Efficient-Large-Model/Fast_dLLM_v2_1.5B
Full fine-tune vs Qwen/Qwen2.5-1.5B-Instruct. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 57/57, MLP 84/84, Attention 196/196
1 derivatives checked
- divelab/relay-fastdllm-v2-c40m60-distilled-step200: Full fine-tune