MiniMax-H3 DynTime sQKV Quants
Experimental: a ComfyUI core patch is required. These FL2VA and Ref2VA checkpoints retain the original FP32 runtime time MLP and physically separate Q, K, and V projections. They do not execute correctly in stock ComfyUI.
Community mixed-precision INT8 conversions of
MiniMaxAI/MiniMax-H3.
All 50 transformer blocks are retained. The repository is separate from the
stock-compatible quants
so patch-required files cannot be confused with ordinary ComfyUI checkpoints.
These are community derivatives, not official MiniMax or ComfyOrg releases.
Naming
FL2VAis text/first-frame/last-frame-to-audio-video generation.Ref2VAis reference-image/video/audio-to-audio-video generation.DT-sQKVmeans dynamic-time conditioning with physically separate Q, K, and V projections.- A filename without
DT-sQKVbelongs to the stock-compatible repository. - Exact INT8/BF16 inventories and GPU classes are documented here instead of being encoded in the filenames.
Choose a checkpoint
| Profile | Direct downloads | File size | GPU class and quant layout |
|---|---|---|---|
| DT-sQKV INT8 ConvRot | FL2VA · Ref2VA | 20.999 GiB | 24 GB · RTX 30/40/50. 170 INT8 + 30 BF16 main semantic matrices; 270 physical INT8 modules; BF16 token refiner. Patch required. |
| DT-sQKV INT8 ConvRot HQ | FL2VA · Ref2VA | 27.994 GiB | 32 GB+ · RTX 30/40/50. 74 INT8 + 126 BF16 main semantic matrices; 122 physical INT8 modules; BF16 token refiner. Patch required. |
The 24/32 GB classes are capacity guidance, not full-workflow peak guarantees. Resolution, frame count, the text encoder, VAEs, attention backend, and offload settings also affect memory use. Moving a 21–28 GiB model across PCIe at every denoising step can make these editions impractical on 8–16 GB GPUs; use a stock-compatible W4/W8W4/NVFP4 checkpoint for those memory classes.
Required ComfyUI patch
Apply:
patches/ComfyUI-MiniMax-H3-DT-sQKV.patch
The patch changes MiniMax-H3 model detection, construction, loading, and forward execution. A loader merely accepting the checkpoint is not sufficient: the patched forward path must evaluate the original time MLP and the separate Q/K/V modules.
What DT-sQKV preserves
| Feature | Stock-compatible quants | These DT-sQKV quants |
|---|---|---|
| Transformer blocks | All 50 retained | All 50 retained |
| Attention storage | Fused qkv_proj = cat(Q,K,V) | Physical q_proj, k_proj, v_proj tensors |
| Attention execution | One fused projection call | Three projection calls |
Original time_embedder | Replaced by measured table | Retained in FP32 |
time_embedder.proj_in | Absent | FP32 weight [5376,256], bias [5376] |
time_embedder.proj_out | Absent | FP32 weight [2688,5376], bias [2688] |
adaln_t_table | FP32 [4097,16] | Absent |
adaln_curve_basis | Absent | FP32 [2688,16] |
adaln_curve_mean | Absent | FP32 [2688] |
| Per-block AdaLN | 51 independent FP32 rank-16 projections | 51 independent FP32 rank-16 projections |
| ComfyUI | Stock | Included core patch required |
The original FP32 time path runs for every requested timestep:
256 -> 5,376 -> 2,688
full_t = SiLU(original_time_embedder(t))
coords = (full_t - mean) @ basis[2,688 x 16]
AdaLN_i(t) = independent_projection_i(coords)
The shared rank-16 basis removes redundant input width from the 51 large AdaLN
projections. It does not replace the original time MLP and does not merge the
per-block AdaLN layers. Full-time relative reconstruction error is about
3e-7; measured basis orthogonality residual is below 6e-7.
Separate Q/K/V layout
The original Diffusers checkpoints contain separate to_q, to_k, and
to_v tensors. These files retain that layout through loading and execution:
- 50 main transformer attention blocks;
- 2 token-refiner attention blocks;
- 156 physical Q/K/V weights;
- three projection calls per attention block;
- no fused
qkv_projmodules.
Released Q/K/V tensors were checked bit-for-bit against their corresponding contiguous slices in the stock-compatible fused checkpoint.
Quantization profiles
Both profiles use ConvRot/Hadamard group size 256, deterministic scale search, and per-row FP32 scales for INT8 weights. Norms, patch projections, output heads, the time MLP, rank-16 basis, and all AdaLN projections retain source precision.
| Profile | Main semantic matrices | Physical INT8 modules | Token refiner | Time/AdaLN path |
|---|---|---|---|---|
DT-sQKV-INT8-ConvRot | 170 INT8 + 30 BF16 | 270 | BF16 | Original FP32 time MLP, basis, mean, and 51 FP32 AdaLN projections |
DT-sQKV-INT8-ConvRot-HQ | 74 INT8 + 126 BF16 | 122 | BF16 | Original FP32 time MLP, basis, mean, and 51 FP32 AdaLN projections |
The standard profile keeps 30 high-risk main matrices in BF16. The HQ profile
keeps every attention-output projection and every MLP fc2 projection in
BF16, together with the 26 highest-error QKV groups. All 50 HQ fc1
projections remain INT8.
Validation
Every checkpoint passed:
- exact key, shape, dtype, and quantization-inventory validation;
- bitwise Q/K/V split verification;
- dynamic-time reconstruction comparison;
- complete CPU load through patched ComfyUI as
MiniMaxH3Model; - remote LFS byte-size and SHA-256 verification.
Reports under reports/ retain their historical internal profile names so the
validation provenance remains intact. Tests used clean ComfyUI commit
14b05228 plus the included patch. Future ComfyUI revisions may require the
same small core changes to be forward-ported.
No full prompt-to-decoded-video perceptual A/B score is claimed. The 32 GB profile was structurally validated but is not claimed to remain fully resident on a 24 GB GPU.
Installation
- Use a ComfyUI revision compatible with the included patch.
- Apply
patches/ComfyUI-MiniMax-H3-DT-sQKV.patchand restart ComfyUI. - Place one selected checkpoint in
ComfyUI/models/diffusion_models/. - Use the matching FL2VA or Ref2VA workflow.
A complete workflow also requires the MiniMax-H3 Qwen3-VL text encoder and the video/audio VAEs from the stock-compatible repository. They are not duplicated here.
License and attribution
Use is subject to the included MiniMax-H3 community license. The base model is by MiniMax. ComfyUI and its quantization runtimes are separate upstream projects. This community conversion is not endorsed by MiniMax or ComfyOrg.