Model ( PreTrainedModel or torch.nn.Module, optional) – Trainer is a simple but feature-complete training and eval loop for PyTorch, optimized for 🤗 Transformers. Trainer ( model: torch.nn. = None, args: aining_args.TrainingArguments = None, data_collator: Optional = None, train_dataset: Optional = None, eval_dataset: Optional = None, tokenizer: Optional = None, model_init: Callable = None, compute_metrics: Optional] = None, callbacks: Optional] = None, optimizers: Tuple = (None, None) ) ¶ Predict – Returns predictions (with metrics if labels are available) on a test set. Run_model (TensorFlow only) – Basic pass through the model.Įvaluate – Runs an evaluation loop and returns metrics. Prediction_step – Performs an evaluation/test step. Training_step – Performs a training step. Log – Logs information on the various objects watching training.Ĭreate_optimizer_and_scheduler – Sets up the optimizer and learning rate scheduler if they were not passed atĬompute_loss - Computes the loss on a batch of training inputs. Get_test_dataloader/ get_test_tfdataset – Creates the test DataLoader (PyTorch) or TF Dataset. Get_eval_dataloader/ get_eval_tfdataset – Creates the evaluation DataLoader (PyTorch) or TF Dataset. Get_train_dataloader/ get_train_tfdataset – Creates the training DataLoader (PyTorch) or TF Dataset. To inject custom behavior you can subclass them and override the following methods: The API supports distributed training on multiple GPUs/TPUs, mixed precision through NVIDIA Apex and Native AMP for PyTorch and tf.keras.mixed_precision for TensorFlow.īoth Trainer and TFTrainer contain the basic training loop which supports TrainingArguments/ TFTrainingArguments to access all the points of It’s used in most of the example scripts.īefore instantiating your Trainer/ TFTrainer, create a The Trainer and TFTrainer classes provide an API for feature-complete
0 kommentar(er)
