MCPcopy Create free account
hub / github.com/ROCm/AMDMIGraphX / quantize_bf16

Function quantize_bf16

src/quantization.cpp:78–86  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

76}
77
78void quantize_bf16(program& prog, const std::vector<std::string>& ins_names)
79{
80 run_passes(prog,
81 {normalize_ops{},
82 optimize_module{{"quantizelinear", "dequantizelinear"}},
83 truncate_float_pass{ins_names, shape::bf16_type},
84 optimize_module{{"quantizelinear", "dequantizelinear"}}},
85 quant_tracer());
86}
87
88static void quantize_8bits(program& prog,
89 const target& t,

Callers 5

create_programMethod · 0.50
create_programMethod · 0.50
create_programMethod · 0.50
create_programMethod · 0.50

Calls 2

run_passesFunction · 0.85
quant_tracerFunction · 0.85

Tested by 4

create_programMethod · 0.40
create_programMethod · 0.40
create_programMethod · 0.40
create_programMethod · 0.40