ASLP-lab/FM-Speech

Partial fine-tune vs Qwen/Qwen3-Omni-30B-A3B-Instruct. Partial change: 60% of the weights changed (attention, mlp, other), in 48 of 48 layers.

Tensors changed: other 3/148, MLP 838/1557, Norms 0/562, Attention 320/696, Embeddings 0/21, lm_head 0/16

1 derivatives checked

• mradermacher/FM-Speech-GGUF: Quantization

On the Hub · Downloads on Model Pulse