GitLab

Call str on id to make sure any ints are converted into the expected format for squad datasets

* Add run_mim_no_trainer.py draft from #20412

Add parse_args method and copy over other dependencies

Add Method call for sending telemetry

Initialize Accelerator

Make one log on every process

Set seed and Handle repository creation

Initialize dataset and Set validation split

Create Config

Adapt Config

Update Config

Create Feature Extractor

Create model

Set column names

Create transforms

Create mask generator

Create method to preprocess images

Shuffle datasets if needed and set transforms

Create Dataloaders

Add optimizer

Add learning rate scheduler

Prepare everything with our accelerator

Tie weights for TPU training

Recalculate training steps and training epochs

Set accelerator checkpointing steps

Initialize trackers and store configuration

Set total batch size

Fix typo: mlm -> mim

Log info at the start of training

Load in the weights and states from previous save

update the progress_bar if load from checkpoint

Define train loop

Add evaluation loop to training

Add to parse_args method

Push repo to hub

Save accelerator state

End training and save model and feature extractor

Remove unused imports

Fix trailing whitespace

* Update code based on comments, Rename feature_extractor to image_processor

* Fix linting

* Add argument for learning rate

* Add argument for setting number of training epochs

* Remove incorrect logger argument

* Convert max_train_steps to int for tqdm

---------
Co-authored-by: Saad Mahmud <shuvro.mahmud79@gmail.com>

Migration to Evaluate for metric is not quite complete

Save the tokenizer and image preprocessor after training a model with the contrastive image-text example (#23035)

Save tokenizer and image preprocessor

* Replace -100s in predictions by the pad token

* Style

* Try to catch them all

* Update run_speech_recognition_ctc.py

Make sure all processes wait until data is saved before loading the processor from the output_dit

* Make sure all processes wait until data is saved before loading the processor from the output_dit

* Update run_speech_recognition_ctc.py

* Update run_speech_recognition_seq2seq.py

Add id2label and label2id to config in run_xnil

allow only test_file in pytorch and flax summarization

* add low_cpu_mem_usage option in run_clm.py example which will benefit LLM loading
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* update all the example and README under language-modeling
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

---------
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* Add specaugment to run_speech_recognition_seq2seq.py

* Remove useless argument: text_column

* Fix quality

* Update return_attention_mask condition

* Update specaugment arguments only for whisper models

* Remove SpecAugment arguments from ModelArguments, only leave default values for simplicity

* Apply suggestions from code review
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

* Update apply_spec_augment only for whisper models

* Apply suggestions from code review
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

* Rename return_attention_mask to forward_attention_mask to avoid confusion with wav2vec2 models

---------
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

* Override the decoding parameters of Seq2SeqTrainer

* Fix quality

* Fix max_length parameter

* Fix quality

* Remove redundant parameter max_length

* Separate the preprocess of train and validation to use different max_target_length

* [Examples] Generalise run audio classification for log-mel models

* batch feature extractor

* make style

* Respect documentation on passive log level

* Fix test and set log level in examples

* Add doc

Update run_mae.py

Update run_speech_recognition_ctc.py

There should be `# limitations under the License` line at the end of the documentation section.

docs: fix some references in PyTorch text classification readme

* Result of black 23.1

* Update target to Python 3.7

* Switch flake8 to ruff

* Configure isort

* Configure isort

* Apply isort with line limit

* Put the right black version

* adapt black in check copies

* Fix copies

* simplify column_names in run_clm

* simplify column_names in run_mlm

* minor

* [run_clm example] add streaming dataset support

* unrefactor kwargs

* fix

* fix

* require datasets>=2.0.0

* port to mlm

* Update examples to use image processors

* Small fixes

* Resolve conflicts

* Adapt repository creation to latest hf_hub

* Update all examples

* Fix other tests, add Flax examples

* Address review comments

fix args passed to predict function

[NumPy] Remove references to deprecated NumPy type aliases.

This change replaces references to a number of deprecated NumPy type aliases (np.bool, np.int, np.float, np.complex, np.object, np.str) with their recommended replacement (bool, int, float, complex, object, str).

NumPy 1.24 drops the deprecated aliases, so we must remove uses before updating NumPy.
Co-authored-by: Peter Hawkins <phawkins@google.com>
Co-authored-by: Peter Hawkins <phawkins@google.com>

Added mask_time_prob and mask_time_length arguments to wav2vec2 pretraining script and readme - new branch

* [run_clm example] add torch_dtype option for model load.
for BLOOM 175B model. peak memory will reduce about 350G for inference. the weight of BLOOM in model hub is bfloat16
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* add other type in option

* fix style
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

Update big table
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MBP.localdomain>