[Bugfix] Fix NVFP4+MTP crash: force unquantized mtp.fc for Qwen3Next (#46316)

Signed-off-by: Ranran Haoran Zhang <ranzhang@redhat.com>
This commit is contained in:
Ranran
2026-06-25 09:05:04 -07:00
committed by GitHub
parent d490b98162
commit e45b279928
+11 -2
View File
@@ -64,13 +64,22 @@ class Qwen3NextMultiTokenPredictor(nn.Module):
config.hidden_size,
)
# Workaround: mtp.fc is stored as BF16 in NVFP4 checkpoints but is
# missing from the checkpoint quant exclude list (its `ignore` glob
# does not cover `mtp.fc`). Force unquantized to match the weights,
# mirroring the Qwen3.5 MTP handling (PR #38832).
fc_quant = (
None
if (quant_config and quant_config.get_name() == "modelopt_fp4")
else quant_config
)
self.fc = ColumnParallelLinear(
self.config.hidden_size * 2,
self.config.hidden_size,
gather_output=True,
bias=False,
return_bias=False,
quant_config=quant_config,
quant_config=fc_quant,
prefix=f"{prefix}.fc",
)
@@ -242,7 +251,7 @@ class Qwen3NextMTP(nn.Module, QwenNextMixtureOfExperts):
"k_proj",
"v_proj",
],
"gate_up_proj": ["up_proj", "down_proj"],
"gate_up_proj": ["gate_proj", "up_proj"],
}
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):