MCPcopy
hub / github.com/NVIDIA/TensorRT-LLM / preprocess_init_params

Function preprocess_init_params

tensorrt_llm/quantization/quantize.py:117–123  ·  view source on GitHub ↗
(init_params, name, module)

Source from the content-addressed store, hash-verified

115 }
116
117 def preprocess_init_params(init_params, name, module):
118 init_params["quant_mode"] = quant_config.quant_mode
119 if isinstance(module, ColumnLinear):
120 module_name = name.rsplit('.', 1)[-1]
121 init_params["transb"] = module_name == "lm_head"
122 if "tp_rank" in init_params:
123 init_params["tp_rank"] = model_cfg.mapping.tp_rank
124
125 model = quantize_layers(
126 model,

Callers 1

quantize_layersFunction · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected