amazingvince/zephyr-smol_llama-100m-sft-full
Partial fine-tune vs BEE-spoke-data/smol_llama-101M-GQA. Partial change: 76% of the weights changed (attention, lm_head, mlp, norms), in 6 of 6 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 13/13, MLP 18/18, Attention 24/24
3 derivatives checked
- afrideva/zephyr-smol_llama-100m-sft-full-GGUF: Same weights, converted
- tensorblock/zephyr-smol_llama-100m-sft-full-GGUF: Quantization
- amazingvince/zephyr-smol_llama-100m-dpo-full: Full fine-tune
1 other repos hold byte-for-byte identical weight files.