WAS commited on
Commit
dcc864e
·
verified ·
1 Parent(s): dc4bebf

Kandinsky 6: replace raw transformers with int8_convrot and W6A8 quantisations

Browse files
README.md CHANGED
@@ -100,6 +100,8 @@ licence files sit beside the files they cover.
100
  | `models/kandinsky6/diffusion_models/kandinsky6_lite_5s_w6a8.safetensors` | [kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers), Kandinsky Lab, quantised for ComfyUI (w6a8) | MIT |
101
  | `models/kandinsky6/diffusion_models/kandinsky6_pro_distill_5s_int8_convrot.safetensors` | [kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers), Kandinsky Lab, quantised for ComfyUI (int8_convrot) | MIT |
102
  | `models/kandinsky6/diffusion_models/kandinsky6_pro_distill_5s_w6a8.safetensors` | [kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers), Kandinsky Lab, quantised for ComfyUI (w6a8) | MIT |
 
 
103
  | `models/kandinsky6/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors` | [Qwen/Qwen2.5-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct), as repackaged by [Comfy-Org](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI) | Apache-2.0 |
104
  | `models/kandinsky6/text_encoders/clip_l.safetensors` | [openai/clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14), as repackaged by [Comfy-Org](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged) | MIT |
105
  | `models/kandinsky6/vae/kandinsky6_audio_vae.safetensors` | [MMAudio](https://huggingface.co/hkchengrex/MMAudio)'s 44.1 kHz autoencoder and [BigVGAN v2](https://huggingface.co/nvidia/bigvgan_v2_44khz_128band_512x), as packed by Kandinsky Lab | **CC BY-NC 4.0, non-commercial** (MMAudio); MIT (BigVGAN) |
 
100
  | `models/kandinsky6/diffusion_models/kandinsky6_lite_5s_w6a8.safetensors` | [kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers), Kandinsky Lab, quantised for ComfyUI (w6a8) | MIT |
101
  | `models/kandinsky6/diffusion_models/kandinsky6_pro_distill_5s_int8_convrot.safetensors` | [kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers), Kandinsky Lab, quantised for ComfyUI (int8_convrot) | MIT |
102
  | `models/kandinsky6/diffusion_models/kandinsky6_pro_distill_5s_w6a8.safetensors` | [kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers), Kandinsky Lab, quantised for ComfyUI (w6a8) | MIT |
103
+ | `models/kandinsky6/diffusion_models/kandinsky6_pro_5s_int8_convrot.safetensors` | [kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers), Kandinsky Lab, quantised for ComfyUI (int8_convrot) | MIT |
104
+ | `models/kandinsky6/diffusion_models/kandinsky6_pro_5s_w6a8.safetensors` | [kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers), Kandinsky Lab, quantised for ComfyUI (w6a8) | MIT |
105
  | `models/kandinsky6/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors` | [Qwen/Qwen2.5-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct), as repackaged by [Comfy-Org](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI) | Apache-2.0 |
106
  | `models/kandinsky6/text_encoders/clip_l.safetensors` | [openai/clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14), as repackaged by [Comfy-Org](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged) | MIT |
107
  | `models/kandinsky6/vae/kandinsky6_audio_vae.safetensors` | [MMAudio](https://huggingface.co/hkchengrex/MMAudio)'s 44.1 kHz autoencoder and [BigVGAN v2](https://huggingface.co/nvidia/bigvgan_v2_44khz_128band_512x), as packed by Kandinsky Lab | **CC BY-NC 4.0, non-commercial** (MMAudio); MIT (BigVGAN) |
models/kandinsky6/NOTICE.md CHANGED
@@ -13,6 +13,8 @@ transformers are ComfyUI quantisations of the upstream file, described below.
13
  | `diffusion_models/kandinsky6_lite_5s_w6a8.safetensors` | W6A8 of [kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers) at `5686616f04318cbec7b0a4978b4ff1697395de15`, `transformer/diffusion_pytorch_model.safetensors` | MIT, Kandinsky Lab: `diffusion_models/LICENSE-Kandinsky6` |
14
  | `diffusion_models/kandinsky6_pro_distill_5s_int8_convrot.safetensors` | int8 ConvRot of [kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers) at `1018d5828716e8bea3de99609f45c924de9c5564`, `transformer/diffusion_pytorch_model.safetensors` | MIT, Kandinsky Lab: `diffusion_models/LICENSE-Kandinsky6` |
15
  | `diffusion_models/kandinsky6_pro_distill_5s_w6a8.safetensors` | W6A8 of [kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers) at `1018d5828716e8bea3de99609f45c924de9c5564`, `transformer/diffusion_pytorch_model.safetensors` | MIT, Kandinsky Lab: `diffusion_models/LICENSE-Kandinsky6` |
 
 
16
  | `text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors` | [Comfy-Org/Qwen-Image_ComfyUI](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI) at `1f12b17be14c89b026c51a91d67c32f84bb047bc`, `split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors`, Comfy-Org's fp8 repackaging of [Qwen/Qwen2.5-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) | Apache-2.0, Alibaba Cloud: `text_encoders/LICENSE-Qwen2.5-VL` |
17
  | `text_encoders/clip_l.safetensors` | [Comfy-Org/HunyuanVideo_repackaged](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged) at `533a43813f01517d9e624f9e824a85bb94094120`, `split_files/text_encoders/clip_l.safetensors`, the text tower of [openai/clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) | MIT, OpenAI: `text_encoders/LICENSE-CLIP` |
18
  | `vae/kandinsky6_audio_vae.safetensors` | [kandinskylab/Kandinsky-6.0-Lite-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Lite-distill-5s-Diffusers) at `efbaeb9961770cc82fece5b9cb2197049d8c983c`, `audio_vae/diffusion_pytorch_model.safetensors` | **CC BY-NC 4.0, non-commercial**, and MIT: see below |
@@ -31,11 +33,15 @@ Quantised: the query, key, value and output projections and both feed-forward we
31
  transformer block whose input width is a multiple of 256. Every other tensor is bfloat16. On Lite the
32
  audio stream is 896 wide, so its block weights stay bfloat16 too.
33
 
 
 
 
34
  Source transformers, SHA-256:
35
 
36
- b1b52668561d9138d93bb4da13c9999e615bf569be9f8cf65732d54652447f53 kandinskylab/Kandinsky-6.0-Lite-distill-5s-Diffusers@efbaeb9961 transformer
37
  f84ea3266560bbd1f203a874b8cee43e080073820facaf3306cab8fae2798a97 kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers@1018d58287 transformer
 
38
  1550090ca56fed8c75dda0d30f8aa7b46775e8e1ebb086b70ad532e40eb878f9 kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers@5686616f04 transformer
 
39
 
40
  ## SHA-256
41
 
@@ -45,6 +51,8 @@ Source transformers, SHA-256:
45
  375b16bd1f3981806d7f9513978f0c1e0a460d074d080fe1f267c189cbb895f4 diffusion_models/kandinsky6_lite_5s_w6a8.safetensors
46
  8b96fc841b9ecd1d96218b70ecffd846ba8b620fa24d087fc99a547fbdc4edf9 diffusion_models/kandinsky6_pro_distill_5s_int8_convrot.safetensors
47
  d26c82be04d5e2d97f4bcf184f0af09768f3af92ade3b7fff52ee181f0a0b968 diffusion_models/kandinsky6_pro_distill_5s_w6a8.safetensors
 
 
48
  cb5636d852a0ea6a9075ab1bef496c0db7aef13c02350571e388aea959c5c0b4 text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors
49
  660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd text_encoders/clip_l.safetensors
50
  dfccdf9d0d8d3de270a23b097400f21adb91f4ec683bdc26369cb65b08a5af87 vae/kandinsky6_audio_vae.safetensors
@@ -70,4 +78,4 @@ non-commercial use, unchanged from Kandinsky Lab's file.
70
  | File | Where it is | Why |
71
  |---|---|---|
72
  | `hunyuan_video_vae_bf16.safetensors`, the video VAE | [Comfy-Org/HunyuanVideo_repackaged](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged), `split_files/vae/` | Tencent Hunyuan Community License |
73
- | Kandinsky 6 Pro, not distilled | [kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers) | Gated upstream |
 
13
  | `diffusion_models/kandinsky6_lite_5s_w6a8.safetensors` | W6A8 of [kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers) at `5686616f04318cbec7b0a4978b4ff1697395de15`, `transformer/diffusion_pytorch_model.safetensors` | MIT, Kandinsky Lab: `diffusion_models/LICENSE-Kandinsky6` |
14
  | `diffusion_models/kandinsky6_pro_distill_5s_int8_convrot.safetensors` | int8 ConvRot of [kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers) at `1018d5828716e8bea3de99609f45c924de9c5564`, `transformer/diffusion_pytorch_model.safetensors` | MIT, Kandinsky Lab: `diffusion_models/LICENSE-Kandinsky6` |
15
  | `diffusion_models/kandinsky6_pro_distill_5s_w6a8.safetensors` | W6A8 of [kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers) at `1018d5828716e8bea3de99609f45c924de9c5564`, `transformer/diffusion_pytorch_model.safetensors` | MIT, Kandinsky Lab: `diffusion_models/LICENSE-Kandinsky6` |
16
+ | `diffusion_models/kandinsky6_pro_5s_int8_convrot.safetensors` | int8 ConvRot of [kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers) at `295ca976810e318fb45c870ba4fc72c9a37fd548`, `transformer/diffusion_pytorch_model.safetensors` | MIT, Kandinsky Lab: `diffusion_models/LICENSE-Kandinsky6` |
17
+ | `diffusion_models/kandinsky6_pro_5s_w6a8.safetensors` | W6A8 of [kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers) at `295ca976810e318fb45c870ba4fc72c9a37fd548`, `transformer/diffusion_pytorch_model.safetensors` | MIT, Kandinsky Lab: `diffusion_models/LICENSE-Kandinsky6` |
18
  | `text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors` | [Comfy-Org/Qwen-Image_ComfyUI](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI) at `1f12b17be14c89b026c51a91d67c32f84bb047bc`, `split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors`, Comfy-Org's fp8 repackaging of [Qwen/Qwen2.5-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) | Apache-2.0, Alibaba Cloud: `text_encoders/LICENSE-Qwen2.5-VL` |
19
  | `text_encoders/clip_l.safetensors` | [Comfy-Org/HunyuanVideo_repackaged](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged) at `533a43813f01517d9e624f9e824a85bb94094120`, `split_files/text_encoders/clip_l.safetensors`, the text tower of [openai/clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) | MIT, OpenAI: `text_encoders/LICENSE-CLIP` |
20
  | `vae/kandinsky6_audio_vae.safetensors` | [kandinskylab/Kandinsky-6.0-Lite-distill-5s-Diffusers](https://huggingface.co/kandinskylab/Kandinsky-6.0-Lite-distill-5s-Diffusers) at `efbaeb9961770cc82fece5b9cb2197049d8c983c`, `audio_vae/diffusion_pytorch_model.safetensors` | **CC BY-NC 4.0, non-commercial**, and MIT: see below |
 
33
  transformer block whose input width is a multiple of 256. Every other tensor is bfloat16. On Lite the
34
  audio stream is 896 wide, so its block weights stay bfloat16 too.
35
 
36
+ Kandinsky-6.0-Pro-5s-Diffusers is gated on Hugging Face behind an access form; its licence is MIT,
37
+ the same as every other Kandinsky 6 transformer here.
38
+
39
  Source transformers, SHA-256:
40
 
 
41
  f84ea3266560bbd1f203a874b8cee43e080073820facaf3306cab8fae2798a97 kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers@1018d58287 transformer
42
+ b1b52668561d9138d93bb4da13c9999e615bf569be9f8cf65732d54652447f53 kandinskylab/Kandinsky-6.0-Lite-distill-5s-Diffusers@efbaeb9961 transformer
43
  1550090ca56fed8c75dda0d30f8aa7b46775e8e1ebb086b70ad532e40eb878f9 kandinskylab/Kandinsky-6.0-Lite-5s-Diffusers@5686616f04 transformer
44
+ f0568ea71b9d754b9538808fc61662c22123fd852e9f7c667f6150ed5f725967 kandinskylab/Kandinsky-6.0-Pro-5s-Diffusers@295ca97681 transformer
45
 
46
  ## SHA-256
47
 
 
51
  375b16bd1f3981806d7f9513978f0c1e0a460d074d080fe1f267c189cbb895f4 diffusion_models/kandinsky6_lite_5s_w6a8.safetensors
52
  8b96fc841b9ecd1d96218b70ecffd846ba8b620fa24d087fc99a547fbdc4edf9 diffusion_models/kandinsky6_pro_distill_5s_int8_convrot.safetensors
53
  d26c82be04d5e2d97f4bcf184f0af09768f3af92ade3b7fff52ee181f0a0b968 diffusion_models/kandinsky6_pro_distill_5s_w6a8.safetensors
54
+ e21a0a4eeab4c25491d30b0af05358830cdd556ffe434c0866e41866c2ad694b diffusion_models/kandinsky6_pro_5s_int8_convrot.safetensors
55
+ 30ad0006e81bec43d00afa0c6c97b64dd94b47481287ec8b97d5e91a0ecab815 diffusion_models/kandinsky6_pro_5s_w6a8.safetensors
56
  cb5636d852a0ea6a9075ab1bef496c0db7aef13c02350571e388aea959c5c0b4 text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors
57
  660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd text_encoders/clip_l.safetensors
58
  dfccdf9d0d8d3de270a23b097400f21adb91f4ec683bdc26369cb65b08a5af87 vae/kandinsky6_audio_vae.safetensors
 
78
  | File | Where it is | Why |
79
  |---|---|---|
80
  | `hunyuan_video_vae_bf16.safetensors`, the video VAE | [Comfy-Org/HunyuanVideo_repackaged](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged), `split_files/vae/` | Tencent Hunyuan Community License |
81
+
models/kandinsky6/diffusion_models/kandinsky6_pro_5s_int8_convrot.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e21a0a4eeab4c25491d30b0af05358830cdd556ffe434c0866e41866c2ad694b
3
+ size 34898934560
models/kandinsky6/diffusion_models/kandinsky6_pro_5s_w6a8.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30ad0006e81bec43d00afa0c6c97b64dd94b47481287ec8b97d5e91a0ecab815
3
+ size 29342827504