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"
}
}
}

33
cargos/tvai-v2/models/ViT-SO400M-16-SigLIP2-512.json Normal file
View File

@@ -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"
}
}
}

27
cargos/tvai-v2/models/ViT-bigG-14-CLIPA-336.json Normal file
View File

@@ -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
}
}

27
cargos/tvai-v2/models/ViT-bigG-14-CLIPA.json Normal file
View File

@@ -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
}
}

19
cargos/tvai-v2/models/ViT-bigG-14-quickgelu.json Normal file
View File

@@ -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
}
}

18
cargos/tvai-v2/models/ViT-bigG-14.json Normal file
View File

@@ -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
}
}

18
cargos/tvai-v2/models/ViT-e-14.json Normal file
View File

@@ -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
}
}

18
cargos/tvai-v2/models/ViT-g-14.json Normal file
View File

@@ -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
}
}

33
cargos/tvai-v2/models/ViT-gopt-16-SigLIP2-256.json Normal file
View File

@@ -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"
}
}
}

33
cargos/tvai-v2/models/ViT-gopt-16-SigLIP2-384.json Normal file
View File

@@ -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"
}
}
}

20
cargos/tvai-v2/models/ViTamin-B-LTT.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-B.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-L-256.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-L-336.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-L-384.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-L.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-L2-256.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-L2-336.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-L2-384.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-L2.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-S-LTT.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-S.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-XL-256.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-XL-336.json Normal file
View File

@@ -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
}

20
cargos/tvai-v2/models/ViTamin-XL-384.json Normal file
View File

@@ -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
}

538
cargos/tvai-v2/models/aaa-10.json Normal file
View File

@@ -0,0 +1,538 @@
{
"backends": {
"coreml": {
"model": "v3s",
"parallel": 1,
"scales": {
"1": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fnet-fp16-[H]x[W]-[S]x-ml.tz",
"gnet-fp16-[H]x[W]-[S]x-ml.tz"
],
"outputs": [
"fnet/output",
"generator/output"
]
},
"2": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fnet-fp16-[H]x[W]-[S]x-ml.tz",
"gnet-fp16-[H]x[W]-[S]x-ml.tz"
],
"outputs": [
"fnet/output",
"generator/output"
]
},
"4": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fnet-fp16-[H]x[W]-[S]x-ml.tz",
"gnet-fp16-[H]x[W]-[S]x-ml.tz"
],
"outputs": [
"fnet/output",
"generator/output"
]
}
}
},
"onnx": {
"model": "v3",
"parallel": 1,
"scales": {
"1": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp32-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
},
"2": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp32-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
},
"4": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp32-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
}
}
},
"onnx16": {
"model": "v3",
"parallel": 1,
"scales": {
"1": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp16-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
},
"2": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp16-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
},
"4": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp16-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
}
}
},
"openvino": {
"model": "v3",
"parallel": 0,
"scales": {
"1": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp16-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
},
"2": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp16-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
},
"4": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp16-[H]x[W]-[S]x-ox.tz"
],
"outputs": [
"generator/output"
]
}
}
},
"tensorrt": {
"capabilities": [
1200,
809,
806,
705
],
"model": "v3",
"parallel": 1,
"scales": {
"1": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp32-[H]x[W]-[S]x-rt[C]-[R]-rt.tz"
],
"outputs": [
"generator/output"
]
},
"2": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp32-[H]x[W]-[S]x-rt[C]-[R]-rt.tz"
],
"outputs": [
"generator/output"
]
},
"4": {
"blocks": [
288,
288,
352,
256,
480,
384,
384,
480,
384,
576,
416,
576,
576,
448,
672,
384,
672,
576
],
"nets": [
"fgnet-fp32-[H]x[W]-[S]x-rt[C]-[R]-rt.tz"
],
"outputs": [
"generator/output"
]
}
}
}
},
"enabled": 0,
"gui": {
"afterImg": "/tldb/images/model-thumbnails/AAA_Enhanced.png",
"beforeImg": "/tldb/images/model-thumbnails/AAA_Original.png",
"desc": "Upscale video with aliasing or moire patterns. Aliasing is common in computer generated (CG) content, or from line-skipping cameras.",
"displayPri": 30,
"hiddenKeywords": [
"AA"
],
"maxScale": 6,
"minScale": 0.25,
"modelPickerParams": {
"vidArtifactType": [
"Noise",
"Blurry"
],
"vidQuality": [
0,
1,
2
],
"vidType": [
"P",
"CG"
]
},
"name": "Artemis - Aliasing or Moire",
"searchableKeywords": [
"Artemis",
"Antialias"
],
"showScale": 1
},
"plugin" : {
"displayName" : "Artemis - Aliasing or Moire",
"videoQuality": [0,1,2]
},
"inputs": {
"fnetInput": "fnet/input",
"gnetInput": "generator/input",
"input": "fgwnet/input",
"previousInput": "pre_inputs",
"previousOutput": "pre_outputs"
},
"extraParameters" : [
{
"default": 0,
"max": 0.1,
"min": 0,
"name": "Add Noise",
"guiName": "Add Noise"
},
{
"default": 0.2,
"max": 1,
"min": 0,
"name": "Recover Details",
"guiName": "Recover Details"
}
],
"interlacedFrames": 0,
"jsonVersion": "1.0.0",
"keepColor": 1,
"key": "CHJPPIOMOMCPNLKEOLFNAPEBPGDDEIGBOOIDFPNMKAMDOEEFMIJJOOEFMEGNMAJJGIGJEPNJKCLDBKJIONNHGKDLPMAOGCHBGFBIHKGHACOBNLOBBONCFFCLBLDIAJNJCDBGFGABNBPJFEFONMKGPEPLBPPKEEGPDEPGOMGEMHBINMAACMKGGPOMAFFJFIDOCGKNALLLBHOIBCGMEJAGGJJGIIALINECCBHKDPKCFNLCICEMMMBJJDEHGMPCJEEJEOBCKDJEEIBLCHJHFONCIKKGLFOKLMNIEKPHDCNHKDFEJOPEMEJJKHCAMEKLKBBM",
"minAppVersion": "1.5.0",
"modelType": 1,
"noiseLevel": 0.4,
"postflight": 0,
"preflight": 5,
"shortName": "aaa",
"displayName": "Aliasing or Moire",
"version": "10"
}

Some files were not shown because too many files have changed in this diff Show More