wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-03-16 22:58:19 +08:00
Code Issues Actions 11 Packages Projects Releases Wiki Activity

fix ACE-Step 1.5: max_tokens typo and lyrics embedding truncation

Browse Source 
Two bugs in ace15.py encode_token_weights():

1. max_tokens parameter received min_tokens value:
   `max_tokens=lm_metadata["min_tokens"]` → `max_tokens=lm_metadata["max_tokens"]`
   This caused the LM to always generate minimum-length audio codes
   regardless of the requested duration.

2. lyrics_embeds[:, 0] discarded the full lyric sequence, passing only
   the first token embedding to the diffusion model's lyric encoder.
   Changed to pass the full lyrics_embeds tensor for proper lyrics
   conditioning.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Mark Ledford 2026-02-19 01:36:45 -08:00
parent 6d11cc7354
commit ddd0cd7521
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
comfy/text_encoders/ace15.py
View File

@ -298,11 +298,11 @@ class ACE15TEModel(torch.nn.Module):
self.qwen3_06b.set_clip_options({"layer": [0]})
lyrics_embeds, _, extra_l = self.qwen3_06b.encode_token_weights(token_weight_pairs_lyrics)
out = {"conditioning_lyrics": lyrics_embeds[:, 0]}
out = {"conditioning_lyrics": lyrics_embeds}
lm_metadata = token_weight_pairs["lm_metadata"]
if lm_metadata["generate_audio_codes"]:
audio_codes = generate_audio_codes(getattr(self, self.lm_model, self.qwen3_06b), token_weight_pairs["lm_prompt"], token_weight_pairs["lm_prompt_negative"], min_tokens=lm_metadata["min_tokens"], max_tokens=lm_metadata["min_tokens"], seed=lm_metadata["seed"], cfg_scale=lm_metadata["cfg_scale"], temperature=lm_metadata["temperature"], top_p=lm_metadata["top_p"], top_k=lm_metadata["top_k"], min_p=lm_metadata["min_p"])
audio_codes = generate_audio_codes(getattr(self, self.lm_model, self.qwen3_06b), token_weight_pairs["lm_prompt"], token_weight_pairs["lm_prompt_negative"], min_tokens=lm_metadata["min_tokens"], max_tokens=lm_metadata["max_tokens"], seed=lm_metadata["seed"], cfg_scale=lm_metadata["cfg_scale"], temperature=lm_metadata["temperature"], top_p=lm_metadata["top_p"], top_k=lm_metadata["top_k"], min_p=lm_metadata["min_p"])
out["audio_codes"] = [audio_codes]
return base_out, None, out