strisunshinewentingwang/pytorch

Pull Requests Commits

address review comments

saketh-are committed 2 years ago

7fcac020

Merge remote-tracking branch 'origin/viable/strict' into activation_functions_i_opinfos

saketh-are committed 2 years ago

cccf0824

Revert D32541986: [pytorch][PR] [opinfo] use dtypes instead of dtypesIfCPU

albanD committed 2 years ago

148f3238

[BE] minor improvement to dist quantization (#67401)

wanchaol committed 2 years ago

7c6a8a47

[BE] delete frontend.cpp (#67400)

wanchaol committed 2 years ago

fb556c91

[opinfo] use dtypes instead of dtypesIfCPU (#67619)

kshitij12345 committed 2 years ago

d2a90f91

Fix test issue with householder_product for non-contiguous inputs. (#68231)

v0dro committed 2 years ago

2d06c081

Sparse CSR CPU: add `triangular_solve_out` (#62180)

IvanYashchuk committed 2 years ago

3b3dc1ad

dbr quant overhead[9/x]: precalculate when to skip op_convert_after_hook (#68432)

vkuzo committed 2 years ago

e1c449ff

dbr quant: remove more asserts from hot paths (#68431)

vkuzo committed 2 years ago

ba230de1

speed up quantized relu6 inplace kernel (#68404)

vkuzo committed 2 years ago

95c00cf0

dbr quant: simplify relatedness logic (#68374)

vkuzo committed 2 years ago

592053f1

dbr quant overhead[8/x]: small speedup in op_needs_quantization (#68373)

vkuzo committed 2 years ago

f1021bcf

dbr quant overhead[7/x]: speed up AutoQuantizationState.reset_to_new_call (#68372)

vkuzo committed 2 years ago

74ba1067

dbr quant overhead[6/x]: remove unneeded isinstance checks in `op_convert_before_hook` (#68371)

vkuzo committed 2 years ago

b7d58745

dbr quant overhead[5/x]: remove unnecessary asserts (#68370)

vkuzo committed 2 years ago

b3a7d696

dbr quant: clean up key types in AutoQuantizationState mappings (#68369)

vkuzo committed 2 years ago

16a6e061

dbr quant overhead[4/x]: speed up hook type calculations (#68351)

vkuzo committed 2 years ago

3fc9bc43

dbr quant overhead[3/x]: speed up AutoQuantizationState.mark_cur_op_complete (#68350)

vkuzo committed 2 years ago

c72ffee4

dbr quant overhead[2/x]: precalculate op_convert_info (#68347)

vkuzo committed 2 years ago

c7ecf149

dbr quant: move model level utils into own file (#68346)

vkuzo committed 2 years ago

9fba8971

dbr quant: clean up AutoQuantizationState.get_op_convert_info flag (#68345)

vkuzo committed 2 years ago

629f9a55

dbr quant: refactor AutoQuantizationState._get_packed_param_name (#68344)

vkuzo committed 2 years ago

52cc9cb0

dbr quant: refactor AutoQuantizationState._get_input_args_quant_dequant_info (#68343)

vkuzo committed 2 years ago

2755cf45

dbr quant: refactor `get_quantized_op` to only use `seen_op_info` (#68342)

vkuzo committed 2 years ago

57472ec4

dbr quant: refactor `get_func_output_obs_type` to only use `seen_op_info` (#68341)

vkuzo committed 2 years ago

9cf4779e

dbr quant overhead[1/x]: remove expensive calls to named_modules (#68309)

vkuzo committed 2 years ago

f8b084c5

dbr quantization: inline scale and zp (#68251)

vkuzo committed 2 years ago

ed6ef0ee

barebones numeric suite for quantization with dynamic tracing (#67776)

vkuzo committed 2 years ago

ca499567

Strided masked softmin. (#68463)

pearu committed 2 years ago

d0eff8d8

Older