inference-optimization/Kimi-K3-0.40B-MXFP4
73.1k
Remove weight_scale from decompressed routed_expert weights
Decompress routed_expert weights (MXFP4 -> bf16); ignore routed_expert in qconfig
Upload folder using huggingface_hub
Update config.json
Upload folder using huggingface_hub
Add files using upload-large-folder tool
initial commit
