pytorch/benchmark

Pull Requests Commits

Add mirage superoptimize

xuzhao9 committed 1 year ago

1f234771

Implement gated mlp

xuzhao9 committed 1 year ago

e2f1d022

Add mirage submodule

xuzhao9 committed 1 year ago

a7e2de92

add rmsnorm_linear

xuzhao9 committed 1 year ago

6d25de8a

Make sure all ci-enabled impls are in the output

xuzhao9 committed 1 year ago

00c9b9ed

Test backward pass in unit test.

xuzhao9 committed 1 year ago

eec86128

Fix the broken gemm test

xuzhao9 committed 1 year ago

384a43d8

Add doc for adding custom ops (#2509)

FindHao committed 1 year ago

d933cedc

jasonjk-park committed 1 year ago

2feadb6a

Add --op-collection option (#2503)

FindHao committed 1 year ago

58f3b1f3

Update 2.5.0.yaml (#2498)

juliagmt-google committed 1 year ago

9e670cd2

Change default gpu metric backend (#2501)

FindHao committed 1 year ago

c3961913

Add ncu report analyzer (#2497)

FindHao committed 1 year ago

21cc30dc

Add host-side Triton TMA support to Dynamo (#137677)

aakhundov committed 1 year ago

db41e776

Update the flash-attention submodule (#2500)

xuzhao9 committed 1 year ago

ea4433fd

Add fbscribelogger to Dynamo benchmark runner (#137867)

ezyang committed 1 year ago

12e1d263

sakipping null values in scribe message

adamomainz committed 1 year ago

509e94f0

Add entire _dynamo.config as a json for logging (#137216)

jovianjaison committed 1 year ago

7cb1c0ac

making CI more flexible for extra data in tritonbench

adamomainz committed 1 year ago

28d301a4

change GPT2ForSequenceClassification inference accuracy tolerance (#136749)

Valentine233 committed 1 year ago

680d64ea

Add AtenOp Benchmarking (#2495)

FindHao committed 1 year ago

34d4f94d

Format `.ci/` / `.github/` / `benchmarks/` / `functorch/` / `tools/` / `torchgen/` with `ruff format` (#132577)

XuehaiPan committed 1 year ago

f9f52f64

differentiating between some Fbsource only targets and OSS for CI

adamomainz committed 1 year ago

3900904e

use 3.13 multiline traceback in get_instruction_source_311 (#137617)

williamwen42 committed 1 year ago

f3921ca7

fixing typo in fp8_gemm

adamomainz committed 1 year ago

b56e2eed

combining CI and servicelab configs

adamomainz committed 1 year ago

3a7a4fea

Check dyno and dcgm existence before disable them (#2496)

FindHao committed 1 year ago

dcd3d319

Improve release benchmark suites with a lower value of epoch (#2482)

juliagmt-google committed 1 year ago

7742ef2f

adding new configs for servicelab

adamomainz committed 1 year ago

533d2588

Handle torch function subclass/mode dispatch on generic tensor methods (#137119)

mlazos committed 1 year ago

4fd7c743

Older