agurung/v4_savebestearly_sft_qwen7B_25percent_lr_1e4_bptt_offset

Same weights, repacked vs Qwen/Qwen2.5-7B-Instruct. Same weights in different files (re-saved, re-sharded or new metadata), plus 9 new tensors.

Tensors changed: Embeddings 0/1, Norms 0/57, MLP 0/84, Attention 0/196, lm_head 0/1

On the Hub · Downloads on Model Pulse