Update model card with H100 skill 5s validation

README.md

@@ -3,46 +3,63 @@ license: other
 base_model: hunyuanvideo-community/HunyuanVideo
 library_name: sglang
 pipeline_tag: text-to-video
+tags:
+- sglang
+- diffusion
+- hunyuanvideo
+- modelopt
+- fp8
 ---
 
 # HunyuanVideo ModelOpt FP8 SGLang Transformer
 
-This repository contains a SGLang-ready ModelOpt FP8 transformer override for `hunyuanvideo-community/HunyuanVideo`.
-Load it together with the original base model:
+This repository contains a SGLang-ready ModelOpt FP8 transformer override for [`hunyuanvideo-community/HunyuanVideo`](https://huggingface.co/hunyuanvideo-community/HunyuanVideo).
+It only replaces the DiT/transformer weights; text encoders, VAE, scheduler, tokenizer, and other non-transformer components are loaded from the original base model.
+
+The checkpoint is intended for SGLang Diffusion with the HunyuanVideo FP8 support from [sgl-project/sglang#23199](https://github.com/sgl-project/sglang/pull/23199).
+
+## Usage
 
 ```bash
-sglang generate \
-  --backend=sglang \
-  --model-path hunyuanvideo-community/HunyuanVideo \
-  --transformer-path BBuf/HunyuanVideo-ModelOpt-FP8-SGLang \
-  --height 544 --width 960 --num-frames 17 \
-  --num-inference-steps 8 --seed 0 \
-  --prompt "A cinematic shot of a red sports car driving through rain at night, reflections on wet streets, smooth camera movement." \
-  --save-output
+sglang generate   --backend=sglang   --model-path hunyuanvideo-community/HunyuanVideo   --transformer-path BBuf/HunyuanVideo-ModelOpt-FP8-SGLang   --prompt "A cat and a dog baking a cake together in a kitchen. The cat is carefully measuring flour, while the dog is stirring the batter with a wooden spoon. The kitchen is cozy, with sunlight streaming through the window."   --seed=42   --text-encoder-cpu-offload   --pin-cpu-memory   --num-frames=65   --fps=13   --width=848   --height=480   --num-inference-steps=30   --save-output   --warmup   --enable-torch-compile
 ```
 
-## Validation Snapshot
+The command above follows the HunyuanVideo preset used by the `sglang-diffusion-benchmark-profile` skill. `65` frames at `13` fps gives an exact `5.000s` video.
+
+## H100 Validation Snapshot
+
+Validation was run on one H100 GPU using rank0 (`CUDA_VISIBLE_DEVICES=0`) with `--backend=sglang`. Logs show `Using pipeline from model_index.json: HunyuanVideoPipeline`; no diffusers fallback markers were observed.
 
-Validated on one H100 GPU with SGLang main plus the Qwen Image ModelOpt FP8 support commit and the HunyuanVideo FP8 changes.
+Artifacts:
 
-- Base model: `hunyuanvideo-community/HunyuanVideo`
-- Precision: ModelOpt FP8 transformer override, non-transformer components from the base model
-- Resolution: 960x544, 17 frames
-- Steps: 8
-- Seed: 0
-- Offload flags: `--dit-cpu-offload false --dit-layerwise-offload false`
+- Validation tree: [`validation/h100_skill_5s_20260420`](https://huggingface.co/BBuf/HunyuanVideo-ModelOpt-FP8-SGLang/tree/main/validation/h100_skill_5s_20260420)
+- Full command and run summary: [`result_summary_skill_5s.md`](https://huggingface.co/BBuf/HunyuanVideo-ModelOpt-FP8-SGLang/blob/main/validation/h100_skill_5s_20260420/result_summary_skill_5s.md)
+- BF16 video: [`hunyuanvideo_bf16_skill_5s.mp4`](https://huggingface.co/BBuf/HunyuanVideo-ModelOpt-FP8-SGLang/resolve/main/validation/h100_skill_5s_20260420/artifacts/hunyuanvideo_bf16_skill_5s.mp4)
+- FP8 video: [`hunyuanvideo_fp8_skill_5s.mp4`](https://huggingface.co/BBuf/HunyuanVideo-ModelOpt-FP8-SGLang/resolve/main/validation/h100_skill_5s_20260420/artifacts/hunyuanvideo_fp8_skill_5s.mp4)
+- Profiler traces: [BF16](https://huggingface.co/BBuf/HunyuanVideo-ModelOpt-FP8-SGLang/resolve/main/validation/h100_skill_5s_20260420/profiler/bf16/445b0a02-cb9c-4793-9b49-1e5a102facb4-5_steps-global-rank0.trace.json.gz), [FP8](https://huggingface.co/BBuf/HunyuanVideo-ModelOpt-FP8-SGLang/resolve/main/validation/h100_skill_5s_20260420/profiler/fp8/0bb0f929-c3dc-42f5-b834-cfffb815c3e0-5_steps-global-rank0.trace.json.gz), [kernel summary](https://huggingface.co/BBuf/HunyuanVideo-ModelOpt-FP8-SGLang/blob/main/validation/h100_skill_5s_20260420/profiler/kernel_summary_skill_5s.md)
 
-Offline throughput, random dataset, 3 prompts, batch size 1:
+![BF16 vs FP8 5s contact sheet](https://huggingface.co/BBuf/HunyuanVideo-ModelOpt-FP8-SGLang/resolve/main/validation/h100_skill_5s_20260420/artifacts/hunyuanvideo_bf16_fp8_skill_5s_contact_sheet.png)
 
-| checkpoint | total duration | latency/request | MP/s | denoise steady avg |
-| --- | ---: | ---: | ---: | ---: |
-| BF16 | 26.11 s | 8.70 s | 1.02 | ~0.379 s/step |
-| FP8 | 24.90 s | 8.30 s | 1.07 | ~0.319 s/step |
+Benchmark, warmup excluded:
 
-Approximate speedup: 1.05x end-to-end throughput and 1.19x steady denoise step time.
-Transformer memory at load dropped from 23.88 GB to 15.45 GB.
+| Metric | BF16 | FP8 | Delta | Speedup |
+|---|---:|---:|---:|---:|
+| E2E latency | 59.546 s | 54.748 s | -4.798 s (-8.1%) | 1.09x |
+| Denoising stage | 42.542 s | 37.980 s | -4.562 s (-10.7%) | 1.12x |
+| Avg denoise step | 1.4180 s | 1.2659 s | -0.1521 s | 1.12x |
+| Decoding stage | 16.692 s | 16.458 s | -0.233 s (-1.4%) | 1.01x |
+| Text encoding | 0.308 s | 0.306 s | -0.002 s (-0.7%) | 1.01x |
+
+Profiler kernel share over 5 profiled denoise timesteps. Profiler timings include profiling overhead and are not used as benchmark latency numbers.
+
+| Precision | Total CUDA op time | Top CUDA/kernel shares |
+|---|---:|---|
+| BF16 | 17.055 s | `cudaMemcpyAsync` 41.54%; FlashAttention 31.99%; BF16 GEMM kernels 9.77%, 8.16%, 2.11% |
+| FP8 | 15.324 s | `cudaMemcpyAsync` 40.62%; FlashAttention 36.80%; FP8 Cutlass GEMM 12.83%; `_static_quant_fp8` 1.37% |
 
 ## Conversion Notes
 
 The checkpoint was converted from a ModelOpt FP8 export with SGLang's `build_modelopt_fp8_transformer` tool using the `hunyuan-video` preset.
-The preset keeps numerically sensitive embedder/modulation/output layers in BF16 and maps ModelOpt/diffusers module names to SGLang runtime module names for fused QKV and fused QKV+MLP projections.
+The preset keeps numerically sensitive embedder, modulation, and output layers in BF16, and maps ModelOpt/diffusers module names to SGLang runtime module names for fused QKV and fused QKV+MLP projections.
+
+One runtime caveat: the CLI can keep the same offload flags as the BF16 skill preset, but ModelOpt FP8 checkpoints currently force `dit_cpu_offload` off while preserving layerwise offload behavior for restored FP8 tensor strides.