Enable Online FP8&NF4 on Minimax-H3

examples/minimax_h3/README.md

@@ -291,6 +291,68 @@ The Ulysses degree must divide 56 attention heads. Scripts must run from their g
 processes can spawn safely. H100 examples request packed FlashAttention 4 and fall back to packed PyTorch SDPA when
 FlashAttention 4 is unavailable.
 
+## Online DiT Quantization
+
+MiniMax H3 supports two single-GPU online quantization backends for the DiT transformer Linear layers:
+
+| CLI value | Backend | Weight/activation path |
+|---|---|---|
+| torchao-fp8 | TorchAO | FP8 dynamic activation and FP8 weight when supported, otherwise TorchAO's FP8 weight-only path |
+| bnb-nf4 | bitsandbytes | NF4 weight-only with BF16 compute |
+
+Both paths convert the 258 Linear layers in the main and token-refiner transformer blocks. The FP32 video/audio
+patch projections, timestep embedding, output projections, text encoder, and VAEs retain their reference dtypes.
+The BF16 DiT is loaded from the original shards, moved to CUDA after text encoding, quantized on first denoising use,
+and then kept resident for the pipeline lifetime. This ordering avoids a simultaneous BF16 text encoder and DiT on
+one GPU and avoids unsupported CPU transfers of quantized tensor subclasses.
+TorchAO conversion has a transient memory peak near the BF16 footprint; use the full 80 GB device without colocated workloads.
+
+Use the dedicated TorchAO FP8 example:
+
+~~~bash
+python examples/minimax_h3/minimax_h3_fl2va_torchao_fp8_h100.py \
+  --mode t2va \
+  --duration 5 \
+  --output outputs/minimax_h3_torchao_fp8.mp4
+~~~
+
+Or the dedicated bitsandbytes NF4 example:
+
+~~~bash
+python examples/minimax_h3/minimax_h3_fl2va_bnb_nf4_h100.py \
+  --mode t2va \
+  --duration 5 \
+  --output outputs/minimax_h3_bnb_nf4.mp4
+~~~
+
+The standard FL2VA, Ref2VA, and JSON request CLIs also accept
+--quantization with either torchao-fp8 or bnb-nf4. The Python loader accepts the same names:
+
+~~~python
+from examples.minimax_h3.common import load_minimax_h3_pipeline
+
+pipeline = load_minimax_h3_pipeline(
+    "/path/to/MiniMaxAI_MiniMax-H3",
+    partition="FL2VA",
+    quantization="torchao-fp8",
+)
+~~~
+
+Online quantization currently requires ulysses_degree=1, tp_degree=1, and FSDP disabled. Quantizing before TP/FSDP
+would invalidate those wrappers' BF16 parameter-sharding contract, so unsupported combinations fail before checkpoint
+loading.
+
+For matched BF16/FP8/NF4 profiling, use the validation benchmark. It writes the synchronized MP4 plus a JSON report
+containing load time, end-to-end generation time, stage timings, and denoising allocator peaks:
+
+~~~bash
+python tools/validation/benchmark_minimax_h3_quantization.py \
+  --backend torchao-fp8 \
+  --duration 5 \
+  --steps 50 \
+  --output outputs/minimax_h3_torchao_fp8_50step.mp4
+~~~
+
 For multi-GPU resident profiles, `WorkerTensorChannel` transports text conditioning, visual condition rows, and the
 final video latent directly between worker groups. CUDA intermediates therefore do not stage through the parent
 process or CPU. The pipeline reports media, text, condition VAE, denoising, video/audio decode, allocator peak, and
@@ -314,10 +376,10 @@ The standard four-GPU profile already uses Ulysses2 x TP2 and therefore leaves F
 `load_minimax_h3_pipeline` directly to construct another supported combination; the product of Ulysses and TP degrees
 must be 1, 2, or 4.
 
-Ring attention, CFG parallelism, pipeline parallelism, sparse attention, quantization, and `torch.compile` are not
-enabled for H3. Video-VAE parallelism is spatial tiling over the existing TP process group, not parameter tensor
-parallelism. The dedicated service manifests expose the pipeline without adding framework-level configuration fields
-or changing the shared request schema.
+Ring attention, CFG parallelism, pipeline parallelism, sparse attention, and `torch.compile` are not enabled for H3.
+Video-VAE parallelism is spatial tiling over the existing TP process group, not parameter tensor parallelism. The
+dedicated service manifests expose the pipeline without adding framework-level configuration fields or changing the
+shared request schema.
 
 ## Four-GPU Regression
 

examples/minimax_h3/common.py

@@ -19,6 +19,9 @@
     ModelRuntimeConfig,
     OffloadConfig,
     ParallelConfig,
+    QuantConfig,
+    QuantKernelBackend,
+    QuantType,
     WeightOffloadType,
 )
 from telefuser.core.module_manager import ModuleManager
@@ -127,6 +130,34 @@ def _checkpoint_shards(component: Path) -> list[str]:
     return shards
 
 
+def minimax_h3_quant_config(quantization: str | QuantType | None) -> QuantConfig:
+    """Resolve a public MiniMax H3 online-quantization name to runtime config."""
+    if quantization is None:
+        return QuantConfig()
+    if isinstance(quantization, str):
+        normalized = quantization.strip().lower().replace("_", "-")
+        names = {
+            "torchao-fp8": QuantType.TORCHAO_FP8,
+            "bnb-nf4": QuantType.BNB_NF4,
+        }
+        try:
+            quant_type = names[normalized]
+        except KeyError as exc:
+            raise ValueError("quantization must be 'torchao-fp8', 'bnb-nf4', or None") from exc
+    elif isinstance(quantization, QuantType):
+        quant_type = quantization
+    else:
+        raise TypeError("quantization must be a string, QuantType, or None")
+
+    backends = {
+        QuantType.TORCHAO_FP8: QuantKernelBackend.TORCHAO,
+        QuantType.BNB_NF4: QuantKernelBackend.BITSANDBYTES,
+    }
+    if quant_type not in backends:
+        raise ValueError(f"MiniMax H3 does not support online quantization type {quant_type.name}")
+    return QuantConfig(enabled=True, quant_type=quant_type, kernel_backend=backends[quant_type])
+
+
 def load_minimax_h3_pipeline(
     model_root: str | Path,
     *,
@@ -141,6 +172,7 @@ def load_minimax_h3_pipeline(
     feature_cache_config: FeatureCacheConfig | None = None,
     adaln_cache_path: str | Path | None = None,
     online_adaln_cache: bool = False,
+    quantization: str | QuantType | None = None,
 ) -> MiniMaxH3Pipeline:
     if adaln_cache_path is not None and online_adaln_cache:
         raise ValueError("Choose either adaln_cache_path or online_adaln_cache, not both.")
@@ -162,6 +194,11 @@ def load_minimax_h3_pipeline(
         raise ValueError("enable_fsdp requires multi-GPU sequence parallelism without tensor parallelism")
     if (adaln_cache_path is not None or online_adaln_cache) and resolved_enable_fsdp:
         raise ValueError("AdaLN cache modes do not yet support FSDP deployment.")
+    quant_config = minimax_h3_quant_config(quantization)
+    if quant_config.enabled and world_size != 1:
+        raise ValueError("MiniMax H3 online quantization currently requires a single-GPU profile")
+    if quant_config.enabled and resolved_enable_fsdp:
+        raise ValueError("MiniMax H3 online quantization cannot be combined with FSDP")
     if isinstance(attn_impl, str):
         try:
             attn_impl = AttnImplType[attn_impl]
@@ -171,13 +208,20 @@ def load_minimax_h3_pipeline(
     if not component_root.is_dir():
         raise FileNotFoundError(f"MiniMax H3 partition not found: {component_root}")
     runtime_device = torch.device(device)
+    if quant_config.enabled and runtime_device.type != "cuda":
+        raise ValueError("MiniMax H3 online quantization requires a CUDA device")
     use_resident_modules = world_size > 1 or resolved_enable_fsdp
     resident_offload = OffloadConfig(
         offload_type=(
             WeightOffloadType.NO_CPU_OFFLOAD if use_resident_modules else WeightOffloadType.MODEL_CPU_OFFLOAD
         ),
         pin_cpu_memory=False,
     )
+    dit_offload = (
+        OffloadConfig(offload_type=WeightOffloadType.NO_CPU_OFFLOAD, pin_cpu_memory=False)
+        if quant_config.enabled
+        else resident_offload
+    )
     text_parallel = (
         ParallelConfig(
             device_ids=list(range(resolved_encoder_tp)),
@@ -203,9 +247,10 @@ def load_minimax_h3_pipeline(
         device_type=runtime_device.type,
         device_id=runtime_device.index or 0,
         torch_dtype=torch.bfloat16,
-        offload_config=resident_offload,
+        offload_config=dit_offload,
         attention_config=AttentionConfig.dense_attention(attn_impl),
         feature_cache_config=feature_cache_config or FeatureCacheConfig(),
+        quant_config=quant_config,
         parallel_config=ParallelConfig(
             device_ids=list(range(world_size)),
             sp_ulysses_degree=ulysses_degree,
@@ -344,6 +389,7 @@ def save_generation(result: MiniMaxH3Generation, output_path: str | Path) -> Non
     "load_minimax_h3_pipeline",
     "load_minimax_h3_request",
     "minimax_h3_adaln_cache_timesteps",
+    "minimax_h3_quant_config",
     "partition_for_minimax_h3_request",
     "run_minimax_h3_request",
     "save_generation",

examples/minimax_h3/minimax_h3_fl2va_bnb_nf4_h100.py

@@ -0,0 +1,55 @@
+# SPDX-License-Identifier: Apache-2.0
+"""MiniMax H3 FL2VA example with bitsandbytes NF4 online quantization."""
+
+from __future__ import annotations
+
+from copy import deepcopy
+from functools import wraps
+
+if __package__:
+    from . import minimax_h3_fl2va_h100 as base
+else:
+    try:
+        from examples.minimax_h3 import minimax_h3_fl2va_h100 as base
+    except ModuleNotFoundError:
+        import minimax_h3_fl2va_h100 as base
+
+PPL_CONFIG = {
+    **base.PPL_CONFIG,
+    "name": "minimax_h3_fl2va_bnb_nf4_h100",
+    "quantization": "bnb-nf4",
+}
+PIPELINE_MANIFEST = deepcopy(base.PIPELINE_MANIFEST)
+PIPELINE_MANIFEST["pipeline_name"] = PPL_CONFIG["name"]
+
+
+@wraps(base.run)
+def run(*args: object, **kwargs: object) -> base.MiniMaxH3Generation:
+    return base.run(*args, **kwargs)
+
+
+@wraps(base.run_with_file)
+def run_with_file(*args: object, **kwargs: object) -> dict[str, str]:
+    return base.run_with_file(*args, **kwargs)
+
+
+def get_pipeline(
+    parallelism: int = 1,
+    model_root: str = PPL_CONFIG["model_root"],
+    **kwargs: object,
+) -> base.MiniMaxH3Pipeline:
+    """Load the single-GPU bitsandbytes NF4 FL2VA pipeline."""
+    return base.get_pipeline(
+        parallelism,
+        model_root,
+        quantization=PPL_CONFIG["quantization"],
+        **kwargs,
+    )
+
+
+def main() -> None:
+    base._main(PPL_CONFIG["quantization"])
+
+
+if __name__ == "__main__":
+    main()

examples/minimax_h3/minimax_h3_fl2va_h100.py

@@ -38,6 +38,7 @@
     "feature_cache_model_type": "MiniMax-H3-Base",
     "feature_cache_n_derivatives": 1,
     "feature_cache_taylor_threshold": 2,
+    "quantization": None,
 }
 
 
@@ -88,6 +89,7 @@ def get_pipeline(
     feature_cache_model_type: str = PPL_CONFIG["feature_cache_model_type"],
     feature_cache_n_derivatives: int = PPL_CONFIG["feature_cache_n_derivatives"],
     feature_cache_taylor_threshold: int = PPL_CONFIG["feature_cache_taylor_threshold"],
+    quantization: str | None = PPL_CONFIG["quantization"],
 ) -> MiniMaxH3Pipeline:
     """Load the FL2VA checkpoint partition for one, two, or four GPUs."""
     tp_degree = 2 if parallelism == 4 else 1
@@ -108,6 +110,7 @@ def get_pipeline(
             n_derivatives=feature_cache_n_derivatives,
             taylor_threshold=feature_cache_taylor_threshold,
         ),
+        quantization=quantization,
     )
 
 
@@ -241,7 +244,7 @@ def run_with_file(
     return {"output_path": str(Path(output_path))}
 
 
-def main() -> None:
+def _main(default_quantization: str | None = PPL_CONFIG["quantization"]) -> None:
     parser = argparse.ArgumentParser(description="Generate MiniMax H3 T2VA/FL2VA audio-video on H100 GPUs")
     parser.add_argument("--model-root", default=PPL_CONFIG["model_root"])
     parser.add_argument("--mode", choices=("t2va", "first-frame", "last-frame", "first-last"))
@@ -265,6 +268,12 @@ def main() -> None:
     parser.add_argument("--flow-shift", type=float, default=PPL_CONFIG["flow_shift"])
     parser.add_argument("--audio-flow-shift", type=float, default=PPL_CONFIG["audio_flow_shift"])
     parser.add_argument("--device", default=PPL_CONFIG["device"])
+    parser.add_argument(
+        "--quantization",
+        choices=("torchao-fp8", "bnb-nf4"),
+        default=default_quantization,
+        help="Online DiT Linear quantization backend (single GPU only).",
+    )
     parser.add_argument("--gpu-num", "--ulysses-degree", dest="gpu_num", type=int, choices=(1, 2, 4), default=1)
     parser.add_argument(
         "--attn-impl",
@@ -319,6 +328,7 @@ def main() -> None:
         feature_cache_model_type=args.feature_cache_model_type,
         feature_cache_n_derivatives=args.feature_cache_n_derivatives,
         feature_cache_taylor_threshold=args.feature_cache_taylor_threshold,
+        quantization=args.quantization,
     )
     try:
         result = run_with_file(
@@ -339,5 +349,9 @@ def main() -> None:
         pipeline.stop()
 
 
+def main() -> None:
+    _main()
+
+
 if __name__ == "__main__":
     main()

examples/minimax_h3/minimax_h3_fl2va_torchao_fp8_h100.py

@@ -0,0 +1,55 @@
+# SPDX-License-Identifier: Apache-2.0
+"""MiniMax H3 FL2VA example with TorchAO FP8 online quantization."""
+
+from __future__ import annotations
+
+from copy import deepcopy
+from functools import wraps
+
+if __package__:
+    from . import minimax_h3_fl2va_h100 as base
+else:
+    try:
+        from examples.minimax_h3 import minimax_h3_fl2va_h100 as base
+    except ModuleNotFoundError:
+        import minimax_h3_fl2va_h100 as base
+
+PPL_CONFIG = {
+    **base.PPL_CONFIG,
+    "name": "minimax_h3_fl2va_torchao_fp8_h100",
+    "quantization": "torchao-fp8",
+}
+PIPELINE_MANIFEST = deepcopy(base.PIPELINE_MANIFEST)
+PIPELINE_MANIFEST["pipeline_name"] = PPL_CONFIG["name"]
+
+
+@wraps(base.run)
+def run(*args: object, **kwargs: object) -> base.MiniMaxH3Generation:
+    return base.run(*args, **kwargs)
+
+
+@wraps(base.run_with_file)
+def run_with_file(*args: object, **kwargs: object) -> dict[str, str]:
+    return base.run_with_file(*args, **kwargs)
+
+
+def get_pipeline(
+    parallelism: int = 1,
+    model_root: str = PPL_CONFIG["model_root"],
+    **kwargs: object,
+) -> base.MiniMaxH3Pipeline:
+    """Load the single-GPU TorchAO FP8 FL2VA pipeline."""
+    return base.get_pipeline(
+        parallelism,
+        model_root,
+        quantization=PPL_CONFIG["quantization"],
+        **kwargs,
+    )
+
+
+def main() -> None:
+    base._main(PPL_CONFIG["quantization"])
+
+
+if __name__ == "__main__":
+    main()