jsyeom/

fp8-wfp8a16kvfp8

	FP8 W-FP8 / A-FP16 / KV-FP8 (Llama-3.1-8B-Instruct)




	


	w_of_wfp8a16kvfp8_llama_31_8b/ — FP8 weights

The per-tensor scaled FP8 (E4M3) weights of Llama-3.1-8B-Instruct — a type-conversion variant (weights cast to FP8). Stored per layer: layer_0.safetensors … layer_31.safetensors + embeddings.safetensors. The 7 linears per layer are cast to FP8 (fp8 weight + fp32 per-tensor scale); everything else stays fp16:

key dtype shape

self_attn.q_proj.weight… See the full description on the dataset page: https://huggingface.co/datasets/jsyeom/fp8-wfp8a16kvfp8.