Update formatting (#271)

.github/workflows/check_code_quality.yml

@@ -48,7 +48,7 @@ jobs:
       run: |
         source venv/bin/activate
         black --check .
-    - name: Check style with isort
+    - name: Check style with ruff
       run: |
         source venv/bin/activate
-        isort --check .
+        ruff .

Makefile

@@ -22,11 +22,11 @@ REAL_CLONE_URL = $(if $(CLONE_URL),$(CLONE_URL),$(DEFAULT_CLONE_URL))
 # Run code quality checks
 style_check:
 	black --check .
-	isort --check .
+	ruff .
 
 style:
 	black .
-	isort .
+	ruff . --fix
 
 # Run tests for the library
 test:

docs/source/inference.mdx

@@ -70,7 +70,7 @@ When fixing the shapes with the `reshape()` method, inference cannot be performe
 from datasets import load_dataset
 from transformers import AutoTokenizer, pipeline
 from evaluate import evaluator
-from optimum.intel.openvino import OVModelForQuestionAnswering
+from optimum.intel import OVModelForQuestionAnswering
 
 model_id = "distilbert-base-cased-distilled-squad"
 model = OVModelForQuestionAnswering.from_pretrained(model_id, export=True)
@@ -92,7 +92,7 @@ metric = task_evaluator.compute(model_or_pipeline=qa_pipe, data=eval_dataset, me
 By default the model will be compiled when instantiating our `OVModel`. In the case where the model is reshaped, placed to an other device or if FP16 precision is enabled, the model will need to be recompiled again, which will happen by default before the first inference (thus inflating the latency of the first inference). To avoid an unnecessary compilation, you can disable the first compilation by setting `compile=False`. The model should also be compiled before the first inference with `model.compile()`.
 
 ```python
-from optimum.intel.openvino import OVModelForSequenceClassification
+from optimum.intel import OVModelForSequenceClassification
 
 model_id = "distilbert-base-uncased-finetuned-sst-2-english"
 # Load the model and disable the model compilation
@@ -111,7 +111,7 @@ Here is an example on how you can run inference for a translation task using an
 
 ```python
 from transformers import AutoTokenizer, pipeline
-from optimum.intel.openvino import OVModelForSeq2SeqLM
+from optimum.intel import OVModelForSeq2SeqLM
 
 model_id = "t5-small"
 model = OVModelForSeq2SeqLM.from_pretrained(model_id, export=True)
@@ -147,7 +147,7 @@ pip install diffusers
 Here is an example of how you can load an OpenVINO Stable Diffusion model and run inference using OpenVINO Runtime:
 
 ```python
-from optimum.intel.openvino import OVStableDiffusionPipeline
+from optimum.intel import OVStableDiffusionPipeline
 
 model_id = "echarlaix/stable-diffusion-v1-5-openvino"
 stable_diffusion = OVStableDiffusionPipeline.from_pretrained(model_id)

docs/source/optimization_ov.mdx

@@ -16,7 +16,7 @@ limitations under the License.
 
 # Optimization
 
-🤗 Optimum Intel provides an `optimum.openvino` package that enables you to apply a variety of model compression methods such as quantization, pruning, on many models hosted on the 🤗 hub using the [NNCF](https://docs.openvino.ai/2022.1/docs_nncf_introduction.html) framework.
+🤗 Optimum Intel provides an `openvino` package that enables you to apply a variety of model compression methods such as quantization, pruning, on many models hosted on the 🤗 hub using the [NNCF](https://docs.openvino.ai/2022.1/docs_nncf_introduction.html) framework.
 
 
 ## Post-training optimization
@@ -27,7 +27,7 @@ Here is how to apply static quantization on a fine-tuned DistilBERT:
 ```python
 from functools import partial
 from transformers import AutoModelForSequenceClassification, AutoTokenizer
-from optimum.intel.openvino import OVConfig, OVQuantizer
+from optimum.intel import OVConfig, OVQuantizer
 
 model_id = "distilbert-base-uncased-finetuned-sst-2-english"
 model = AutoModelForSequenceClassification.from_pretrained(model_id)
@@ -83,7 +83,7 @@ from transformers import (
 )
 from datasets import load_dataset
 - from transformers import Trainer
-+ from optimum.intel.openvino import OVConfig, OVTrainer, OVModelForSequenceClassification
++ from optimum.intel import OVConfig, OVTrainer, OVModelForSequenceClassification
 
 model_id = "distilbert-base-uncased-finetuned-sst-2-english"
 model = AutoModelForSequenceClassification.from_pretrained(model_id)
@@ -168,7 +168,7 @@ Once we have the config ready, we can start develop the training pipeline like t
 
 ```diff
 - from transformers import Trainer, TrainingArguments
-+ from optimum.intel.openvino import OVConfig, OVTrainer, OVTrainingArguments
++ from optimum.intel import OVConfig, OVTrainer, OVTrainingArguments
 
 # Load teacher model
 + teacher_model = AutoModelForSequenceClassification.from_pretrained(teacher_model_or_path)
@@ -210,7 +210,7 @@ After applying quantization on our model, we can then easily load it with our `O
 
 ```python
 from transformers import pipeline
-from optimum.intel.openvino import OVModelForSequenceClassification
+from optimum.intel import OVModelForSequenceClassification
 
 model_id = "helenai/distilbert-base-uncased-finetuned-sst-2-english-ov-int8"
 ov_model = OVModelForSequenceClassification.from_pretrained(model_id)

examples/neural_compressor/language-modeling/run_clm.py

@@ -29,9 +29,16 @@
 from typing import Optional
 
 import datasets
+import evaluate
 import torch
 import transformers
 from datasets import load_dataset
+from neural_compressor import (
+    DistillationConfig,
+    PostTrainingQuantConfig,
+    QuantizationAwareTrainingConfig,
+    WeightPruningConfig,
+)
 from transformers import (
     CONFIG_MAPPING,
     MODEL_FOR_CAUSAL_LM_MAPPING,
@@ -49,13 +56,6 @@
 from transformers.utils import check_min_version
 from transformers.utils.versions import require_version
 
-import evaluate
-from neural_compressor import (
-    DistillationConfig,
-    PostTrainingQuantConfig,
-    QuantizationAwareTrainingConfig,
-    WeightPruningConfig,
-)
 from optimum.intel.neural_compressor import INCModelForCausalLM, INCQuantizer, INCTrainer
 
 
@@ -441,7 +441,7 @@ def main():
         )
     else:
         model = AutoModelForCausalLM.from_config(config)
-        n_params = sum(dict((p.data_ptr(), p.numel()) for p in model.parameters()).values())
+        n_params = sum({p.data_ptr(): p.numel() for p in model.parameters()}.values())
         logger.info(f"Training new model from scratch - Total size={n_params/2**20:.2f}M params")
 
     # We resize the embeddings only when necessary to avoid index errors. If you are creating a model from scratch

examples/neural_compressor/language-modeling/run_mlm.py

@@ -29,9 +29,16 @@
 from typing import Optional
 
 import datasets
+import evaluate
 import torch
 import transformers
 from datasets import load_dataset
+from neural_compressor import (
+    DistillationConfig,
+    PostTrainingQuantConfig,
+    QuantizationAwareTrainingConfig,
+    WeightPruningConfig,
+)
 from transformers import (
     CONFIG_MAPPING,
     MODEL_FOR_MASKED_LM_MAPPING,
@@ -48,13 +55,6 @@
 from transformers.utils import check_min_version
 from transformers.utils.versions import require_version
 
-import evaluate
-from neural_compressor import (
-    DistillationConfig,
-    PostTrainingQuantConfig,
-    QuantizationAwareTrainingConfig,
-    WeightPruningConfig,
-)
 from optimum.intel.neural_compressor import INCModelForMaskedLM, INCQuantizer, INCTrainer
 
 

examples/neural_compressor/multiple-choice/run_swag.py

@@ -29,12 +29,12 @@
 import torch
 import transformers
 from datasets import load_dataset
+from neural_compressor import DistillationConfig, QuantizationAwareTrainingConfig, WeightPruningConfig
 from transformers import (
     AutoConfig,
     AutoModelForMultipleChoice,
     AutoTokenizer,
     HfArgumentParser,
-    PreTrainedModel,
     TrainingArguments,
     default_data_collator,
     set_seed,
@@ -45,7 +45,6 @@
 from transformers.utils import check_min_version
 from transformers.utils.versions import require_version
 
-from neural_compressor import DistillationConfig, QuantizationAwareTrainingConfig, WeightPruningConfig
 from optimum.intel.neural_compressor import INCModelForMultipleChoice, INCTrainer
 
 
@@ -548,14 +547,14 @@ def compute_metrics(eval_predictions):
         trainer.log_metrics("eval", metrics)
         trainer.save_metrics("eval", metrics)
 
-    kwargs = dict(
-        finetuned_from=model_args.model_name_or_path,
-        tasks="multiple-choice",
-        dataset_tags="swag",
-        dataset_args="regular",
-        dataset="SWAG",
-        language="en",
-    )
+    kwargs = {
+        "finetuned_from": model_args.model_name_or_path,
+        "tasks": "multiple-choice",
+        "dataset_tags": "swag",
+        "dataset_args": "regular",
+        "dataset": "SWAG",
+        "language": "en",
+    }
 
     if training_args.push_to_hub:
         trainer.push_to_hub(**kwargs)

examples/neural_compressor/multiple-choice/run_swag_post_training.py

@@ -25,17 +25,18 @@
 from typing import Optional, Union
 
 import datasets
-import numpy as np
+import evaluate
 import torch
 import transformers
+from accelerate import Accelerator
 from datasets import load_dataset
+from neural_compressor import PostTrainingQuantConfig
 from torch.utils.data import DataLoader
 from transformers import (
     AutoConfig,
     AutoModelForMultipleChoice,
     AutoTokenizer,
     HfArgumentParser,
-    PreTrainedModel,
     TrainingArguments,
     default_data_collator,
     set_seed,
@@ -45,9 +46,6 @@
 from transformers.utils import check_min_version
 from transformers.utils.versions import require_version
 
-import evaluate
-from accelerate import Accelerator
-from neural_compressor import PostTrainingQuantConfig
 from optimum.intel.neural_compressor import INCModelForMultipleChoice, INCQuantizer
 
 

examples/neural_compressor/optical-character-recognition/run_ocr_post_training.py

@@ -5,16 +5,16 @@
 from dataclasses import dataclass, field
 from typing import Optional
 
+import evaluate
 import pandas as pd
 import torch
 import transformers
+from neural_compressor import PostTrainingQuantConfig
 from PIL import Image
 from torch.utils.data import DataLoader, Dataset
 from transformers import HfArgumentParser, TrainingArguments, TrOCRProcessor, VisionEncoderDecoderModel
 from transformers.utils import check_min_version
 
-import evaluate
-from neural_compressor import PostTrainingQuantConfig
 from optimum.intel.neural_compressor import INCModelForVision2Seq, INCQuantizer
 
 
@@ -284,7 +284,7 @@ def eval_func(model, iters=None):
             if result_loaded_model != result_optimized_model:
                 logger.error("The quantized model was not successfully loaded.")
             else:
-                logger.info(f"The quantized model was successfully loaded.")
+                logger.info("The quantized model was successfully loaded.")
 
 
 def _mp_fn(index):

examples/neural_compressor/question-answering/run_qa.py

@@ -25,20 +25,23 @@
 from typing import Optional
 
 import datasets
+import evaluate
 import numpy as np
 import torch
 import transformers
+from accelerate import Accelerator
 from datasets import load_dataset
+from neural_compressor import DistillationConfig, QuantizationAwareTrainingConfig, WeightPruningConfig
 from torch.utils.data.dataloader import DataLoader
 from tqdm.auto import tqdm
+from trainer_qa import QuestionAnsweringINCTrainer
 from transformers import (
     AutoConfig,
     AutoModelForQuestionAnswering,
     AutoTokenizer,
     DataCollatorWithPadding,
     EvalPrediction,
     HfArgumentParser,
-    PreTrainedModel,
     PreTrainedTokenizerFast,
     TrainingArguments,
     default_data_collator,
@@ -47,13 +50,9 @@
 from transformers.trainer_utils import get_last_checkpoint
 from transformers.utils import check_min_version
 from transformers.utils.versions import require_version
+from utils_qa import postprocess_qa_predictions
 
-import evaluate
-from accelerate import Accelerator
-from neural_compressor import DistillationConfig, QuantizationAwareTrainingConfig, WeightPruningConfig
 from optimum.intel.neural_compressor import INCModelForQuestionAnswering
-from trainer_qa import QuestionAnsweringINCTrainer
-from utils_qa import postprocess_qa_predictions
 
 
 # Will be removed when neural-compressor next release is out
@@ -553,7 +552,10 @@ def move_input_to_device(input, device):
         )
         teacher_model_qa = QAModel(teacher_model)
         teacher_model_qa = accelerator.prepare(teacher_model_qa)
-        num_param = lambda model: sum(p.numel() for p in model.parameters())
+
+        def num_param(model):
+            return sum(p.numel() for p in model.parameters())
+
         logger.info(
             "***** Number of teacher model parameters: {:.2f}M *****".format(num_param(teacher_model_qa) / 10**6)
         )
@@ -662,9 +664,33 @@ def prepare_validation_features(examples):
                 load_from_cache_file=not data_args.overwrite_cache,
                 desc="Running tokenizer on validation dataset",
             )
+
         if data_args.max_eval_samples is not None:
             # During Feature creation dataset samples might increase, we will select required samples again
-            eval_dataset = eval_dataset.select(range(data_args.max_eval_samples))
+            max_eval_samples = min(len(eval_dataset), data_args.max_eval_samples)
+            eval_dataset = eval_dataset.select(range(max_eval_samples))
+
+    if training_args.do_predict:
+        if "test" not in raw_datasets:
+            raise ValueError("--do_predict requires a test dataset")
+        predict_examples = raw_datasets["test"]
+        if data_args.max_predict_samples is not None:
+            # We will select sample from whole data
+            predict_examples = predict_examples.select(range(data_args.max_predict_samples))
+        # Predict Feature Creation
+        with training_args.main_process_first(desc="prediction dataset map pre-processing"):
+            predict_dataset = predict_examples.map(
+                prepare_validation_features,
+                batched=True,
+                num_proc=data_args.preprocessing_num_workers,
+                remove_columns=column_names,
+                load_from_cache_file=not data_args.overwrite_cache,
+                desc="Running tokenizer on prediction dataset",
+            )
+        if data_args.max_predict_samples is not None:
+            # During Feature creation dataset samples might increase, we will select required samples again
+            max_predict_samples = min(len(predict_dataset), data_args.max_predict_samples)
+            predict_dataset = predict_dataset.select(range(max_predict_samples))
 
     # Post-processing:
     def post_processing_function(examples, features, predictions, stage="eval"):