Team Ai
Modelpublic

ProCreations/Qwen3.8-27B-Escha-W2-WebGPU

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face
0likes51downloads
Model Card

Qwen3.8 27B Escha W2 — WebGPU Pack

Lossless browser-streaming container for `EschaLabs/Qwen3.8-27B-Escha-W2`, used by `ProCreations/Qwen3.8-27b-W2`.

The original Escha 2/3-bit codes, Hadamard signs, scales, biases, norms, and int8 embedding/head values are copied byte-for-byte. No additional model quantization or approximation is applied. Only file grouping changes.

  • —Source revision: b1fff11fb1606a0ba6965f4c76751cfabe64d5c4
  • —Tensor bytes: 10.15 GB
  • —Format: escha-webgpu-pack-v1