seanghay/Qwen3-ASR-0.6B-Khmer

Full fine-tune vs Qwen/Qwen3-ASR-0.6B. Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: other 11/11, MLP 156/156, Norms 107/187, Attention 256/256, Embeddings 1/1

1 derivatives checked

• mradermacher/Qwen3-ASR-0.6B-Khmer-GGUF: Same weights, converted (fewer layers)

On the Hub · Downloads on Model Pulse