From 96ae5d44485f21dbe135c771d4780ea36d350132 Mon Sep 17 00:00:00 2001 From: Ray Tien Date: Tue, 11 Aug 2026 14:31:45 +0800 Subject: [PATCH] Fix dependency versions and Diffusers links --- README.ja.md | 4 ++-- README.ko.md | 4 ++-- README.md | 4 ++-- README.zh-CN.md | 4 ++-- requirements.txt | 18 ++++++++++-------- 5 files changed, 18 insertions(+), 16 deletions(-) diff --git a/README.ja.md b/README.ja.md index 80b4a26..360b711 100644 --- a/README.ja.md +++ b/README.ja.md @@ -211,7 +211,7 @@ hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" "Ref2VA/ hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" --local-dir MiniMax-H3 ``` -diffusers ユーザーは手動でダウンロードする必要はありません。`ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` が必要なコンポーネントだけを取得します。読み込み方法は [diffusers documentation](https://github.com/huggingface/diffusers/blob/minimax-h3/docs/source/en/api/pipelines/minimax_h3.md) を参照してください。 +diffusers ユーザーは手動でダウンロードする必要はありません。`ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` が必要なコンポーネントだけを取得します。読み込み方法は [diffusers documentation](https://github.com/huggingface/diffusers/blob/f53d552036a0d1bd5570782a39cd40cfabf112bc/docs/source/en/api/pipelines/minimax_h3.md) を参照してください。 モデルのサービングには以下の推論フレームワークを推奨します: @@ -219,7 +219,7 @@ diffusers ユーザーは手動でダウンロードする必要はありませ - [vLLM](https://github.com/vllm-project/vllm) \- see [vllm recipes](https://recipes.vllm.ai/MiniMaxAI/MiniMax-H3) -- [diffusers](https://github.com/huggingface/diffusers) \- see [diffusers docs](https://github.com/huggingface/diffusers/blob/minimax-h3/docs/source/en/api/pipelines/minimax_h3.md) +- [diffusers](https://github.com/huggingface/diffusers) \- see [diffusers docs](https://github.com/huggingface/diffusers/blob/f53d552036a0d1bd5570782a39cd40cfabf112bc/docs/source/en/api/pipelines/minimax_h3.md) - [ComfyUI](https://github.com/Comfy-Org/ComfyUI) \- see [Comfy tutorial](https://docs.comfy.org/tutorials/video/minimax/minimax-h3); use [R2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_r2v.json) / [T2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_t2v.json) diff --git a/README.ko.md b/README.ko.md index 0370b94..60b8ef1 100644 --- a/README.ko.md +++ b/README.ko.md @@ -211,7 +211,7 @@ hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" "Ref2VA/ hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" --local-dir MiniMax-H3 ``` -diffusers 사용자는 수동 다운로드가 필요하지 않습니다. `ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")`가 필요한 구성 요소만 가져옵니다. 로딩 방법은 [diffusers documentation](https://github.com/huggingface/diffusers/blob/minimax-h3/docs/source/en/api/pipelines/minimax_h3.md)을 참고하세요. +diffusers 사용자는 수동 다운로드가 필요하지 않습니다. `ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")`가 필요한 구성 요소만 가져옵니다. 로딩 방법은 [diffusers documentation](https://github.com/huggingface/diffusers/blob/f53d552036a0d1bd5570782a39cd40cfabf112bc/docs/source/en/api/pipelines/minimax_h3.md)을 참고하세요. 모델 서빙에는 다음 추론 프레임워크를 권장합니다: @@ -219,7 +219,7 @@ diffusers 사용자는 수동 다운로드가 필요하지 않습니다. `Modula - [vLLM](https://github.com/vllm-project/vllm) \- see [vllm recipes](https://recipes.vllm.ai/MiniMaxAI/MiniMax-H3) -- [diffusers](https://github.com/huggingface/diffusers) \- see [diffusers docs](https://github.com/huggingface/diffusers/blob/minimax-h3/docs/source/en/api/pipelines/minimax_h3.md) +- [diffusers](https://github.com/huggingface/diffusers) \- see [diffusers docs](https://github.com/huggingface/diffusers/blob/f53d552036a0d1bd5570782a39cd40cfabf112bc/docs/source/en/api/pipelines/minimax_h3.md) - [ComfyUI](https://github.com/Comfy-Org/ComfyUI) \- see [Comfy tutorial](https://docs.comfy.org/tutorials/video/minimax/minimax-h3); use [R2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_r2v.json) / [T2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_t2v.json) diff --git a/README.md b/README.md index 9789aa4..ada93ad 100644 --- a/README.md +++ b/README.md @@ -215,7 +215,7 @@ hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" "Ref2VA/ hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" --local-dir MiniMax-H3 ``` -diffusers users do not need a manual download: `ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` fetches exactly the components it needs. See the [diffusers documentation](https://github.com/huggingface/diffusers/blob/minimax-h3/docs/source/en/api/pipelines/minimax_h3.md) for loading recipes. +diffusers users do not need a manual download: `ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` fetches exactly the components it needs. See the [diffusers documentation](https://github.com/huggingface/diffusers/blob/f53d552036a0d1bd5570782a39cd40cfabf112bc/docs/source/en/api/pipelines/minimax_h3.md) for loading recipes. We recommend the following inference frameworks to serve the model: @@ -223,7 +223,7 @@ We recommend the following inference frameworks to serve the model: - [vLLM](https://github.com/vllm-project/vllm) \- see [vllm recipes](https://recipes.vllm.ai/MiniMaxAI/MiniMax-H3) -- [diffusers](https://github.com/huggingface/diffusers) \- see [diffusers docs](https://github.com/huggingface/diffusers/blob/minimax-h3/docs/source/en/api/pipelines/minimax_h3.md) +- [diffusers](https://github.com/huggingface/diffusers) \- see [diffusers docs](https://github.com/huggingface/diffusers/blob/f53d552036a0d1bd5570782a39cd40cfabf112bc/docs/source/en/api/pipelines/minimax_h3.md) - [ComfyUI](https://github.com/Comfy-Org/ComfyUI) \- see [Comfy tutorial](https://docs.comfy.org/tutorials/video/minimax/minimax-h3); use [R2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_r2v.json) / [T2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_t2v.json) diff --git a/README.zh-CN.md b/README.zh-CN.md index 9721242..f85a16d 100644 --- a/README.zh-CN.md +++ b/README.zh-CN.md @@ -211,7 +211,7 @@ hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" "Ref2VA/ hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" --local-dir MiniMax-H3 ``` -diffusers 用户无需手动下载：`ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` 会准确拉取所需组件。加载方法请参阅 [diffusers documentation](https://github.com/huggingface/diffusers/blob/minimax-h3/docs/source/en/api/pipelines/minimax_h3.md)。 +diffusers 用户无需手动下载：`ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` 会准确拉取所需组件。加载方法请参阅 [diffusers documentation](https://github.com/huggingface/diffusers/blob/f53d552036a0d1bd5570782a39cd40cfabf112bc/docs/source/en/api/pipelines/minimax_h3.md)。 我们推荐使用以下推理框架来服务模型： @@ -219,7 +219,7 @@ diffusers 用户无需手动下载：`ModularPipeline.from_pretrained("MiniMaxAI - [vLLM](https://github.com/vllm-project/vllm) \- see [vllm recipes](https://recipes.vllm.ai/MiniMaxAI/MiniMax-H3) -- [diffusers](https://github.com/huggingface/diffusers) \- see [diffusers docs](https://github.com/huggingface/diffusers/blob/minimax-h3/docs/source/en/api/pipelines/minimax_h3.md) +- [diffusers](https://github.com/huggingface/diffusers) \- see [diffusers docs](https://github.com/huggingface/diffusers/blob/f53d552036a0d1bd5570782a39cd40cfabf112bc/docs/source/en/api/pipelines/minimax_h3.md) - [ComfyUI](https://github.com/Comfy-Org/ComfyUI) \- see [Comfy tutorial](https://docs.comfy.org/tutorials/video/minimax/minimax-h3); use [R2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_r2v.json) / [T2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_t2v.json) diff --git a/requirements.txt b/requirements.txt index 7f5c20c..f2e39c6 100644 --- a/requirements.txt +++ b/requirements.txt @@ -11,16 +11,13 @@ # tighter if you need an exact CUDA build (e.g. torch==2.4.1+cu121). torch>=2.4.0 -# diffusers is required for the H3 pipeline / modular pipeline. The official -# docs reference the `minimax-h3` branch of diffusers for the latest class -# additions; once that lands on PyPI, tighten the upper bound accordingly. -# Until then, install from source: `pip install "git+https://github.com/huggingface/diffusers.git@minimax-h3"` -diffusers>=0.32.2 +# MiniMax-H3 was merged after diffusers 0.39.0. Pin the merge +# commit so installs contain the H3 modular pipeline and remain reproducible. +diffusers @ git+https://github.com/huggingface/diffusers.git@f53d552036a0d1bd5570782a39cd40cfabf112bc # transformers provides the Qwen3-VL text encoder / processor / tokenizer used -# by the H3 pipeline. Newer is generally better here; keep >=4.45 to get -# Qwen3-VL support. -transformers>=4.45.0 +# by the H3 pipeline. Qwen3-VL support starts in transformers 4.57. +transformers>=4.57.0 # safetensors is used to load the VAE / Transformer weights directly. safetensors>=0.4.3 @@ -36,6 +33,11 @@ huggingface_hub>=0.25.0 # numpy is used throughout the audio / video VAE code. numpy>=1.24.0 +# Direct dependencies used by the released Visual and Audio VAE code. +torchvision>=0.19.0 +einops>=0.8.0 +PyYAML>=6.0.1 + # Pillow is required by the Qwen3-VL processor for image decoding. Pillow>=10.0.0