optimum

a49a1165 - Added cache_block_outputs parameter to handle models with non-regular structure such as ChatGLM (#1479)

Added cache_block_outputs parameter to handle models with non-regular structure such as ChatGLM (#1479) * Added cache_block_outputs parameter to handle models with non-regular structure in GPTQ * Code style * Added variable description * Applied comments * Changed default. Added more docstring * Added a test for cache_block_outputs feature * Style