[quantization] Add per-token NVFP4 quantization MoE support for ReLU2 (#56740)
Signed-off-by: Xuanteng Huang <xuantengh@nvidia.com> Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: OpenAI Codex <noreply@openai.com>
This commit is contained in:
co-authored by
Claude Fable 5.1
OpenAI Codex
parent
b803edbc9f
commit
f38c6778ce
@@ -297,3 +297,16 @@ def test_nemotron_fp4_moe_flashinfer_trtllm(monkeypatch: pytest.MonkeyPatch):
|
||||
hf_overrides=HF_OVERRIDE_TEXT,
|
||||
extra_args=["--moe-backend=flashinfer_trtllm"],
|
||||
)
|
||||
|
||||
|
||||
def test_nemotron_nvfp4_per_token_moe_flashinfer_trtllm(
|
||||
monkeypatch: pytest.MonkeyPatch,
|
||||
):
|
||||
can_initialize(
|
||||
"nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4",
|
||||
hf_overrides=HF_OVERRIDE_TEXT,
|
||||
extra_args=[
|
||||
"--moe-backend=flashinfer_trtllm",
|
||||
"--quantization-config.moe.activation=nvfp4_per_token",
|
||||
],
|
||||
)
|
||||
|
||||
Reference in New Issue
Block a user