ModelCloud/GPTQModel

gptqmodel-torch-oracle-numerics

Validate changed quantization math, packed formats, and inference kernels against a Torch oracle; adjudicate near-tie mismatches before rejecting optimized arithmetic.

لغة النص الأصلي: الإنجليزية

gptqmodel-lazy-load-smoke-testing

End-to-end smoke test for GPT-QModel's lazy checkpoint load (LazyTurtle) and first-layer materialization through StageInputsCapture.cache_inputs. Use when verifying lazy-load behavior, first-layer input capture, shell_module_materialize module_path…

لغة النص الأصلي: الإنجليزية

gptqmodel-tokenizer-normalization

Diagnose and correct GPT-QModel tokenizer initialization, tokenization normalization, special-token handling, prompt rendering, and chat-template problems. Use when inference or evaluation quality suggests wrong token IDs or prompts, when a model needs…

لغة النص الأصلي: الإنجليزية