simaai/Phi-4-mini-instruct-Autoround-Safetensors
Text Generation • 4B • Updated • 79
Pre-quantized safetensor models (GPTQ/AutoRound) and their optimization scripts, ready to be compiled for SiMa.ai hardware.
Note Qwen2.5-VL-7B-Instruct: GPTQ INT4/G256 text decoder and lm_head, GPTQ INT8 per-channel vision stack.