Skip to content

[benchmark_inference] torchao nvfp4 #2770

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
crcrpar wants to merge 3 commits into
base: main
Choose a base branch
from
Draft

[benchmark_inference] torchao nvfp4 #2770

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a839011
add option to use torchao nvfp4
crcrpar Nov 12, 2025
410b32e
fix typo in torchao quantization of model
crcrpar Nov 12, 2025
a6b6538
Merge branch 'main' into crpa/inference_torchao_nvfp4
crcrpar Nov 26, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 19 additions & 4 deletions thunder/benchmarks/benchmark_inference.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,9 +18,6 @@
import os
import statistics
import time
import warnings
from typing import Any
from collections.abc import Callable
from looseversion import LooseVersion

import torch
Expand Down Expand Up @@ -52,6 +49,7 @@
from thunder.torch.custom_op import _register_custom_op, _register_nvfuser_translator

if TYPE_CHECKING:
from collections.abc import Callable
from typing import Any


Expand Down Expand Up @@ -217,6 +215,7 @@ class InferenceBenchmarkConfig:
num_layers: int | None
num_iterations: int
warmup_iterations: int
use_torchao_nvfp4: bool
enable_nvfp4: bool # Enable NVFP4 registration and quantize GroupedSwiGLU in MoE
fx_report_folder: str | None
enable_nv_linear: bool
Expand Down Expand Up @@ -345,7 +344,21 @@ def __init__(self, config: InferenceBenchmarkConfig):
self.vocab_size = model.vocab_size

if self.config.enable_nvfp4:
_quantize_llama4(model)
if not self.config.use_torchao_nvfp4:
_quantize_llama4(model)
else:
from torchao.quantization import quantize_
from torchao.prototype.mx_formats.inference_workflow import (
NVFP4InferenceConfig,
NVFP4MMConfig,
)

config = NVFP4InferenceConfig(
mm_config=NVFP4MMConfig.DYNAMIC,
use_triton_kernel=True,
use_dynamic_per_tensor_scale=True,
)
quantize_(model, config=config)
self.model = self._compile_model(model)

@property
Expand Down Expand Up @@ -737,6 +750,7 @@ def parse_args() -> argparse.Namespace:
help="Specify the folder for thunderfx_benchmark_report.",
)

parser.add_argument("--use-torchao-nvfp4", action="store_true", help="Use TorchAO for NVFP4")
parser.add_argument(
"--enable-nvfp4",
action="store_true",
Expand Down Expand Up @@ -796,6 +810,7 @@ def main():
warmup_iterations=args.warmup_iterations,
mode=args.mode,
enable_nvfp4=args.enable_nvfp4,
use_torchao_nvfp4=args.use_torchao_nvfp4,
fx_report_folder=args.fx_report_folder,
enable_nv_linear=args.enable_nv_linear,
disable_moe_replacement=args.disable_moe_replacement,
Expand Down
Loading