vrfai/Qwen3-ASR-0.6B-int8-QAD

Partial fine-tune (fewer layers) vs Qwen/Qwen3-ASR-0.6B. Qwen/Qwen3-ASR-0.6B matches part of vrfai/Qwen3-ASR-0.6B-int8-QAD (612 of its 1200 tensors; the rest of vrfai/Qwen3-ASR-0.6B-int8-QAD has no counterpart). On the shared part: Partial change: 51% of the weights changed (attention, mlp, norms, other), in 28 of 28 layers.

Tensors changed: other 5/11, MLP 94/156, Norms 47/187, Attention 125/256, lm_head 0/1, Embeddings 0/1

On the Hub · Downloads on Model Pulse