Samosa Chat Maple 2-bit SSD streaming pack
This repository contains the production artifact layout used by Samosa Chat's
native Maple runtime. It is a storage-only repack of the MIT-licensed
deepgrove/maple-preview-2bit-mlx
checkpoint pinned at revision
361db5da5e74ff6fcdd852d478e1f266ce11013a.
The model's numerical weights are unchanged. Expert tensors are stored in
fixed-size aligned records in maple-experts.bin, allowing Samosa to read only
the routed experts from SSD. Non-expert tensors are stored in
maple-resident.safetensors. maple-manifest.json describes and validates the
packed layout.
These files are intended for the bundled samosa-maple runtime. They are not a
drop-in MLX checkpoint because the original expert shards have been replaced by
Samosa's streaming container.
Runtime files
maple-experts.bin: aligned expert records streamed on demand from SSDmaple-resident.safetensors: non-expert tensors retained by the runtimemaple-manifest.json: strict layout metadataconfig.jsonand tokenizer files: model configuration and prompt encoding
Samosa verifies every artifact's byte length and SHA-256 digest before atomic installation.
- Downloads last month
- -
Model tree for deepanwa/Samosa-Chat-Maple-2bit-SSD
Base model
deepgrove/maple-preview-2bit-mlx