jinx/ComfyUI
1
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2025-08-02 23:14:49 +08:00
Code Issues Activity

Initial ACE-Step model implementation. (#7972)

Browse Source
This commit is contained in:
comfyanonymous
2025-05-07 05:33:34 -07:00
committed by GitHub
parent 271c9c5b9e
commit 16417b40d9
18 changed files with 19738 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
comfy/model_detection.py
View File

@@ -226,6 +226,31 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
dit_config.update(json.loads(metadata["config"]).get("transformer", {}))
return dit_config
if '{}genre_embedder.weight'.format(key_prefix) in state_dict_keys: #ACE-Step model
dit_config = {}
dit_config["audio_model"] = "ace"
dit_config["attention_head_dim"] = 128
dit_config["in_channels"] = 8
dit_config["inner_dim"] = 2560
dit_config["max_height"] = 16
dit_config["max_position"] = 32768
dit_config["max_width"] = 32768
dit_config["mlp_ratio"] = 2.5
dit_config["num_attention_heads"] = 20
dit_config["num_layers"] = 24
dit_config["out_channels"] = 8
dit_config["patch_size"] = [16, 1]
dit_config["rope_theta"] = 1000000.0
dit_config["speaker_embedding_dim"] = 512
dit_config["text_embedding_dim"] = 768
dit_config["ssl_encoder_depths"] = [8, 8]
dit_config["ssl_latent_dims"] = [1024, 768]
dit_config["ssl_names"] = ["mert", "m-hubert"]
dit_config["lyric_encoder_vocab_size"] = 6693
dit_config["lyric_hidden_size"] = 1024
return dit_config
if '{}t_block.1.weight'.format(key_prefix) in state_dict_keys: # PixArt
patch_size = 2
dit_config = {}