prithivMLmods commited on
Commit
02e2c81
·
verified ·
1 Parent(s): 0d15a30

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +10 -0
README.md CHANGED
@@ -3,6 +3,10 @@ base_model:
3
  - MiniMaxAI/MiniMax-Music3
4
  ---
5
 
 
 
 
 
6
  ## Run with llama.cpp
7
 
8
  ### Install dependencies and clone llama.cpp
@@ -135,3 +139,9 @@ The generated audio file will be saved at:
135
  | MiniMax-Music3.Q6_K.gguf | Q6_K | 7.05 GB | [Link](https://huggingface.co/prithivMLmods/MiniMax-Music3-GGUF/blob/main/MiniMax-Music3.Q6_K.gguf) | Very high quality, near perfect, *recommended*. |
136
  | MiniMax-Music3.Q8_0.gguf | Q8_0 | 9.13 GB | [Link](https://huggingface.co/prithivMLmods/MiniMax-Music3-GGUF/blob/main/MiniMax-Music3.Q8_0.gguf) | Extremely high quality, generally unneeded but max available standard quant. |
137
  | MiniMax-Music3.mmproj-bf16.gguf | mmproj-bf16 | 6.64 GB | [Link](https://huggingface.co/prithivMLmods/MiniMax-Music3-GGUF/blob/main/MiniMax-Music3.mmproj-bf16.gguf) | Multimodal projection file in BF16 format. Used for audio/music capabilities. |
 
 
 
 
 
 
 
3
  - MiniMaxAI/MiniMax-Music3
4
  ---
5
 
6
+ # **MiniMax-Music3-GGUF**
7
+
8
+ > **MiniMax Music 3** is an open-weights text-to-music generation model from MiniMax AI that creates complete, structurally coherent songs up to five minutes long in 32 kHz, 16-bit stereo WAV format. Conditioned on section-tagged lyrics (such as `[Verse]` and `[Chorus]`) and detailed style descriptions covering metadata, vocals, and arrangement, it uses a hierarchical Hybrid-LM architecture pairing a Qwen3-initialized **8B Global LLM** for long-range musical structure with a **0.6B Local LLM** for frame-level acoustic detail. Instead of decoding directly from discrete Residual Vector Quantization (RVQ) tokens, the system fuses the hidden states of both language models and synthesizes high-fidelity audio through a **2.4B Flow Matching** module and **Flow-VAE** decoder, supporting deployment via SGLang-Omni, Hugging Face `diffusers`, and ComfyUI—even scaling down to 8 GB VRAM GPUs with layer offloading.
9
+
10
  ## Run with llama.cpp
11
 
12
  ### Install dependencies and clone llama.cpp
 
139
  | MiniMax-Music3.Q6_K.gguf | Q6_K | 7.05 GB | [Link](https://huggingface.co/prithivMLmods/MiniMax-Music3-GGUF/blob/main/MiniMax-Music3.Q6_K.gguf) | Very high quality, near perfect, *recommended*. |
140
  | MiniMax-Music3.Q8_0.gguf | Q8_0 | 9.13 GB | [Link](https://huggingface.co/prithivMLmods/MiniMax-Music3-GGUF/blob/main/MiniMax-Music3.Q8_0.gguf) | Extremely high quality, generally unneeded but max available standard quant. |
141
  | MiniMax-Music3.mmproj-bf16.gguf | mmproj-bf16 | 6.64 GB | [Link](https://huggingface.co/prithivMLmods/MiniMax-Music3-GGUF/blob/main/MiniMax-Music3.mmproj-bf16.gguf) | Multimodal projection file in BF16 format. Used for audio/music capabilities. |
142
+
143
+ ## llama.cpp
144
+
145
+ LLM inference in C/C++ — https://github.com/ggml-org/llama.cpp
146
+
147
+ Quantized using llama.cpp by [prithivMLmods](https://huggingface.co/prithivMLmods).