pytorch/benchmark

Pull Requests Commits

ignore mark_dynamic() in export (#2451)

pianpwk committed 2 years ago

6558f9e4

Implement bf16xint16 kernel (#2349)

davidberard98 committed 2 years ago

1e6003dd

tritonbench bf16xint16 matmul template (#2348)

davidberard98 committed 2 years ago

ebd00aac

Fix dynamo benchmark skip logic for cpu device (#135193)

zxd1997066 committed 2 years ago

8351474b

Turn on TMA by default for row-wise GEMM (#2450)

htyu committed 2 years ago

e95390a1

reland map/zip iterator related changes (#135074)

williamwen42 committed 2 years ago

5ee35999

Retire CompileProfiler (#135133)

anijain2305 committed 2 years ago

cdceb782

adding buck file for triton-lang/kernels

Adam Mainz committed 2 years ago

ac0cf987

adding arc lint opt in for benchmarking (#2448)

Adam Mainz committed 2 years ago

7bb00692

Support for proxying frozen dataclasses (#134846)

mlazos committed 2 years ago

8c5abe8e

Add tma option for FB8 row-wise GEMM.

htyu committed 2 years ago

71b6e37d

Stop adding useless prefix to error message here, you're pushing the important info off the screen. (#133108)

ezyang committed 2 years ago

a95718f5

Add to torchbench relative path search (#134871)

yiming0416 committed 2 years ago

2927a2b8

adding llama inference specific shapes for flash attention

Adam Mainz committed 2 years ago

fdb33106

move loop ordering after fusion (#126254)

shunting314 committed 2 years ago

49bcc381

Switch benchmarking to use export non-strict mode (#130977)

desertfire committed 2 years ago

1f4461fb

Enable flash_v3 backward (#2445)

bertmaher committed 2 years ago

e6251f18

Add FlexAttention (#2443)

bertmaher committed 2 years ago

c0409aaf

Format flash_attention with black (#2444)

bertmaher committed 2 years ago

31d07c69

Add compile time instruction count metric (#133834)

laithsakka committed 2 years ago

d8735143

fix test_functional_call_sequential_params_and_buffers expectation on Windows (#134394)

xuhancn committed 2 years ago

ae902e4c

fixing OSS ci for tritonbench (#2439)

Adam Mainz committed 2 years ago

b0ae45a8

PR#4179 (#2435)

Elliot Gorokhovsky committed 2 years ago

babb1285

Install the benchmark before running. (#2436)

xuzhao9 committed 2 years ago

52103b5e

Enable cublas for FP8 rowwise

htyu committed 2 years ago

ef3ce35e

typing for decorators - fx/_compatibility (part 1) (#134202)

aorenste committed 2 years ago

f4f13306

Always emit end events even on failure, use thread local storage for stack (#2432)

jamesjwu committed 2 years ago

0968f5ed

fix PR #2428 (#2433)

sfzhu93 committed 2 years ago

ea7a2d35

Collect Export pass rate separately (#134076)

desertfire committed 2 years ago

41cf189a

add support for auto-tune TMA grid constant (#2428)

sfzhu93 committed 2 years ago

fa08aeb5

Older