Commitbed2fa7

committed

reduce tests

Signed-off-by: Yibin Li <109242046+yibinl-nvidia@users.noreply.github.com>

1 parentb78bc41 commitbed2fa7Copy full SHA for bed2fa7

File tree

1 file changed

-93

lines changed

tests/unittest/_torch/executor
- test_chunked_logits.py

1 file changed

-93

lines changed

`‎tests/unittest/_torch/executor/test_chunked_logits.py‎`

Lines changed: 1 addition & 93 deletions

Original file line number	Diff line number	Diff line change
`@@ -20,52 +20,9 @@ def sample_logits():`
`20`	`20`	`returntorch.randn(1,1,1000,device='cuda')`
`21`	`21`
`22`	`22`
`23`		`-@pytest.fixture`
`24`		`-defchunked_request():`
`25`		`-"""Create LlmRequest with chunked logits enabled"""`
`26`		`-returnLlmRequest(request_id=100,`
`27`		`-max_new_tokens=10,`
`28`		`-input_tokens=[1,2,3],`
`29`		`-sampling_config=SamplingConfig(),`
`30`		`-is_streaming=False,`
`31`		`-return_generation_logits=True,`
`32`		`-use_chunked_generation_logits=True,`
`33`		`-logits_chunk_size=4)`
`34`		`-`
`35`		`-`
`36`		`-@pytest.fixture`
`37`		`-defnon_chunked_request():`
`38`		`-"""Create LlmRequest with chunked logits disabled"""`
`39`		`-returnLlmRequest(request_id=101,`
`40`		`-max_new_tokens=10,`
`41`		`-input_tokens=[1,2,3],`
`42`		`-sampling_config=SamplingConfig(),`
`43`		`-is_streaming=False,`
`44`		`-return_generation_logits=True,`
`45`		`-use_chunked_generation_logits=False)`
`46`		`-`
`47`		`-`
`48`	`23`	`classTestLogitsStorage:`
`49`	`24`	`"""Unit tests for LogitsStorage class"""`
`50`	`25`
`51`		`-deftest_initialization(self):`
`52`		`-"""Test LogitsStorage initialization with different parameters"""`
`53`		`-# Test basic initialization`
`54`		`-storage=LogitsStorage(seq_length=10,`
`55`		`-use_device_memory=True,`
`56`		`-should_exclude_last=False,`
`57`		`-use_chunked_generation_logits=False,`
`58`		`-chunk_size=8)`
`59`		`-`
`60`		`-assertstorage.seq_length==10`
`61`		`-assertstorage.use_device_memoryisTrue`
`62`		`-assertstorage._should_exclude_lastisFalse`
`63`		`-assertstorage.use_chunked_generation_logitsisFalse`
`64`		`-assertstorage.chunk_size==8`
`65`		`-assertstorage._logits_indices== []`
`66`		`-assertstorage.beam_width==-1`
`67`		`-assertstorage.vocab_size==-1`
`68`		`-`
`69`	`26`	`deftest_initialization_chunked_mode(self):`
`70`	`27`	`"""Test LogitsStorage initialization in chunked mode"""`
`71`	`28`	`storage=LogitsStorage(seq_length=10,`
`@@ -79,23 +36,6 @@ def test_initialization_chunked_mode(self):`
`79`	`36`	`assertstorage._device_fragments== []`
`80`	`37`	`assertstorage._current_position==0`
`81`	`38`
`82`		`-deftest_append_3d_logits(self,sample_logits):`
`83`		`-"""Test appending 3D logits"""`
`84`		`-storage=LogitsStorage(seq_length=10,`
`85`		`-use_chunked_generation_logits=False)`
`86`		`-storage.append(sample_logits)`
`87`		`-`
`88`		`-assertstorage.beam_width==1`
`89`		`-assertstorage.vocab_size==1000`
`90`		`-`
`91`		`-deftest_append_invalid_shape(self):`
`92`		`-"""Test appending logits with invalid shape"""`
`93`		`-storage=LogitsStorage(seq_length=10,`
`94`		`-use_chunked_generation_logits=False)`
`95`		`-`
`96`		`-withpytest.raises(AssertionError):`
`97`		`-storage.append(torch.randn(1000))# 1D - should fail`
`98`		`-`
`99`	`39`	`deftest_append_chunked_mode_streaming(self,sample_logits):`
`100`	`40`	`"""Test append behavior in chunked streaming mode"""`
`101`	`41`	`storage=LogitsStorage(seq_length=10,`
`@@ -145,17 +85,6 @@ def test_finalize_chunked_transfer_non_chunked_mode(self):`
`145`	`85`	`# Should not raise any errors`
`146`	`86`	`storage.finalize_chunked_transfer()`
`147`	`87`
`148`		`-deftest_storage_overflow(self,sample_logits):`
`149`		`-"""Test storage overflow handling"""`
`150`		`-storage=LogitsStorage(seq_length=2,`
`151`		`-use_chunked_generation_logits=False)`
`152`		`-storage.append(sample_logits)`
`153`		`-storage.append(sample_logits)`
`154`		`-`
`155`		`-# This should cause overflow`
`156`		`-withpytest.raises(ValueError,match="LogitsStorage overflow"):`
`157`		`-storage.append(sample_logits)`
`158`		`-`
`159`	`88`
`160`	`89`	`classTestPyResult:`
`161`	`90`	`"""Unit tests for PyResult class"""`
`@@ -202,27 +131,6 @@ def test_transfer_remaining_device_logits(self, sample_logits):`
`202`	`131`
`203`	`132`	`# Should not raise errors`
`204`	`133`
`205`		`-deftest_context_generation_logits_property(self,sample_logits):`
`206`		`-"""Test context_logits property"""`
`207`		`-result=PyResult(prompt_len=5,`
`208`		`-max_new_tokens=10,`
`209`		`-return_context_logits=True,`
`210`		`-return_generation_logits=True,`
`211`		`-use_chunked_generation_logits=False)`
`212`		`-`
`213`		`-result.append_context_logits(sample_logits)`
`214`		`-context_logits=result.context_logits`
`215`		`-`
`216`		`-assertcontext_logitsisnotNone`
`217`		`-assertcontext_logits.shape== (1,1000)# Should remove beam dimension`
`218`		`-`
`219`		`-result.append_generation_logits(sample_logits)`
`220`		`-generation_logits=result.generation_logits`
`221`		`-`
`222`		`-assertgeneration_logitsisnotNone`
`223`		`-assertgeneration_logits.shape== (1,1,1000`
`224`		`- )# Should transpose dimensions`
`225`		`-`
`226`	`134`
`227`	`135`	`classTestLlmRequest:`
`228`	`136`	`"""Unit tests for LlmRequest class"""`
`@@ -251,7 +159,7 @@ def test_initialization_chunked_logits(self):`
`251`	`159`	`assertrequest_streaming.py_logits_chunk_size==1# 1 in streaming mode`
`252`	`160`
`253`	`161`
`254`		`-classTestChunkedLogitsIntegration:`
	`162`	`+classTestChunkedLogitsComplicated:`
`255`	`163`	`"""Integration tests for chunked logits functionality"""`
`256`	`164`
`257`	`165`	`deftest_chunked_vs_non_chunked_equivalence(self,sample_logits):`

0 commit comments

Comments

(0)

Movatterモバイル変換

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Commitbed2fa7

File tree

1 file changed

1 file changed

`‎tests/unittest/_torch/executor/test_chunked_logits.py‎`

0 commit comments