chore: fix typos in tests directory (#36785)

* chore: fix typos in tests directory

* chore: fix typos in tests directory

* chore: fix typos in tests directory

* chore: fix typos in tests directory

* chore: fix typos in tests directory

* chore: fix typos in tests directory

* chore: fix typos in tests directory

tests/generation/test_beam_constraints.py

@@ -43,7 +43,7 @@ class ConstraintTest(unittest.TestCase):
             DisjunctiveConstraint([torch.LongTensor([1, 2, 4]), torch.LongTensor([1, 2, 3, 4, 5])])
 
     def test_check_illegal_input(self):
-        # We can't have constraints that are complete subsets of another. This leads to a preverse
+        # We can't have constraints that are complete subsets of another. This leads to a perverse
         # interpretation of "constraint fulfillment": does generating [1,2,3] fulfill the constraint?
         # It would mean that it generated [1,2] which fulfills it, but it's in the middle of potentially
         # fulfilling [1,2,3,4]. If we believe that [1,2,3] does fulfill the constraint, then the algorithm

tests/models/align/test_modeling_align.py

@@ -495,7 +495,7 @@ class AlignModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
     def test_model_get_set_embeddings(self):
         pass
 
-    # override as the `temperature` parameter initilization is different for ALIGN
+    # override as the `temperature` parameter initialization is different for ALIGN
     def test_initialization(self):
         config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
 
@@ -504,7 +504,7 @@ class AlignModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             model = model_class(config=configs_no_init)
             for name, param in model.named_parameters():
                 if param.requires_grad:
-                    # check if `temperature` is initilized as per the original implementation
+                    # check if `temperature` is initialized as per the original implementation
                     if name == "temperature":
                         self.assertAlmostEqual(
                             param.data.item(),

tests/models/altclip/test_modeling_altclip.py

@@ -482,7 +482,7 @@ class AltCLIPModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase)
     def test_model_get_set_embeddings(self):
         pass
 
-    # override as the `logit_scale` parameter initilization is different for AltCLIP
+    # override as the `logit_scale` parameter initialization is different for AltCLIP
     def test_initialization(self):
         config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
         configs_no_init = _config_zero_init(config)
@@ -490,7 +490,7 @@ class AltCLIPModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase)
             model = model_class(config=configs_no_init)
             for name, param in model.named_parameters():
                 if param.requires_grad:
-                    # check if `logit_scale` is initilized as per the original implementation
+                    # check if `logit_scale` is initialized as per the original implementation
                     if name == "logit_scale":
                         self.assertAlmostEqual(
                             param.data.item(),

tests/models/auto/test_processor_auto.py

@@ -186,7 +186,7 @@ class AutoFeatureExtractorTest(unittest.TestCase):
             model_config.save_pretrained(tmpdirname)
             # copy relevant files
             copyfile(SAMPLE_VOCAB, os.path.join(tmpdirname, "vocab.json"))
-            # create emtpy sample processor
+            # create empty sample processor
             with open(os.path.join(tmpdirname, FEATURE_EXTRACTOR_NAME), "w") as f:
                 f.write("{}")
 

tests/pipelines/test_pipelines_common.py

@@ -613,7 +613,7 @@ class PipelineUtilsTest(unittest.TestCase):
         set_seed_fn = lambda: torch.manual_seed(0)  # noqa: E731
         for task in SUPPORTED_TASKS.keys():
             if task == "table-question-answering":
-                # test table in seperate test due to more dependencies
+                # test table in separate test due to more dependencies
                 continue
 
             self.check_default_pipeline(task, "pt", set_seed_fn, self.check_models_equal_pt)
@@ -631,7 +631,7 @@ class PipelineUtilsTest(unittest.TestCase):
         set_seed_fn = lambda: keras.utils.set_random_seed(0)  # noqa: E731
         for task in SUPPORTED_TASKS.keys():
             if task == "table-question-answering":
-                # test table in seperate test due to more dependencies
+                # test table in separate test due to more dependencies
                 continue
 
             self.check_default_pipeline(task, "tf", set_seed_fn, self.check_models_equal_tf)

tests/pipelines/test_pipelines_token_classification.py

@@ -778,7 +778,7 @@ class TokenClassificationPipelineTests(unittest.TestCase):
     @require_tf
     def test_tf_only(self):
         model_name = "hf-internal-testing/tiny-random-bert-tf-only"  # This model only has a TensorFlow version
-        # We test that if we don't specificy framework='tf', it gets detected automatically
+        # We test that if we don't specify framework='tf', it gets detected automatically
         token_classifier = pipeline(task="ner", model=model_name)
         self.assertEqual(token_classifier.framework, "tf")
 

tests/quantization/bnb/README.md

@@ -13,7 +13,7 @@ The following is the recipe on how to effectively debug `bitsandbytes` integrati
 
 The following instructions are tested with 2 NVIDIA-Tesla T4 GPUs. To run successfully `bitsandbytes` you would need a 8-bit core tensor supported GPU. Note that Turing, Ampere or newer architectures - e.g. T4, RTX20s RTX30s, A40-A100, A6000 should be supported. 
 
-## Virutal envs
+## Virtual envs
 
 ```bash
 conda create --name int8-testing python==3.8
@@ -61,7 +61,7 @@ This happens when some Linear weights are set to the CPU when using `accelerate`
 
 Use the latest version of `accelerate` with a command such as: `pip install -U accelerate` and the problem should be solved.
 
-### `Parameter has no attribue .CB` 
+### `Parameter has no attribute .CB` 
 
 Same solution as above.
 
@@ -71,7 +71,7 @@ Run your script by pre-pending `CUDA_LAUNCH_BLOCKING=1` and you should observe a
 
 ### `CUDA illegal memory error: an illegal memory access at line...`:
 
-Check the CUDA verisons with:
+Check the CUDA versions with:
 ```bash
 nvcc --version
 ```

tests/quantization/bnb/test_4bit.py

@@ -179,7 +179,7 @@ class Bnb4BitTest(Base4bitTest):
 
     def test_original_dtype(self):
         r"""
-        A simple test to check if the model succesfully stores the original dtype
+        A simple test to check if the model successfully stores the original dtype
         """
         self.assertTrue(hasattr(self.model_4bit.config, "_pre_quantization_dtype"))
         self.assertFalse(hasattr(self.model_fp16.config, "_pre_quantization_dtype"))
@@ -496,8 +496,8 @@ class Pipeline4BitTest(Base4bitTest):
     def test_pipeline(self):
         r"""
         The aim of this test is to verify that the mixed 4bit is compatible with `pipeline` from transformers. Since
-        we used pipline for inference speed benchmarking we want to make sure that this feature does not break anything
+        we used pipeline for inference speed benchmarking we want to make sure that this feature does not break anything
-        on pipline.
+        on pipeline.
         """
         # self._clear_cuda_cache()
         self.pipe = pipeline(

tests/quantization/bnb/test_mixed_int8.py

@@ -213,7 +213,7 @@ class MixedInt8Test(BaseMixedInt8Test):
 
     def test_original_dtype(self):
         r"""
-        A simple test to check if the model succesfully stores the original dtype
+        A simple test to check if the model successfully stores the original dtype
         """
         self.assertTrue(hasattr(self.model_8bit.config, "_pre_quantization_dtype"))
         self.assertFalse(hasattr(self.model_fp16.config, "_pre_quantization_dtype"))
@@ -655,8 +655,8 @@ class MixedInt8TestPipeline(BaseMixedInt8Test):
     def test_pipeline(self):
         r"""
         The aim of this test is to verify that the mixed int8 is compatible with `pipeline` from transformers. Since
-        we used pipline for inference speed benchmarking we want to make sure that this feature does not break anything
+        we used pipeline for inference speed benchmarking we want to make sure that this feature does not break anything
-        on pipline.
+        on pipeline.
         """
         # self._clear_cuda_cache()
         self.pipe = pipeline(

tests/quantization/gptq/test_gptq.py

@@ -167,7 +167,7 @@ class GPTQTest(unittest.TestCase):
 
     def test_original_dtype(self):
         r"""
-        A simple test to check if the model succesfully stores the original dtype
+        A simple test to check if the model successfully stores the original dtype
         """
         self.assertTrue(hasattr(self.quantized_model.config, "_pre_quantization_dtype"))
         self.assertFalse(hasattr(self.model_fp16.config, "_pre_quantization_dtype"))
@@ -261,7 +261,7 @@ class GPTQTest(unittest.TestCase):
                 if self.device_map == "cpu":
                     quant_type = "ipex" if is_ipex_available() else "torch"
                 else:
-                    # We expecte tritonv2 to be used here, because exllama backend doesn't support packing https://github.com/ModelCloud/GPTQModel/issues/1354
+                    # We expect tritonv2 to be used here, because exllama backend doesn't support packing https://github.com/ModelCloud/GPTQModel/issues/1354
                     # TODO: Remove this once GPTQModel exllama kernels supports packing
                     quant_type = "tritonv2"
                 quantized_model_from_saved = AutoModelForCausalLM.from_pretrained(
@@ -433,7 +433,7 @@ class GPTQTestExllamaV2(unittest.TestCase):
                 "exllamav2",
             )
         else:
-            # We expecte tritonv2 to be used here, because exllama backend doesn't support packing https://github.com/ModelCloud/GPTQModel/issues/1354
+            # We expect tritonv2 to be used here, because exllama backend doesn't support packing https://github.com/ModelCloud/GPTQModel/issues/1354
             # TODO: Remove this once GPTQModel exllama kernels supports packing
             self.assertEqual(
                 self.quantized_model.model.layers[0].self_attn.k_proj.QUANT_TYPE,
@@ -458,7 +458,7 @@ class GPTQTestExllamaV2(unittest.TestCase):
 
     def test_generate_quality(self):
         """
-        Simple test to check the quality of the model by comapring the the generated tokens with the expected tokens
+        Simple test to check the quality of the model by comparing the the generated tokens with the expected tokens
         """
         self.check_inference_correctness(self.quantized_model)
 

tests/quantization/higgs/test_higgs.py

@@ -184,7 +184,7 @@ class HiggsTest(unittest.TestCase):
             output = model.generate(**input_ids, max_new_tokens=self.max_new_tokens)
             self.assertEqual(self.tokenizer.decode(output[0], skip_special_tokens=True), self.EXPECTED_OUTPUT)
 
-    @unittest.skip("This will almost surely OOM. Enable when swithed to a smaller model")
+    @unittest.skip("This will almost surely OOM. Enable when switched to a smaller model")
     def test_dequantize(self):
         """
         Test the ability to dequantize a model

tests/quantization/torchao_integration/test_torchao.py

@@ -202,7 +202,7 @@ class TorchAoGPUTest(TorchAoTest):
 
     def test_int4wo_offload(self):
         """
-        Simple test that checks if the quantized model int4 wieght only is working properly with cpu/disk offload
+        Simple test that checks if the quantized model int4 weight only is working properly with cpu/disk offload
         """
 
         device_map_offload = {
@@ -254,7 +254,7 @@ class TorchAoGPUTest(TorchAoTest):
     @require_torch_multi_gpu
     def test_int4wo_quant_multi_gpu(self):
         """
-        Simple test that checks if the quantized model int4 wieght only is working properly with multiple GPUs
+        Simple test that checks if the quantized model int4 weight only is working properly with multiple GPUs
         set CUDA_VISIBLE_DEVICES=0,1 if you have more than 2 GPUS
         """