README.md

---
license: mit
language:
- en
- zh
pipeline_tag: text-to-image
tags:
- comfyui
- diffusion-single-file
base_model:
- ACE-Step/Ace-Step1.5
base_model_relation: quantized
---
For more information (including how to compress models yourself), check out https://huggingface.co/DFloat11 and https://github.com/LeanModels/DFloat11

Feel free to request for other models for compression as well, although models whose architecture I am unfamiliar with might be slightly tricky for me.

### How to Use

#### ComfyUI
Install the ComfyUI DFloat11 Extended node via the ComfyUI manager. After installing, simply replace the "Load Diffusion Model" node of an existing workflow with the "Load Diffusion Model" node. If you run into any issues, feel free to leave a comment.

#### Official implementation
This is coming soon, but I suspect that these existing compressed weights might be compatible out-of-the-box with the official implementation.

### Compression Details

This is the `pattern_dict` for compressing ACEStep15-based models in ComfyUI:

```python
pattern_dict_comfyui = {
    r"decoder\.time_embed": (
        "linear_1",
        "linear_2",
        "time_proj",
    ),
    r"decoder\.time_embed_r": (
        "linear_1",
        "linear_2",
        "time_proj",
    ),
    
    r"decoder\.layers\.\d+": (
        "self_attn.q_proj",
        "self_attn.k_proj",
        "self_attn.v_proj",
        "self_attn.o_proj",
        "cross_attn.q_proj",
        "cross_attn.k_proj",
        "cross_attn.v_proj",
        "cross_attn.o_proj",
        "mlp.gate_proj",
        "mlp.up_proj",
        "mlp.down_proj",
    ),
    
    r"encoder\.lyric_encoder\.layers\.\d++": (
        "self_attn.q_proj",
        "self_attn.k_proj",
        "self_attn.v_proj",
        "self_attn.o_proj",
        "mlp.gate_proj",
        "mlp.up_proj",
        "mlp.down_proj",
    ),
    r"encoder\.timbre_encoder\.layers\.\d+": (
        "self_attn.q_proj",
        "self_attn.k_proj",
        "self_attn.v_proj",
        "self_attn.o_proj",
        "mlp.gate_proj",
        "mlp.up_proj",
        "mlp.down_proj",
    ),

    r"tokenizer\.attention_pooler\.layers\.\d+": (
        "self_attn.q_proj",
        "self_attn.k_proj",
        "self_attn.v_proj",
        "self_attn.o_proj",
        "mlp.gate_proj",
        "mlp.up_proj",
        "mlp.down_proj",
    ),

    r"detokenizer\.layers\.\d+": (
        "self_attn.q_proj",
        "self_attn.k_proj",
        "self_attn.v_proj",
        "self_attn.o_proj",
        "mlp.gate_proj",
        "mlp.up_proj",
        "mlp.down_proj",
    ),
}
```