quark-torch-file2file-quantization

SkillDocs & knowledge

Low-memory file2file quantization for very large safetensors LLMs that cannot be loaded whole. Use when the user wants to run file2file quantization, adapt a new safetensors checkpoint without loading the full model, register an external LLMTemplate, inspect sharded checkpoint naming, generate wrapper or conversion scripts, or validate low-memory sharded quantization outputs. Trigger for "run file2file quantization", "quantize without loading the model", "large safetensors low-memory quantization", "file2file for DeepSeek/Qwen/MoE", "safetensors naming incompatible", "register LLMTemplate externally".

Available today. Use it from your connected AI after setup.

No other account needed.

Then ask your AI: use the quark-torch-file2file-quantization skill