[Speculate Decoding] Fix reasoning_phase_token_constraint call args in SpeculativeSampler (#7402)

This commit is contained in:
lonelygsh
2026-04-15 12:45:23 +08:00
committed by GitHub
parent 8995a38fa4
commit f7a2418ce2
@@ -996,7 +996,8 @@ class SpeculativeSampler(nn.Layer):
if self.enf_gen_phase_tag:
reasoning_phase_token_constraint(
logits,
sampling_metadata.pre_token_ids,
token_ids_all,
prompt_lens,
share_inputs["stop_flags"],
share_inputs["seq_lens_this_time"],
share_inputs["seq_lens_encoder"],