@xiciyang/training-qat
Training Qat
QAT thin plugin: corpus O1 de-leakage → shared monolithic training loop (training-core, torchao QAT fake quantization in the container) → apply true INT8 quantization to the QAT-trained weights (the repository’s own llmcompressor INT8, producing compressed-tensors) → serving deployment + dual gates
The plugin will be installed here. Keep web if you are unsure.
npx -y @deepseek-ai/dsh plugin --profile web add @xiciyang/training-qat@0.1.0Description
QAT thin plugin: corpus O1 de-leakage → shared monolithic training loop (training-core, torchao QAT fake quantization in the container) → apply true INT8 quantization to the QAT-trained weights (the repository’s own llmcompressor INT8, producing compressed-tensors) → serving deployment + dual gates + cross-plugin reconciliation. Depends on the six @ai-infra/* shared libraries frozen by long-running task 1.
Compatibility and provenance
Training Qat is published as @xiciyang/training-qat and currently resolves to version 0.1.0. The Hub verifies its manifest and preserves the exact installation source for reproducible installs.
- DSH compatibility
- *
- Runtime surfaces
- any
- Release source
- npm
- Registry updated