pytorch/benchmark

Pull Requests Commits

yolov3: reduce batch size due to OOM

xmfan committed 2 years ago

c5fb66e1

move tf override tensor to torch_function.py (#111714)

mlazos committed 2 years ago

b5631ed9

Don't suppress original error message for data-dependent value (#111596)

ezyang committed 2 years ago

5a391f5c

Add `is_` support for `Tensor`s, force `get_fake_value` to reuse previously computed `example_value` if available (#111565)

jon-chuang committed 2 years ago

ea31e729

Refactor TensorWithTFOverrideVariable (#109556)

mlazos committed 2 years ago

a80732ca

Attach cause to some exceptions and enable RUFF TRY200 (#111496)

Skylion007 committed 2 years ago

613a950a

Fix JSON file format Torch-TRT (#1998)

gs-olive committed 2 years ago

f953f810

Benchmark to store test data along exported model (#111095)

BowenBao committed 2 years ago

8339eef6

replace numpy.bool with bool 3/N

igorsugak committed 2 years ago

e5bfd3dd

Add dynamo_onnx_aot_inline to bench (#110183)

BowenBao committed 2 years ago

5248206b

fix a max-autotune rng state related bug (#111381)

shunting314 committed 2 years ago

ff41192e

Fix 'stable_diffusion_xl' (#1994)

BowenBao committed 2 years ago

5f1ea34c

make disable_cahce_limit also disable accumulated cache limit (#111334)

ydwu4 committed 2 years ago

7617d3f5

Add phi_1_5 hugging face model (#1982)

Pranav Sharma committed 2 years ago

4dbfb75b

Build the nightly docker (#1985)

xuzhao9 committed 2 years ago

eb1adce5

patch the pt2_compilation_time property

xuzhao9 committed 2 years ago

0d7b6afc

Add ttfb to available metrics

xuzhao9 committed 2 years ago

7dc8e811

update doctr to commit 56c8356 to fix BF16 mode (#1979)

chunyuan-w committed 2 years ago

c744f0c9

Update `metadata.yaml` for `squeezenet1_1` (#1980)

hmellor committed 2 years ago

09c938cd

Revert "Reland 3rd try [finishing colesbury's PR 100642] Guard on nn.Module dicts and type (#109323)" + Forward fixes + test (#110964)

voznesenskym committed 2 years ago

31a15827

Fix & update `metadata.yaml` for hf_T5` variants and `timm_nfnet` (#1976)

hmellor committed 2 years ago

7eb245a1

Add new model: simple_gpt_tp_manual (#1969)

xmfan committed 2 years ago

cdd87f0d

Simplify the multiple batches design by removing the "train_dynamic" test

xuzhao9 committed 2 years ago

fd7f14e1

Revert "compulsory case names, allow multiple (#110733)" (#110783)

huydhn committed 2 years ago

d4f79d9c

Broader error catching in Torch-TRT `userbenchmark` (#1974)

gs-olive committed 2 years ago

eb8f9526

Fix benchmarks with self.autocast (#110490)

angelayi committed 2 years ago

cbcc103b

compulsory case names, allow multiple (#110733)

avikchaudhuri committed 2 years ago

ed0f6baa

add tb test for ig-ctr

dshi7 committed 2 years ago

99d1e7a7

Upgrade pin numpy version (#1966)

xuzhao9 committed 2 years ago

8d7fa97f

Switch Torch-TRT batch size to None (#1967)

gs-olive committed 2 years ago

5e8c08fb

Older