Update to use ruff format

.lintrunner.toml

@@ -62,7 +62,7 @@ init_command = [
 is_formatter = true
 
 [[linter]]
-code = 'BLACK-ISORT'
+code = 'RUFF-FORMAT'
 include_patterns = [
   '**/*.py'
 ]
@@ -74,7 +74,7 @@ command = [
   '-m',
   'lintrunner_adapters',
   'run',
-  'black_isort_linter',
+  'ruff_format_linter',
   '--',
   '@{{PATHSFILE}}'
 ]

.pre-commit-config.yaml

@@ -7,29 +7,11 @@ repos:
         args: [--markdown-linebreak-ext=md]
       - id: check-yaml
       - id: requirements-txt-fixer
-
-  - repo: https://github.com/psf/black
-    rev: 24.1.0
-    hooks:
-      - id: black
-        name: Format code
-        exclude: "hadamard_utils.py"
-  - repo: https://github.com/pycqa/isort
-    rev: 5.11.5
-    hooks:
-      - id: isort
-        name: Format imports
   - repo: https://github.com/MarcoGorelli/absolufy-imports
     rev: v0.3.1
     hooks:
       - id: absolufy-imports
         exclude: examples/
-  - repo: https://github.com/astral-sh/ruff-pre-commit
-    # Ruff version.
-    rev: v0.6.0
-    hooks:
-      - id: ruff
-        args: [ --fix ]
   - repo: local
     hooks:
       - id: format-json

docs/source/exts/auto_config_doc/__init__.py

@@ -73,9 +73,9 @@ def run(self):
         (class_name,) = self.arguments
         package_config = OlivePackageConfig.load_default_config()
         auto_config_class = import_class(class_name, package_config)
-        assert issubclass(auto_config_class, AutoConfigClass) or issubclass(
-            auto_config_class, Pass
-        ), f"{class_name} is not a subclass of AutoConfigClass or Pass"
+        assert issubclass(auto_config_class, AutoConfigClass) or issubclass(auto_config_class, Pass), (
+            f"{class_name} is not a subclass of AutoConfigClass or Pass"
+        )
 
         node = nodes.section()
         node.document = self.state.document

docs/source/exts/gallery_directive.py

@@ -107,7 +107,7 @@ def run(self) -> Optional[List[nodes.Node]]:
 
         # Parse the template with Sphinx Design to create an output container
         # Prep the options for the template grid
-        class_ = "gallery-directive" + f' {self.options.get("class-container", "")}'
+        class_ = "gallery-directive" + f" {self.options.get('class-container', '')}"
         options = {"gutter": 2, "class-container": class_}
         options_str = "\n".join(f":{k}: {v}" for k, v in options.items())
 

examples/directml/llm/chat_app/interface/hddr_llm_onnx_dml_interface.py

@@ -288,10 +288,13 @@ def predict(
                         sentence = sentence[: sentence.index(ai_token)].strip()
                         break
                 sentence = sentence.strip()
-                a, b = [[y[0], convert_to_markdown(y[1])] for y in history] + [[text, convert_to_markdown(sentence)]], [
-                    *history,
-                    [text, sentence],
-                ]
+                a, b = (
+                    [[y[0], convert_to_markdown(y[1])] for y in history] + [[text, convert_to_markdown(sentence)]],
+                    [
+                        *history,
+                        [text, sentence],
+                    ],
+                )
                 yield a, b, "Generating..."
 
             if shared_state.interrupted:

examples/inception/prepare_config.py

@@ -10,7 +10,6 @@
 
 
 def resolve_windows_config():
-
     with Path("inception_config.json").open() as f:
         snpe_windows_config = json.load(f)
 

examples/llama2/tensor_parallel_generate.py

@@ -42,8 +42,8 @@
 torch.cuda.empty_cache()
 
 model_id = "meta-llama/Llama-2-7b-hf"
-model_path = "models/tensor_parallel/tensor_parallel-conversion-transformers_optimization_fp16/gpu-cuda_model/model_{:02d}".format(  # noqa: E501
-    rank
+model_path = (
+    f"models/tensor_parallel/tensor_parallel-conversion-transformers_optimization_fp16/gpu-cuda_model/model_{rank:02d}"
 )
 prompt = "What is an apple?"
 # prompt = "Is it normal to have a dark ring around the iris of my eye?"

examples/open_llama/user_script.py

@@ -54,10 +54,13 @@ def __iter__(self):
                 inp = trainenc["input_ids"][i:j].unsqueeze(0)
                 mask = torch.ones(inp.shape)
                 if self.sess is None:
-                    yield {
-                        "input_ids": inp.detach().cpu().numpy().astype("int64"),
-                        "attention_mask": mask.detach().cpu().numpy().astype("int64"),
-                    }, 0
+                    yield (
+                        {
+                            "input_ids": inp.detach().cpu().numpy().astype("int64"),
+                            "attention_mask": mask.detach().cpu().numpy().astype("int64"),
+                        },
+                        0,
+                    )
                 else:
                     outputs = self.sess.run(
                         None,

examples/phi2/generate.py

@@ -2,7 +2,7 @@
 # Copyright (c) Microsoft Corporation. All rights reserved.
 # Licensed under the MIT License.
 # --------------------------------------------------------------------------
-# copied from https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/python/tools/transformers/models/phi2/inference_example.py  # noqa: E501
+# copied from https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/python/tools/transformers/models/phi2/inference_example.py
 import numpy as np
 import onnxruntime as ort
 import torch
@@ -40,7 +40,6 @@ def __init__(self, decoder_path):
         self.tokenizer = None
 
     def get_initial_inputs_and_outputs(self, encodings_dict):
-
         input_ids = torch.tensor(encodings_dict["input_ids"], device=self.device, dtype=torch.int32)
         attention_mask = torch.tensor(encodings_dict["attention_mask"], device=self.device, dtype=torch.int32)
         step = torch.tensor([0], device=self.device, dtype=torch.int64)
@@ -278,7 +277,7 @@ def genai_run(prompts, model_path, max_length=200):
     app_started_timestamp = time.time()
     model = og.Model(model_path)
     model_loaded_timestamp = time.time()
-    print("Model loaded in {:.2f} seconds".format(model_loaded_timestamp - app_started_timestamp))
+    print(f"Model loaded in {model_loaded_timestamp - app_started_timestamp:.2f} seconds")
     tokenizer = og.Tokenizer(model)
 
     input_tokens = tokenizer.encode_batch(prompts)
@@ -299,7 +298,7 @@ def genai_run(prompts, model_path, max_length=200):
     output_token_count = 0
 
     for i, prompt in enumerate(prompts):
-        print(f"Prompt #{i+1:02d}: {prompt}")
+        print(f"Prompt #{i + 1:02d}: {prompt}")
         print(tokenizer.decode(generator.get_sequence(i)))
 
     output_token_count = sum(len(generator.get_sequence(i)) for i in range(len(prompts)))

examples/phi3/phi3.py

@@ -199,7 +199,6 @@ def use_passes(template_json, *passes):
 
 
 def generate_config(args):
-
     json_file_template = "phi3_template.json"
     with open(json_file_template) as f:
         template_json = json.load(f)
@@ -261,14 +260,13 @@ def generate_config(args):
 
 
 def genai_run(prompt, model_path, max_length):
-
     print("\nModel inference starts...")
 
     print("Loading model...")
     app_started_timestamp = time.time()
     model = og.Model(model_path)
     model_loaded_timestamp = time.time()
-    print("Model loaded in {:.2f} seconds".format(model_loaded_timestamp - app_started_timestamp))
+    print(f"Model loaded in {model_loaded_timestamp - app_started_timestamp:.2f} seconds")
 
     print("Creating tokenizer...")
     tokenizer = og.Tokenizer(model)
@@ -320,7 +318,7 @@ def genai_run(prompt, model_path, max_length):
             "\n\n"
             f"Prompt tokens: {len(input_tokens)}, New tokens: {len(new_tokens)},"
             f" Time to first: {(first_token_timestamp - started_timestamp):.2f}s,"
-            f" New tokens per second: {len(new_tokens)/run_time:.2f} tps"
+            f" New tokens per second: {len(new_tokens) / run_time:.2f} tps"
         )
 
 

examples/red_pajama/user_script.py

@@ -7,6 +7,6 @@
 MIN_TRANSFORMERS_VERSION = "4.30.2"
 
 # check transformers version
-assert (
-    transformers.__version__ >= MIN_TRANSFORMERS_VERSION
-), f"Please upgrade transformers to version {MIN_TRANSFORMERS_VERSION} or higher."
+assert transformers.__version__ >= MIN_TRANSFORMERS_VERSION, (
+    f"Please upgrade transformers to version {MIN_TRANSFORMERS_VERSION} or higher."
+)

examples/resnet/prepare_model_data.py

@@ -105,11 +105,7 @@ def prepare_model(num_epochs=1, models_dir="models", data_dir="data"):
             loss.backward()
             optimizer.step()
             if (i + 1) % 100 == 0:
-                print(
-                    "Epoch [{}/{}], Step [{}/{}] Loss: {:.4f}".format(
-                        epoch + 1, num_epochs, i + 1, total_step, loss.item()
-                    )
-                )
+                print(f"Epoch [{epoch + 1}/{num_epochs}], Step [{i + 1}/{total_step}] Loss: {loss.item():.4f}")
         # Decay learning rate
         if (epoch + 1) % 20 == 0:
             curr_lr /= 3
@@ -129,7 +125,7 @@ def prepare_model(num_epochs=1, models_dir="models", data_dir="data"):
                 total += labels.size(0)
                 correct += (predicted == labels).sum().item()
 
-            print("Accuracy of the model on the test images: {} %".format(100 * correct / total))
+            print(f"Accuracy of the model on the test images: {100 * correct / total} %")
 
     # Save the model
     model.to("cpu")

examples/utils/generator.py

@@ -428,9 +428,9 @@ def get_initial_inputs(
                 else [apply_template(template, p) for p in prompt]
             )
         else:
-            assert isinstance(prompt, str) or (
-                isinstance(prompt, list) and all(isinstance(p, str) for p in prompt)
-            ), "tuple prompts require a template"
+            assert isinstance(prompt, str) or (isinstance(prompt, list) and all(isinstance(p, str) for p in prompt)), (
+                "tuple prompts require a template"
+            )
 
         encodings_dict = self.tokenizer(prompt, return_tensors="np", padding=True)
         input_ids = encodings_dict["input_ids"].astype(self.input_info["input_ids"]["dtype"])

examples/utils/kv_cache_utils.py

@@ -160,18 +160,18 @@ def update(self, present_kvs: List["NDArray"]):
         assert present_len > 0, "present_kvs cannot be empty"
 
         if self.seen_len == 0:
-            assert (
-                present_len <= self.max_cache_len
-            ), "present_kvs is longer than max_cache_len during prompt processing"
+            assert present_len <= self.max_cache_len, (
+                "present_kvs is longer than max_cache_len during prompt processing"
+            )
             # prompt processing
             for k, v in zip(self.past_names, present_kvs):
                 self.cache[k][:, :, :present_len] = v
             self.seen_len = present_len
             return
 
-        assert (
-            present_len == self.max_cache_len + 1
-        ), "present_kvs must be one step longer than max_cache_len in token generation"
+        assert present_len == self.max_cache_len + 1, (
+            "present_kvs must be one step longer than max_cache_len in token generation"
+        )
         for k, v in zip(self.past_names, present_kvs):
             self.cache[k][:, :, self.seen_len] = v[:, :, -1]
         self.seen_len += 1
@@ -414,9 +414,9 @@ def update(self, present_kvs: List[OrtValue]):
             # prompt processing
             present_len = present_kvs[0].shape()[2]
             assert present_len > 0, "present_kvs cannot be empty"
-            assert (
-                present_len <= self.max_cache_len
-            ), "present_kvs is longer than max_cache_len during prompt processing"
+            assert present_len <= self.max_cache_len, (
+                "present_kvs is longer than max_cache_len during prompt processing"
+            )
             for k, v in zip(self.past_names, present_kvs):
                 if self.backend == "torch":
                     import torch
@@ -431,9 +431,9 @@ def update(self, present_kvs: List[OrtValue]):
 
         # token generation
         for past_k, present_k, v in zip(self.past_names, self.present_names, present_kvs):
-            assert (
-                self.output_cache[present_k].data_ptr() == v.data_ptr()
-            ), "out cache ortvalue should be same as present ortvalue"
+            assert self.output_cache[present_k].data_ptr() == v.data_ptr(), (
+                "out cache ortvalue should be same as present ortvalue"
+            )
             if self.backend == "torch":
                 self.cache[past_k][:, :, self.seen_len] = self.output_cache[present_k][:, :, -1]
             else:

examples/vgg/prepare_config.py

@@ -10,7 +10,6 @@
 
 
 def resolve_windows_config():
-
     with Path("vgg_config.json").open() as f:
         snpe_windows_config = json.load(f)
 

examples/vit/val_tiny_imagenet/val_tiny_imagenet.py

@@ -99,7 +99,7 @@ def evaluate_onnx_model(session, dataloader):
         ground_truth = val_idx_to_name[label[0]]
         pred_label = vit_id2label["id2label"][str(top1_pred)]
 
-        print(f"Image {i+1}: {img_name[0]}")
+        print(f"Image {i + 1}: {img_name[0]}")
         print(f"  Ground Truth: {ground_truth}")
         print(f"  Top-1 Prediction: {pred_label}")
         print(f"  Top-5 Predictions: {[vit_id2label['id2label'][str(pred)] for pred in top5_preds]}\n")

olive/auto_optimizer/regulate_mixins.py

@@ -69,9 +69,9 @@ def _regulate_precision(self, pass_config, pass_flows):
 
         is_cuda_ep = self.accelerator_spec.execution_provider != "TensorrtExecutionProvider"
         is_trt_ep = self.accelerator_spec.execution_provider == "TensorrtExecutionProvider"
-        assert (
-            not is_cuda_ep or not is_trt_ep
-        ), "can not support CUDA/DmlExecutionProvider and TensorrtExecutionProvider at the same time"
+        assert not is_cuda_ep or not is_trt_ep, (
+            "can not support CUDA/DmlExecutionProvider and TensorrtExecutionProvider at the same time"
+        )
 
         customized_fp16 = self._allow_precision("fp16")
         cuda_fp16 = customized_fp16 and is_cuda_ep

olive/cli/auto_opt.py

@@ -27,7 +27,6 @@
 
 
 class AutoOptCommand(BaseOliveCLICommand):
-
     @staticmethod
     def register_subcommand(parser: ArgumentParser):
         sub_parser = parser.add_parser(

olive/cli/generate_adapter.py

@@ -49,9 +49,9 @@ def run(self):
 
     def _get_run_config(self, tempdir: str) -> Dict:
         input_model_config = get_input_model_config(self.args)
-        assert (
-            input_model_config["type"].lower() == "onnxmodel"
-        ), "Only ONNX models are supported in generate-adapter command."
+        assert input_model_config["type"].lower() == "onnxmodel", (
+            "Only ONNX models are supported in generate-adapter command."
+        )
 
         to_replace = [
             ("input_model", input_model_config),

olive/cli/quantize.py

@@ -31,7 +31,6 @@
 
 
 class QuantizeCommand(BaseOliveCLICommand):
-
     @staticmethod
     def register_subcommand(parser: ArgumentParser):
         sub_parser = parser.add_parser(
@@ -130,8 +129,7 @@ def _get_pass_list(self, precision, algo, impl, is_hf_model):
 
         if not pass_list:
             raise ValueError(
-                f"Quantiation for precision {precision}, algorithm {algo} "
-                f"and implementation {impl} is not supported"
+                f"Quantiation for precision {precision}, algorithm {algo} and implementation {impl} is not supported"
             )
         logger.info("pass list: %s", pass_list)
         return pass_list

olive/cli/session_params_tuning.py

@@ -26,7 +26,6 @@
 
 
 class SessionParamsTuningCommand(BaseOliveCLICommand):
-
     @staticmethod
     def register_subcommand(parser: ArgumentParser):
         sub_parser = parser.add_parser(

olive/data/component/dataset.py

@@ -157,9 +157,9 @@ def __init__(
         self.annotations = None
         if annotations_file is not None:
             self.annotations = np.load(self.data_dir / annotations_file)
-            assert len(self.annotations) == len(
-                self.input_files
-            ), "Number of annotations should be equal to number of input files."
+            assert len(self.annotations) == len(self.input_files), (
+                "Number of annotations should be equal to number of input files."
+            )
 
     def __len__(self):
         return len(self.input_files)

olive/data/component/load_dataset.py

@@ -26,7 +26,7 @@ def huggingface_dataset(
     split: Optional[str] = "validation",
     data_files: Optional[Union[str, Sequence[str], Mapping[str, Union[str, Sequence[str]]]]] = None,
     col_filters: Optional[Mapping[str, Union[str, int, float]]] = None,
-    **kwargs
+    **kwargs,
 ):
     """Create a dataset from huggingface datasets."""
     from datasets.utils.logging import disable_progress_bar, set_verbosity_error

olive/data/component/pre_process_data.py

@@ -189,7 +189,7 @@ def audio_classification_pre_process(
     max_samples: Optional[int] = None,
     trust_remote_code: Optional[bool] = None,
     feature_extractor_args: Optional[Dict[str, Any]] = None,
-    **kwargs
+    **kwargs,
 ):
     """Pre-process data for audio classification task.