wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-02-16 00:12:33 +08:00
Code Issues Actions 22 Packages Projects Releases Wiki Activity

Update ace15.py to allow min_p sampling (#12373)

Browse Source
This commit is contained in:
askmyteapot 2026-02-12 11:28:48 +10:00 committed by GitHub
parent 3fe61cedda
commit e5ae670a40
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 15 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
comfy/text_encoders/ace15.py
View File

@ -16,6 +16,7 @@ def sample_manual_loop_no_classes(
temperature: float = 0.85,
top_p: float = 0.9,
top_k: int = None,
min_p: float = 0.000,
seed: int = 1,
min_tokens: int = 1,
max_new_tokens: int = 2048,
@ -80,6 +81,12 @@ def sample_manual_loop_no_classes(
min_val = top_k_vals[..., -1, None]
cfg_logits[cfg_logits < min_val] = remove_logit_value
if min_p is not None and min_p > 0:
probs = torch.softmax(cfg_logits, dim=-1)
p_max = probs.max(dim=-1, keepdim=True).values
indices_to_remove = probs < (min_p * p_max)
cfg_logits[indices_to_remove] = remove_logit_value
if top_p is not None and top_p < 1.0:
sorted_logits, sorted_indices = torch.sort(cfg_logits, descending=True)
cumulative_probs = torch.cumsum(torch.softmax(sorted_logits, dim=-1), dim=-1)
@ -110,7 +117,7 @@ def sample_manual_loop_no_classes(
return output_audio_codes
def generate_audio_codes(model, positive, negative, min_tokens=1, max_tokens=1024, seed=0, cfg_scale=2.0, temperature=0.85, top_p=0.9, top_k=0):
def generate_audio_codes(model, positive, negative, min_tokens=1, max_tokens=1024, seed=0, cfg_scale=2.0, temperature=0.85, top_p=0.9, top_k=0, min_p=0.000):
positive = [[token for token, _ in inner_list] for inner_list in positive]
positive = positive[0]
@ -134,7 +141,7 @@ def generate_audio_codes(model, positive, negative, min_tokens=1, max_tokens=102
paddings = []
ids = [positive]
return sample_manual_loop_no_classes(model, ids, paddings, cfg_scale=cfg_scale, temperature=temperature, top_p=top_p, top_k=top_k, seed=seed, min_tokens=min_tokens, max_new_tokens=max_tokens)
return sample_manual_loop_no_classes(model, ids, paddings, cfg_scale=cfg_scale, temperature=temperature, top_p=top_p, top_k=top_k, min_p=min_p, seed=seed, min_tokens=min_tokens, max_new_tokens=max_tokens)
class ACE15Tokenizer(sd1_clip.SD1Tokenizer):
@ -192,6 +199,7 @@ class ACE15Tokenizer(sd1_clip.SD1Tokenizer):
temperature = kwargs.get("temperature", 0.85)
top_p = kwargs.get("top_p", 0.9)
top_k = kwargs.get("top_k", 0.0)
min_p = kwargs.get("min_p", 0.000)
duration = math.ceil(duration)
kwargs["duration"] = duration
@ -239,6 +247,7 @@ class ACE15Tokenizer(sd1_clip.SD1Tokenizer):
"temperature": temperature,
"top_p": top_p,
"top_k": top_k,
"min_p": min_p,
}
return out
@ -299,7 +308,7 @@ class ACE15TEModel(torch.nn.Module):
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["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"])
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"])
out["audio_codes"] = [audio_codes]
return base_out, None, out

5
comfy_extras/nodes_ace.py
View File

@ -49,13 +49,14 @@ class TextEncodeAceStepAudio15(io.ComfyNode):
io.Float.Input("temperature", default=0.85, min=0.0, max=2.0, step=0.01, advanced=True),
io.Float.Input("top_p", default=0.9, min=0.0, max=2000.0, step=0.01, advanced=True),
io.Int.Input("top_k", default=0, min=0, max=100, advanced=True),
io.Float.Input("min_p", default=0.000, min=0.0, max=1.0, step=0.001, advanced=True),
],
outputs=[io.Conditioning.Output()],
)
@classmethod
def execute(cls, clip, tags, lyrics, seed, bpm, duration, timesignature, language, keyscale, generate_audio_codes, cfg_scale, temperature, top_p, top_k) -> io.NodeOutput:
tokens = clip.tokenize(tags, lyrics=lyrics, bpm=bpm, duration=duration, timesignature=int(timesignature), language=language, keyscale=keyscale, seed=seed, generate_audio_codes=generate_audio_codes, cfg_scale=cfg_scale, temperature=temperature, top_p=top_p, top_k=top_k)
def execute(cls, clip, tags, lyrics, seed, bpm, duration, timesignature, language, keyscale, generate_audio_codes, cfg_scale, temperature, top_p, top_k, min_p) -> io.NodeOutput:
tokens = clip.tokenize(tags, lyrics=lyrics, bpm=bpm, duration=duration, timesignature=int(timesignature), language=language, keyscale=keyscale, seed=seed, generate_audio_codes=generate_audio_codes, cfg_scale=cfg_scale, temperature=temperature, top_p=top_p, top_k=top_k, min_p=min_p)
conditioning = clip.encode_from_tokens_scheduled(tokens)
return io.NodeOutput(conditioning)