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

Method is_weight_only

tensorrt_llm/quantization/mode.py:127–128  ·  view source on GitHub ↗
(self)

Source from the content-addressed store, hash-verified

125 return self._all(self.INT4_WEIGHTS, self.WEIGHTS_AND_ACTIVATIONS)
126
127 def is_weight_only(self):
128 return self.is_int4_weight_only() or self.is_int8_weight_only()
129
130 def is_int8_weight_only_per_group(self):
131 return self.is_int8_weight_only() and self._any(self.PER_GROUP)

Callers 15

to_dictMethod · 0.95
build_gptFunction · 0.80
enc_dec_build_helperFunction · 0.80
test_weights_onlyMethod · 0.80
test_int8_kv_cacheMethod · 0.80
quant_dequantFunction · 0.80
set_weight_layerMethod · 0.80
add_tllm_weightFunction · 0.80
buildFunction · 0.80
get_weight_dtype_and_idFunction · 0.80
get_quant_methodFunction · 0.80
has_weight_only_quantMethod · 0.80

Calls 2

is_int4_weight_onlyMethod · 0.95
is_int8_weight_onlyMethod · 0.95

Tested by 4

test_weights_onlyMethod · 0.64
test_int8_kv_cacheMethod · 0.64
quant_dequantFunction · 0.64
set_weight_layerMethod · 0.64