From 4d8908df272c0a9db2e5fbcc8aaed73cdf75442a Mon Sep 17 00:00:00 2001 From: Fanli Lin Date: Fri, 20 Sep 2024 02:39:19 +0800 Subject: [PATCH] [tests] enable GemmaIntegrationTest on XPU (#33555) enable GemmaIntegrationTest --- tests/models/gemma/test_modeling_gemma.py | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/tests/models/gemma/test_modeling_gemma.py b/tests/models/gemma/test_modeling_gemma.py index b564d51216..a02541d585 100644 --- a/tests/models/gemma/test_modeling_gemma.py +++ b/tests/models/gemma/test_modeling_gemma.py @@ -528,7 +528,7 @@ class GemmaModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixi @slow -@require_torch_gpu +@require_torch_accelerator class GemmaIntegrationTest(unittest.TestCase): input_text = ["Hello I am doing", "Hi today"] # This variable is used to determine which CUDA device are we using for our runners (A10 or T4) @@ -748,7 +748,6 @@ class GemmaIntegrationTest(unittest.TestCase): output = model.generate(**inputs, max_new_tokens=20, do_sample=False) output_text = tokenizer.batch_decode(output, skip_special_tokens=True) - self.assertEqual(output_text, EXPECTED_TEXTS[self.cuda_compute_capability_major_version]) @require_read_token @@ -770,10 +769,8 @@ class GemmaIntegrationTest(unittest.TestCase): tokenizer = AutoTokenizer.from_pretrained(model_id) inputs = tokenizer(self.input_text, return_tensors="pt", padding=True).to(torch_device) - output = model.generate(**inputs, max_new_tokens=20, do_sample=False) output_text = tokenizer.batch_decode(output, skip_special_tokens=True) - self.assertEqual(output_text, EXPECTED_TEXTS) @require_bitsandbytes