fxmarty

The custom CUDA extensions may be uncompatible with PyTorch 1.10.0, possible missing dependency `ninja`

7

Hello, I run the training in a google colab instance. My code is roughly the following (in different cells): ``` from google.colab import drive drive.mount('/content/drive') %cd "drive/My Drive/DirectVoxGO" pip install...

Support for ununiformly spaced data

10

Hello, Today I tried to plot data for which the time column was not uniformly spaced, there is the result:  Corresponding data: ``` Open High Low Close Volume Date...

`with_format` behavior is inconsistent on different datasets

1

## Describe the bug I found a case where `with_format` does not transform the dataset to the requested format. ## Steps to reproduce the bug Run: ```python from transformers import...

# What does this PR do? Add the folding of `nn.BatchNorm2d` into `nn.Conv2d`, and of `nn.BatchNorm1d` into `nn.Linear`. The former is useful for example for resnet, beit, regnet, dpt, levit,...

[optimum-onnxruntime] The number of calibration samples must be divisible by (num_calibration_shards * calibration_batch_size)

1

### System Info ```shell * pytorch 1.12.0+cu102 * onnxruntime 1.12.0 * optimum 1.4.0.dev0 * onnx-1.12.0 * numpy-1.23.1 ``` ### Who can help? _No response_ ### Information - [X] The official...

Add InferenceSession options and Provider to ORTModel

1

`ORTModel.load_model` was not accepting custom session options, making it hard for example to use a given number of CPU cores with an `ORTModel` for inference. Indeed, onnxruntime obey neither `psutil.cpu_affinity`...

Use evaluators in the benchmark scripts

1

# What does this PR do? This is part of an effort on my end to simplify the benchmark scripts. As much of the logic of the evaluation using pipelines...

Raise warning in examples for classification tasks where the model's label mapping can not automatically match with the dataset labels

1

Following https://github.com/huggingface/optimum/pull/197 , as per the title. For example ```python cfg = AutoConfig.from_pretrained("howey/bert-base-uncased-sst2") print(cfg) """prints BertConfig { "_name_or_path": "howey/bert-base-uncased-sst2", "architectures": [ "BertForSequenceClassification" ], "attention_probs_dropout_prob": 0.1, "classifier_dropout": null, "finetuning_task": "sst2", "gradient_checkpointing":...

Add option to save benchmark results

1

This PR adds an option in the benchmark scripts to save the run configuration, the model obtained, and the evaluation results. ## Before submitting - [x] Did you write any...

Dissociate PyTorch / ONNX Runtime benchmark scripts

1

# What does this PR do? * Dissociate the metrics / time evaluation between PyTorch and ONNX Runtime. * Add a method to perform only perf metrics evaluation, or only...