daniel/h3-blackwell-runtime
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
Direct MiniMax H3 CUDA runtime
9 commits 3 branches 0 tags 34 MiB
Find a file
Daniel Maddern 054b8d58a0 Cache GB10 SageAttention3 wheel
2026-08-12 15:21:03 +07:00
benchmarks Initial direct H3 runtime 2026-08-12 14:12:42 +07:00
src/h3_blackwell_runtime Cache GB10 SageAttention3 wheel 2026-08-12 15:21:03 +07:00
tools Unpack FL2VA sampler reference state 2026-08-12 14:46:43 +07:00
wheels Cache GB10 SageAttention3 wheel 2026-08-12 15:21:03 +07:00
.gitignore Initial direct H3 runtime 2026-08-12 14:12:42 +07:00
compose.spark-comfy-lab.yml Initial direct H3 runtime 2026-08-12 14:12:42 +07:00
compose.spark.yml Use stable H3 artifact mount 2026-08-12 14:27:16 +07:00
Dockerfile.spark Cache GB10 SageAttention3 wheel 2026-08-12 15:21:03 +07:00
PLAN.md Initial direct H3 runtime 2026-08-12 14:12:42 +07:00
pyproject.toml Initial direct H3 runtime 2026-08-12 14:12:42 +07:00
README.md Initial direct H3 runtime 2026-08-12 14:12:42 +07:00

README.md

H3 Blackwell Runtime

Direct MiniMax H3 Ref2VA runtime research project. ComfyUI is the checkpoint and correctness oracle, not the target runtime.

First Gate

Inspect the mounted H3 NVFP4 safetensors headers before designing an importer:

python .\tools\inspect_safetensors.py /runpod-volume/ComfyUI/models/diffusion_models/minimax_h3_ref2va_pruned_nvfp4.safetensors

Write the output to artifacts/checkpoints/ on the mounted volume. The result must identify packed weights, scales, and tensor naming before any kernel conversion work begins.

Benchmark Contract

benchmarks/ref2va-960x544-124f.json is the single-GPU performance contract. Record direct-runner results as JSON and compare them with:

python .\tools\compare_benchmark.py --result direct-result.json

DGX Spark

Dockerfile.spark and compose.spark.yml prepare an ARM64 GB10 development image using the existing AEON CUDA 13/SageAttention3 base. The compose target opens a shell only; it does not start inference.