Add Flax BART pretraining script (#18297)

* add bart pretraining flax script

* fixup

* add bart pretraining flax script

* add BART to README

* add BART to README

* add BART to README

* add BART to README

* add BART to README

* add bos eos document

* Update README.md

* Update README.md

* Update examples/flax/language-modeling/run_bart_dlm_flax.py
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

* final

* final

* final

* remove use_auth_token ing from_config
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

examples/flax/language-modeling/README.md

@@ -338,6 +338,98 @@ of 2.36 and 57.0 respectively after 3 epochs on a single TPUv3-8.
 This should take around 4.5 hours.
 Training statistics can be accessed on directly on the 🤗 [hub](https://huggingface.co/patrickvonplaten/t5-base-norwegian/tensorboard)
 
+## BART: Denoising language modeling
+
+In the following, we demonstrate how to train a BART model 
+using denoising language modeling objective as introduced in [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/abs/1910.13461).
+More specifically, we demonstrate how JAX/Flax can be leveraged 
+to pre-train [**`bart-base`**](https://huggingface.co/facebook/bart-base)
+in Norwegian on a single TPUv3-8 pod.
+
+The example script uses the 🤗 Datasets library. You can easily customize them to your needs if you need extra processing on your datasets.
+
+To setup all relevant files for training, let's create a directory.
+
+```bash
+mkdir ./norwegian-roberta-base
+```
+
+### Train tokenizer
+In the first step, we train a tokenizer to efficiently process the text input for the model. Similar to how it is shown in [How to train a new language model from scratch using Transformers and Tokenizers](https://huggingface.co/blog/how-to-train), we use a **`ByteLevelBPETokenizer`**.
+The tokenizer is trained on the complete Norwegian dataset of OSCAR
+and consequently saved in the cloned model directory.
+This can take up to 10 minutes depending on your hardware ☕.
+
+```python
+from datasets import load_dataset
+from tokenizers import trainers, Tokenizer, normalizers, ByteLevelBPETokenizer
+
+# load dataset
+dataset = load_dataset("oscar", "unshuffled_deduplicated_no", split="train")
+
+# Instantiate tokenizer
+tokenizer = ByteLevelBPETokenizer()
+
+def batch_iterator(batch_size=1000):
+    for i in range(0, len(dataset), batch_size):
+        yield dataset[i: i + batch_size]["text"]
+
+# Customized training
+tokenizer.train_from_iterator(batch_iterator(), vocab_size=50265, min_frequency=2, special_tokens=[
+    "<s>",
+    "<pad>",
+    "</s>",
+    "<unk>",
+    "<mask>",
+])
+
+# Save files to disk
+tokenizer.save("./norwegian-bart-base/tokenizer.json")
+```
+
+### Create configuration
+
+Next, we create the model's configuration file. This is as simple 
+as loading and storing [`**facebook/bart-base**`](https://huggingface.co/facebook/bart-base)
+in the local model folder:
+
+```python
+from transformers import BartConfig
+config = BartConfig.from_pretrained("facebook/bart-base", vocab_size=50265)
+config.save_pretrained("./norwegian-bart-base")
+```
+
+Great, we have set up our model repository. During training, we will automatically
+push the training logs and model weights to the repo.
+
+### Train model
+
+Next we can run the example script to pretrain the model:
+
+```bash
+python run_bart_dlm_flax.py \
+    --output_dir="./norwegian-bart-base" \
+    --config_name="./norwegian-bart-base" \
+    --tokenizer_name="./norwegian-bart-base" \
+    --dataset_name="oscar" \
+    --dataset_config_name="unshuffled_deduplicated_no" \
+    --max_seq_length="1024" \
+    --per_device_train_batch_size="32" \
+    --per_device_eval_batch_size="32" \
+    --learning_rate="1e-4" \
+    --warmup_steps="2000" \
+    --overwrite_output_dir \
+    --logging_steps="500" \
+    --save_steps="2000" \
+    --eval_steps="2000" \
+    --push_to_hub
+```
+
+Training should converge at a loss and accuracy 
+of 1.36 and 0.77 respectively after 3 epochs on a single TPUv3-8.
+This should take less than 6 hours.
+Training statistics can be accessed on [tfhub.dev](https://tensorboard.dev/experiment/Maw62QlaSXWS0MOf2V2lbg/).
+
 ## Runtime evaluation
 
 We also ran masked language modeling using PyTorch/XLA on a TPUv3-8, and PyTorch on 8 V100 GPUs. We report the

examples/flax/language-modeling/run_bart_dlm_flax.py

+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2021 The HuggingFace Team All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Pretraining the library models for denoising language modeling on a text file or a dataset.
+Here is the full list of checkpoints on the hub that can be pretrained by this script:
+https://huggingface.co/models?filter=bart
+"""
+# You can also adapt this script on your own denoising language modeling task. Pointers for this are left as comments.
+
+import json
+import logging
+import math
+import os
+import sys
+import time
+from dataclasses import asdict, dataclass, field
+from enum import Enum
+from itertools import chain
+from pathlib import Path
+from typing import Dict, List, Optional
+
+import nltk
+import numpy as np
+from datasets import load_dataset
+from tqdm import tqdm
+
+import flax
+import jax
+import jax.numpy as jnp
+import optax
+from flax import jax_utils, traverse_util
+from flax.jax_utils import pad_shard_unpad
+from flax.training import train_state
+from flax.training.common_utils import get_metrics, onehot, shard
+from huggingface_hub import Repository
+from transformers import (
+    CONFIG_MAPPING,
+    FLAX_MODEL_FOR_MASKED_LM_MAPPING,
+    AutoTokenizer,
+    BartConfig,
+    BatchEncoding,
+    FlaxBartForConditionalGeneration,
+    HfArgumentParser,
+    PreTrainedTokenizerBase,
+    is_tensorboard_available,
+    set_seed,
+)
+from transformers.models.bart.modeling_flax_bart import shift_tokens_right
+from transformers.utils import get_full_repo_name, send_example_telemetry
+
+
+MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_MASKED_LM_MAPPING.keys())
+MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
+
+
+@dataclass
+class TrainingArguments:
+    output_dir: str = field(
+        metadata={"help": "The output directory where the model predictions and checkpoints will be written."},
+    )
+    overwrite_output_dir: bool = field(
+        default=False,
+        metadata={
+            "help": (
+                "Overwrite the content of the output directory. "
+                "Use this to continue training if output_dir points to a checkpoint directory."
+            )
+        },
+    )
+    do_train: bool = field(default=False, metadata={"help": "Whether to run training."})
+    do_eval: bool = field(default=False, metadata={"help": "Whether to run eval on the dev set."})
+    per_device_train_batch_size: int = field(
+        default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for training."}
+    )
+    per_device_eval_batch_size: int = field(
+        default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for evaluation."}
+    )
+    learning_rate: float = field(default=5e-5, metadata={"help": "The initial learning rate for AdamW."})
+    weight_decay: float = field(default=0.0, metadata={"help": "Weight decay for AdamW if we apply some."})
+    adam_beta1: float = field(default=0.9, metadata={"help": "Beta1 for AdamW optimizer"})
+    adam_beta2: float = field(default=0.999, metadata={"help": "Beta2 for AdamW optimizer"})
+    adam_epsilon: float = field(default=1e-8, metadata={"help": "Epsilon for AdamW optimizer."})
+    adafactor: bool = field(default=False, metadata={"help": "Whether or not to replace AdamW by Adafactor."})
+    num_train_epochs: float = field(default=3.0, metadata={"help": "Total number of training epochs to perform."})
+    warmup_steps: int = field(default=0, metadata={"help": "Linear warmup over warmup_steps."})
+    logging_steps: int = field(default=500, metadata={"help": "Log every X updates steps."})
+    save_steps: int = field(default=500, metadata={"help": "Save checkpoint every X updates steps."})
+    eval_steps: int = field(default=None, metadata={"help": "Run an evaluation every X steps."})
+    seed: int = field(default=42, metadata={"help": "Random seed that will be set at the beginning of training."})
+    push_to_hub: bool = field(
+        default=False, metadata={"help": "Whether or not to upload the trained model to the model hub after training."}
+    )
+    hub_model_id: str = field(
+        default=None, metadata={"help": "The name of the repository to keep in sync with the local `output_dir`."}
+    )
+    hub_token: str = field(default=None, metadata={"help": "The token to use to push to the Model Hub."})
+
+    def __post_init__(self):
+        if self.output_dir is not None:
+            self.output_dir = os.path.expanduser(self.output_dir)
+
+    def to_dict(self):
+        """
+        Serializes this instance while replace `Enum` by their values (for JSON serialization support). It obfuscates
+        the token values by removing their value.
+        """
+        d = asdict(self)
+        for k, v in d.items():
+            if isinstance(v, Enum):
+                d[k] = v.value
+            if isinstance(v, list) and len(v) > 0 and isinstance(v[0], Enum):
+                d[k] = [x.value for x in v]
+            if k.endswith("_token"):
+                d[k] = f"<{k.upper()}>"
+        return d
+
+
+@dataclass
+class ModelArguments:
+    """
+    Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
+    """
+
+    model_name_or_path: Optional[str] = field(
+        default=None,
+        metadata={
+            "help": (
+                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+            )
+        },
+    )
+    model_type: Optional[str] = field(
+        default=None,
+        metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
+    )
+    config_name: Optional[str] = field(
+        default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"}
+    )
+    tokenizer_name: Optional[str] = field(
+        default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"}
+    )
+    cache_dir: Optional[str] = field(
+        default=None, metadata={"help": "Where do you want to store the pretrained models downloaded from s3"}
+    )
+    use_fast_tokenizer: bool = field(
+        default=True,
+        metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
+    )
+    dtype: Optional[str] = field(
+        default="float32",
+        metadata={
+            "help": (
+                "Floating-point format in which the model weights should be initialized and trained. Choose one of"
+                " `[float32, float16, bfloat16]`."
+            )
+        },
+    )
+    use_auth_token: bool = field(
+        default=False,
+        metadata={
+            "help": (
+                "Will use the token generated when running `transformers-cli login` (necessary to use this script "
+                "with private models)."
+            )
+        },
+    )
+
+
+@dataclass
+class DataTrainingArguments:
+    """
+    Arguments pertaining to what data we are going to input our model for training and eval.
+    """
+
+    dataset_name: Optional[str] = field(
+        default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."}
+    )
+    dataset_config_name: Optional[str] = field(
+        default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+    )
+    train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
+    validation_file: Optional[str] = field(
+        default=None,
+        metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
+    )
+    train_ref_file: Optional[str] = field(
+        default=None,
+        metadata={"help": "An optional input train ref data file for whole word masking in Chinese."},
+    )
+    validation_ref_file: Optional[str] = field(
+        default=None,
+        metadata={"help": "An optional input validation ref data file for whole word masking in Chinese."},
+    )
+    overwrite_cache: bool = field(
+        default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
+    )
+    validation_split_percentage: Optional[int] = field(
+        default=5,
+        metadata={
+            "help": "The percentage of the train set used as validation set in case there's no validation split"
+        },
+    )
+    max_seq_length: Optional[int] = field(
+        default=None,
+        metadata={
+            "help": (
+                "The maximum total input sequence length after tokenization and masking. Sequences longer than this"
+                " will be truncated. Default to the max input length of the model."
+            )
+        },
+    )
+    preprocessing_num_workers: Optional[int] = field(
+        default=None,
+        metadata={"help": "The number of processes to use for the preprocessing."},
+    )
+    mlm_probability: float = field(
+        default=0.3, metadata={"help": "Ratio of tokens to mask for span masked language modeling loss"}
+    )
+    permute_sentence_ratio: float = field(
+        default=1.0, metadata={"help": "Ratio of sentences to be permuted in each document"}
+    )
+    poisson_lambda: float = field(
+        default=3.0, metadata={"help": "Mean of Poisson distribution used to generate span-lengths to be masked"}
+    )
+
+    def __post_init__(self):
+        if self.dataset_name is None and self.train_file is None and self.validation_file is None:
+            raise ValueError("Need either a dataset name or a training/validation file.")
+        else:
+            if self.train_file is not None:
+                extension = self.train_file.split(".")[-1]
+                assert extension in ["csv", "json", "txt"], "`train_file` should be a csv, a json or a txt file."
+            if self.validation_file is not None:
+                extension = self.validation_file.split(".")[-1]
+                assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, a json or a txt file."
+
+
+@flax.struct.dataclass
+class FlaxDataCollatorForBartDenoisingLM:
+    """
+    Data collator used for BART denoising language modeling. The code is largely copied from
+    `<https://github.com/morganmcg1/rotobart/blob/main/data_collator.py#L223>`__.
+    For more information on how BART denoising language modeling works, one can take a look
+    at the `official paper <https://arxiv.org/pdf/1910.13461.pdf>`__
+    or the `official code for preprocessing <https://github.com/facebookresearch/fairseq/blob/main/fairseq/data/denoising_dataset.py>`__ .
+    Args:
+        tokenizer (:class:`~transformers.PreTrainedTokenizer` or :class:`~transformers.PreTrainedTokenizerFast`):
+            The tokenizer used for encoding the data
+        mask_ratio (:obj:`float`):
+            The probability with which to (randomly) mask tokens in the input
+        poisson_lambda (:obj:`float`):
+            Mean parameter of Poisson distribution used to generate span-lengths to be masked
+        permute_sentence_ratio (:obj:`float`):
+            Ratio of sentences to be permuted in each document
+        decoder_start_token_id: (:obj:`int):
+            The decoder start token id of the model
+    """
+
+    tokenizer: PreTrainedTokenizerBase
+    decoder_start_token_id: int
+    mask_ratio: float = 0.3
+    poisson_lambda: float = 3.0
+    permute_sentence_ratio: float = 1.0
+
+    def __post_init__(self):
+        if self.tokenizer.mask_token is None or self.tokenizer.eos_token is None:
+            raise ValueError(
+                "This tokenizer does not have a mask token or eos token token which is necessary for denoising"
+                " language modeling. "
+            )
+
+    def __call__(self, examples: List[Dict[str, List[int]]]) -> BatchEncoding:
+        # convert list to dict and tensorize input
+        batch = BatchEncoding(
+            {k: np.array([examples[i][k] for i in range(len(examples))]) for k, v in examples[0].items()}
+        )
+        batch["labels"] = batch["input_ids"].copy()
+        batch["decoder_input_ids"] = shift_tokens_right(
+            batch["labels"], self.tokenizer.pad_token_id, self.decoder_start_token_id
+        )
+        # permuting sentences
+        do_permute = False
+        if self.permute_sentence_ratio > 0.0:
+            batch["input_ids"] = self.permute_sentences(batch["input_ids"])
+            do_permute = True
+
+        # masking span of tokens (text infilling in the paper)
+        if self.mask_ratio:
+            batch["input_ids"], batch["labels"] = self.span_mask_tokens(
+                batch["input_ids"], batch["labels"], do_permute
+            )
+
+        # ignore pad tokens
+        batch["attention_mask"] = (batch["input_ids"] != self.tokenizer.pad_token_id).astype(int)
+        batch["decoder_attention_mask"] = (batch["decoder_input_ids"] != self.tokenizer.pad_token_id).astype(int)
+        return batch
+
+    def permute_sentences(self, input_ids):
+        """
+        Shuffle sentences in each document.
+        """
+        results = input_ids.copy()
+
+        # find end locations of sentences
+        end_sentence_mask = input_ids == self.tokenizer.pad_token_id
+        sentence_ends = np.argwhere(end_sentence_mask)
+        sentence_ends[:, 1] += 1
+        example_has_multiple_sentences, num_sentences = np.unique(sentence_ends[:, 0], return_counts=True)
+        num_sentences_map = {sent_idx: count for sent_idx, count in zip(example_has_multiple_sentences, num_sentences)}
+
+        num_to_permute = np.ceil(num_sentences * self.permute_sentence_ratio).astype(int)
+        num_to_permute_map = {
+            sent_idx: count for sent_idx, count in zip(example_has_multiple_sentences, num_to_permute)
+        }
+
+        sentence_ends = np.split(sentence_ends[:, 1], np.unique(sentence_ends[:, 0], return_index=True)[1][1:])
+        sentence_ends_map = {sent_idx: count for sent_idx, count in zip(example_has_multiple_sentences, sentence_ends)}
+
+        for i in range(input_ids.shape[0]):
+            if i not in example_has_multiple_sentences:
+                continue
+            substitutions = np.random.permutation(num_sentences_map[i])[: num_to_permute_map[i]]
+            ordering = np.arange(0, num_sentences_map[i])
+            ordering[substitutions] = substitutions[np.random.permutation(num_to_permute_map[i])]
+
+            # write shuffled sentences into results
+            index = 0
+            for j in ordering:
+                sentence = input_ids[i, (sentence_ends_map[i][j - 1] if j > 0 else 0) : sentence_ends_map[i][j]]
+                results[i, index : index + sentence.shape[0]] = sentence
+                index += sentence.shape[0]
+        return results
+
+    def span_mask_tokens(self, input_ids, labels, do_permute):
+        """
+        Sampling text spans with span lengths drawn from a Poisson distribution and masking them.
+        """
+        special_tokens_mask_labels = [
+            self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()
+        ]
+        special_tokens_mask_inputs = [
+            self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in input_ids.tolist()
+        ]
+        special_tokens_mask_labels = np.array(special_tokens_mask_labels, dtype=bool)
+        special_tokens_mask_inputs = np.array(special_tokens_mask_inputs, dtype=bool)
+
+        # determine how many tokens we need to mask in total
+        is_token_mask = ~(input_ids == self.tokenizer.pad_token_id) & ~special_tokens_mask_inputs
+        num_tokens_to_mask = int(math.ceil(is_token_mask.astype(float).sum() * self.mask_ratio))
+        if num_tokens_to_mask == 0:
+            return input_ids, labels
+
+        # generate a sufficient number of span lengths
+        span_lengths = np.random.poisson(lam=self.poisson_lambda, size=(num_tokens_to_mask,))
+        while np.cumsum(span_lengths, 0)[-1] < num_tokens_to_mask:
+            span_lengths = np.concatenate(
+                [span_lengths, np.random.poisson(lam=self.poisson_lambda, size=(num_tokens_to_mask,))]
+            )
+
+        # remove all spans of length 0
+        # note that BART inserts additional mask tokens where length == 0,
+        # which we do not implement for now as it adds additional complexity
+        span_lengths = span_lengths[span_lengths > 0]
+
+        # trim to about num_tokens_to_mask tokens
+        cutoff_idx = np.argmin(np.abs(np.cumsum(span_lengths, 0) - num_tokens_to_mask)) + 1
+        span_lengths = span_lengths[:cutoff_idx]
+
+        # randomly choose starting positions for masking
+        token_indices = np.argwhere(is_token_mask == 1)
+        span_starts = np.random.permutation(token_indices.shape[0])[: span_lengths.shape[0]]
+        # prepare mask
+        masked_indices = np.array(token_indices[span_starts])
+        mask = np.full_like(input_ids, fill_value=False)
+
+        # mask starting positions
+        for mi in masked_indices:
+            mask[tuple(mi)] = True
+        span_lengths -= 1
+
+        # fill up spans
+        max_index = input_ids.shape[1] - 1
+        remaining = (span_lengths > 0) & (masked_indices[:, 1] < max_index)
+        while np.any(remaining):
+            masked_indices[remaining, 1] += 1
+            for mi in masked_indices:
+                mask[tuple(mi)] = True
+            span_lengths -= 1
+            remaining = (span_lengths > 0) & (masked_indices[:, 1] < max_index)
+
+        # place the mask tokens
+        mask[np.where(special_tokens_mask_inputs)] = False
+        input_ids[np.where(mask)] = self.tokenizer.mask_token_id
+        if not do_permute:
+            labels[np.where(mask == 0)] = -100
+        else:
+            labels[np.where(special_tokens_mask_labels)] = -100
+
+        # remove mask tokens that are not starts of spans
+        to_remove = (mask == 1) & np.roll((mask == 1), 1, 1)
+        new_input_ids = np.full_like(input_ids, fill_value=self.tokenizer.pad_token_id)
+        for i, example in enumerate(input_ids):
+            new_example = example[~to_remove[i]]
+            new_input_ids[i, : new_example.shape[0]] = new_example
+
+        return new_input_ids, labels
+
+
+def generate_batch_splits(samples_idx: np.ndarray, batch_size: int, drop_last=True) -> np.ndarray:
+    """Generate batches of data for a specified batch size from sample indices. If the dataset size is not divisible by
+    the batch size and `drop_last` is `True`, the last incomplete batch is dropped. Else, it is returned."""
+    num_samples = len(samples_idx)
+    if drop_last:
+        samples_to_remove = num_samples % batch_size
+        if samples_to_remove != 0:
+            samples_idx = samples_idx[:-samples_to_remove]
+        sections_split = num_samples // batch_size
+        samples_idx = samples_idx.reshape((sections_split, batch_size))
+    else:
+        sections_split = math.ceil(num_samples / batch_size)
+        samples_idx = np.array_split(samples_idx, sections_split)
+    return samples_idx
+
+
+def write_train_metric(summary_writer, train_metrics, train_time, step):
+    summary_writer.scalar("train_time", train_time, step)
+
+    train_metrics = get_metrics(train_metrics)
+    for key, vals in train_metrics.items():
+        tag = f"train_{key}"
+        for i, val in enumerate(vals):
+            summary_writer.scalar(tag, val, step - len(vals) + i + 1)
+
+
+def write_eval_metric(summary_writer, eval_metrics, step):
+    for metric_name, value in eval_metrics.items():
+        summary_writer.scalar(f"eval_{metric_name}", value, step)
+
+
+def main():
+    # See all possible arguments in src/transformers/training_args.py
+    # or by passing the --help flag to this script.
+    # We now keep distinct sets of args, for a cleaner separation of concerns.
+
+    parser = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments))
+    if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
+        # If we pass only one argument to the script and it's the path to a json file,
+        # let's parse it to get our arguments.
+        model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
+    else:
+        model_args, data_args, training_args = parser.parse_args_into_dataclasses()
+
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_bart_dlm", model_args, data_args, framework="flax")
+
+    if (
+        os.path.exists(training_args.output_dir)
+        and os.listdir(training_args.output_dir)
+        and training_args.do_train
+        and not training_args.overwrite_output_dir
+    ):
+        raise ValueError(
+            f"Output directory ({training_args.output_dir}) already exists and is not empty."
+            "Use --overwrite_output_dir to overcome."
+        )
+
+    # Setup logging
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        level=logging.INFO,
+        datefmt="[%X]",
+    )
+
+    # Log on each process the small summary:
+    logger = logging.getLogger(__name__)
+
+    # Set the verbosity to info of the Transformers logger (on main process only):
+    logger.info(f"Training/evaluation parameters {training_args}")
+
+    # Set seed before initializing model.
+    set_seed(training_args.seed)
+
+    # Handle the repository creation
+    if training_args.push_to_hub:
+        if training_args.hub_model_id is None:
+            repo_name = get_full_repo_name(
+                Path(training_args.output_dir).absolute().name, token=training_args.hub_token
+            )
+        else:
+            repo_name = training_args.hub_model_id
+        repo = Repository(training_args.output_dir, clone_from=repo_name)
+
+    # Get the datasets: you can either provide your own CSV/JSON/TXT training and evaluation files (see below)
+    # or just provide the name of one of the public datasets available on the hub at https://huggingface.co/datasets/
+    # (the dataset will be downloaded automatically from the datasets Hub).
+    #
+    # For CSV/JSON files, this script will use the column called 'text' or the first column if no column called
+    # 'text' is found. You can easily tweak this behavior (see below).
+    if data_args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        datasets = load_dataset(
+            data_args.dataset_name,
+            data_args.dataset_config_name,
+            cache_dir=model_args.cache_dir,
+            use_auth_token=True if model_args.use_auth_token else None,
+        )
+
+        if "validation" not in datasets.keys():
+            datasets["validation"] = load_dataset(
+                data_args.dataset_name,
+                data_args.dataset_config_name,
+                split=f"train[:{data_args.validation_split_percentage}%]",
+                cache_dir=model_args.cache_dir,
+                use_auth_token=True if model_args.use_auth_token else None,
+            )
+            datasets["train"] = load_dataset(
+                data_args.dataset_name,
+                data_args.dataset_config_name,
+                split=f"train[{data_args.validation_split_percentage}%:]",
+                cache_dir=model_args.cache_dir,
+                use_auth_token=True if model_args.use_auth_token else None,
+            )
+    else:
+        data_files = {}
+        if data_args.train_file is not None:
+            data_files["train"] = data_args.train_file
+        if data_args.validation_file is not None:
+            data_files["validation"] = data_args.validation_file
+        extension = data_args.train_file.split(".")[-1]
+        if extension == "txt":
+            extension = "text"
+        datasets = load_dataset(
+            extension,
+            data_files=data_files,
+            cache_dir=model_args.cache_dir,
+            use_auth_token=True if model_args.use_auth_token else None,
+        )
+
+        if "validation" not in datasets.keys():
+            datasets["validation"] = load_dataset(
+                extension,
+                data_files=data_files,
+                split=f"train[:{data_args.validation_split_percentage}%]",
+                cache_dir=model_args.cache_dir,
+                use_auth_token=True if model_args.use_auth_token else None,
+            )
+            datasets["train"] = load_dataset(
+                extension,
+                data_files=data_files,
+                split=f"train[{data_args.validation_split_percentage}%:]",
+                cache_dir=model_args.cache_dir,
+                use_auth_token=True if model_args.use_auth_token else None,
+            )
+    # See more about loading any type of standard or custom dataset (from files, python dict, pandas DataFrame, etc) at
+    # https://huggingface.co/docs/datasets/loading_datasets.html.
+
+    # Load pretrained model and tokenizer
+
+    if model_args.tokenizer_name:
+        tokenizer = AutoTokenizer.from_pretrained(
+            model_args.tokenizer_name,
+            cache_dir=model_args.cache_dir,
+            use_fast=model_args.use_fast_tokenizer,
+            use_auth_token=True if model_args.use_auth_token else None,
+        )
+    elif model_args.model_name_or_path:
+        tokenizer = AutoTokenizer.from_pretrained(
+            model_args.model_name_or_path,
+            cache_dir=model_args.cache_dir,
+            use_fast=model_args.use_fast_tokenizer,
+            use_auth_token=True if model_args.use_auth_token else None,
+        )
+    else:
+        raise ValueError(
+            "You are instantiating a new tokenizer from scratch. This is not supported by this script."
+            "You can do it from another script, save it, and load it from here, using --tokenizer_name."
+        )
+
+    if model_args.config_name:
+        config = BartConfig.from_pretrained(
+            model_args.config_name,
+            cache_dir=model_args.cache_dir,
+            vocab_size=len(tokenizer),
+            use_auth_token=True if model_args.use_auth_token else None,
+        )
+    elif model_args.model_name_or_path:
+        config = BartConfig.from_pretrained(
+            model_args.model_name_or_path,
+            cache_dir=model_args.cache_dir,
+            use_auth_token=True if model_args.use_auth_token else None,
+        )
+    else:
+        config = CONFIG_MAPPING[model_args.model_type]()
+        logger.warning("You are instantiating a new config instance from scratch.")
+
+    # Preprocessing the datasets.
+    # First we tokenize all the texts.
+    if training_args.do_train:
+        column_names = datasets["train"].column_names
+    else:
+        column_names = datasets["validation"].column_names
+    text_column_name = "text" if "text" in column_names else column_names[0]
+
+    max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
+
+    # Use Punkt Sentence Tokenizer to divide a document into a list of sentences
+    nltk.download("punkt")
+    sentence_tokenizer = nltk.data.load("tokenizers/punkt/english.pickle")
+
+    def sentence_split_function(example):
+        sents = sentence_tokenizer.tokenize(example["text"])
+        # use pad token as end of sentence indicator
+        new_text = tokenizer.bos_token + f"{tokenizer.pad_token}".join(sents) + tokenizer.eos_token
+        return {"text": new_text}
+
+    split_datasets = datasets.map(
+        sentence_split_function,
+        batched=False,
+        num_proc=data_args.preprocessing_num_workers,
+        remove_columns=column_names,
+        load_from_cache_file=not data_args.overwrite_cache,
+    )
+
+    # Tokenize every text, then concatenate them together before splitting them in smaller parts.
+    # Since we make sure that all sequences are of the same length, no attention_mask is needed.
+    def tokenize_function(examples):
+        return tokenizer(examples[text_column_name], add_special_tokens=False, return_attention_mask=False)
+
+    tokenized_datasets = split_datasets.map(
+        tokenize_function,
+        batched=True,
+        num_proc=data_args.preprocessing_num_workers,
+        remove_columns=text_column_name,
+        load_from_cache_file=not data_args.overwrite_cache,
+    )
+
+    # Main data processing function that will concatenate all texts from our dataset and generate chunks of
+    # max_seq_length.
+    def group_texts(examples):
+        # Concatenate all texts.
+        concatenated_examples = {k: list(chain(*examples[k])) for k in examples.keys()}
+        total_length = len(concatenated_examples[list(examples.keys())[0]])
+        # We drop the small remainder, we could add padding if the model supported it instead of this drop, you can
+        # customize this part to your needs.
+        if total_length >= max_seq_length:
+            total_length = (total_length // max_seq_length) * max_seq_length
+        # Split by chunks of max_len.
+        result = {
+            k: [t[i : i + max_seq_length] for i in range(0, total_length, max_seq_length)]
+            for k, t in concatenated_examples.items()
+        }
+        return result
+
+    # Note that with `batched=True`, this map processes 1,000 texts together, so group_texts throws away a
+    # remainder for each of those groups of 1,000 texts. You can adjust that batch_size here but a higher value
+    # might be slower to preprocess.
+    #
+    # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
+    # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+    tokenized_datasets = tokenized_datasets.map(
+        group_texts,
+        batched=True,
+        num_proc=data_args.preprocessing_num_workers,
+        load_from_cache_file=not data_args.overwrite_cache,
+    )
+
+    # Enable tensorboard only on the master node
+    has_tensorboard = is_tensorboard_available()
+    if has_tensorboard and jax.process_index() == 0:
+        try:
+            from flax.metrics.tensorboard import SummaryWriter
+
+            summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
+        except ImportError as ie:
+            has_tensorboard = False
+            logger.warning(
+                f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
+            )
+    else:
+        logger.warning(
+            "Unable to display metrics through TensorBoard because the package is not installed: "
+            "Please run pip install tensorboard to enable."
+        )
+
+    # Initialize our training
+    rng = jax.random.PRNGKey(training_args.seed)
+    dropout_rngs = jax.random.split(rng, jax.local_device_count())
+
+    if model_args.model_name_or_path:
+        model = FlaxBartForConditionalGeneration.from_pretrained(
+            model_args.model_name_or_path,
+            config=config,
+            seed=training_args.seed,
+            dtype=getattr(jnp, model_args.dtype),
+            use_auth_token=True if model_args.use_auth_token else None,
+        )
+    else:
+        config.vocab_size = len(tokenizer)
+        model = FlaxBartForConditionalGeneration(
+            config,
+            seed=training_args.seed,
+            dtype=getattr(jnp, model_args.dtype),
+        )
+
+    # Data collator
+    # This one will take care of randomly masking the tokens and permuting the sentences.
+    data_collator = FlaxDataCollatorForBartDenoisingLM(
+        tokenizer=tokenizer,
+        decoder_start_token_id=model.config.decoder_start_token_id,
+        mask_ratio=data_args.mlm_probability,
+        poisson_lambda=data_args.poisson_lambda,
+        permute_sentence_ratio=data_args.permute_sentence_ratio,
+    )
+
+    # Store some constant
+    num_epochs = int(training_args.num_train_epochs)
+    train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
+    per_device_eval_batch_size = int(training_args.per_device_eval_batch_size)
+    eval_batch_size = per_device_eval_batch_size * jax.device_count()
+
+    num_train_steps = len(tokenized_datasets["train"]) // train_batch_size * num_epochs
+
+    # Create learning rate schedule
+    warmup_fn = optax.linear_schedule(
+        init_value=0.0, end_value=training_args.learning_rate, transition_steps=training_args.warmup_steps
+    )
+    decay_fn = optax.linear_schedule(
+        init_value=training_args.learning_rate,
+        end_value=0,
+        transition_steps=num_train_steps - training_args.warmup_steps,
+    )
+    linear_decay_lr_schedule_fn = optax.join_schedules(
+        schedules=[warmup_fn, decay_fn], boundaries=[training_args.warmup_steps]
+    )
+
+    # We use Optax's "masking" functionality to not apply weight decay
+    # to bias and LayerNorm scale parameters. decay_mask_fn returns a
+    # mask boolean with the same structure as the parameters.
+    # The mask is True for parameters that should be decayed.
+    def decay_mask_fn(params):
+        flat_params = traverse_util.flatten_dict(params)
+        # find out all LayerNorm parameters
+        layer_norm_candidates = ["layernorm", "layer_norm", "ln"]
+        layer_norm_named_params = set(
+            [
+                layer[-2:]
+                for layer_norm_name in layer_norm_candidates
+                for layer in flat_params.keys()
+                if layer_norm_name in "".join(layer).lower()
+            ]
+        )
+        flat_mask = {path: (path[-1] != "bias" and path[-2:] not in layer_norm_named_params) for path in flat_params}
+        return traverse_util.unflatten_dict(flat_mask)
+
+    # create adam optimizer
+    if training_args.adafactor:
+        # We use the default parameters here to initialize adafactor,
+        # For more details about the parameters please check https://github.com/deepmind/optax/blob/ed02befef9bf81cbbf236be3d2b0e032e9ed4a40/optax/_src/alias.py#L74
+        optimizer = optax.adafactor(
+            learning_rate=linear_decay_lr_schedule_fn,
+        )
+    else:
+        optimizer = optax.adamw(
+            learning_rate=linear_decay_lr_schedule_fn,
+            b1=training_args.adam_beta1,
+            b2=training_args.adam_beta2,
+            weight_decay=training_args.weight_decay,
+            mask=decay_mask_fn,
+        )
+
+    # Setup train state
+    state = train_state.TrainState.create(apply_fn=model.__call__, params=model.params, tx=optimizer)
+
+    # Define gradient update step fn
+    def train_step(state, batch, dropout_rng):
+        dropout_rng, new_dropout_rng = jax.random.split(dropout_rng)
+
+        def loss_fn(params):
+            labels = batch.pop("labels")
+
+            logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
+
+            # compute loss, ignore padded input tokens and special tokens
+            label_mask = jnp.where(labels > 0, 1.0, 0.0)
+            loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])) * label_mask
+
+            # take average
+            loss = loss.sum() / label_mask.sum()
+
+            return loss
+
+        grad_fn = jax.value_and_grad(loss_fn)
+        loss, grad = grad_fn(state.params)
+        grad = jax.lax.pmean(grad, "batch")
+        new_state = state.apply_gradients(grads=grad)
+
+        metrics = jax.lax.pmean(
+            {"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}, axis_name="batch"
+        )
+
+        return new_state, metrics, new_dropout_rng
+
+    # Create parallel version of the train step
+    p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
+
+    # Define eval fn
+    def eval_step(params, batch):
+        labels = batch.pop("labels")
+
+        logits = model(**batch, params=params, train=False)[0]
+
+        # compute loss, ignore padded input tokens and special tokens
+        label_mask = jnp.where(labels > 0, 1.0, 0.0)
+        loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])) * label_mask
+
+        # compute accuracy
+        accuracy = jnp.equal(jnp.argmax(logits, axis=-1), labels) * label_mask
+
+        # summarize metrics
+        metrics = {"loss": loss.sum(), "accuracy": accuracy.sum(), "normalizer": label_mask.sum()}
+        metrics = jax.lax.psum(metrics, axis_name="batch")
+
+        return metrics
+
+    p_eval_step = jax.pmap(eval_step, "batch", donate_argnums=(0,))
+
+    # Replicate the train state on each device
+    state = jax_utils.replicate(state)
+
+    train_time = 0
+    epochs = tqdm(range(num_epochs), desc="Epoch ... ", position=0)
+    for epoch in epochs:
+        # ======================== Training ================================
+        train_start = time.time()
+        train_metrics = []
+
+        # Create sampling rng
+        rng, input_rng = jax.random.split(rng)
+
+        # Generate an epoch by shuffling sampling indices from the train dataset
+        num_train_samples = len(tokenized_datasets["train"])
+        # Avoid using jax.numpy here in case of TPU training
+        train_samples_idx = np.random.permutation(np.arange(num_train_samples))
+        train_batch_idx = generate_batch_splits(train_samples_idx, train_batch_size)
+
+        # Gather the indexes for creating the batch and do a training step
+        for step, batch_idx in enumerate(tqdm(train_batch_idx, desc="Training...", position=1)):
+            samples = [tokenized_datasets["train"][int(idx)] for idx in batch_idx]
+            model_inputs = data_collator(samples)
+
+            # Model forward
+            model_inputs = shard(model_inputs.data)
+            state, train_metric, dropout_rngs = p_train_step(state, model_inputs, dropout_rngs)
+            train_metrics.append(train_metric)
+
+            cur_step = epoch * (num_train_samples // train_batch_size) + step
+
+            if cur_step % training_args.logging_steps == 0 and cur_step > 0:
+                # Save metrics
+                train_metric = jax_utils.unreplicate(train_metric)
+                train_time += time.time() - train_start
+                if has_tensorboard and jax.process_index() == 0:
+                    write_train_metric(summary_writer, train_metrics, train_time, cur_step)
+
+                epochs.write(
+                    f"Step... ({cur_step} | Loss: {train_metric['loss']}, Learning Rate:"
+                    f" {train_metric['learning_rate']})"
+                )
+
+                train_metrics = []
+
+            if cur_step % training_args.eval_steps == 0 and cur_step > 0:
+                # ======================== Evaluating ==============================
+                num_eval_samples = len(tokenized_datasets["validation"])
+                # Avoid using jax.numpy here in case of TPU training
+                eval_samples_idx = np.arange(num_eval_samples)
+                eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size, drop_last=False)
+
+                eval_metrics = []
+                for i, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
+                    samples = [tokenized_datasets["validation"][int(idx)] for idx in batch_idx]
+                    model_inputs = data_collator(samples)
+
+                    # Model forward
+                    metrics = pad_shard_unpad(p_eval_step, static_return=True)(
+                        state.params, model_inputs.data, min_device_batch=per_device_eval_batch_size
+                    )
+                    eval_metrics.append(metrics)
+
+                # normalize eval metrics
+                eval_metrics = get_metrics(eval_metrics)
+                eval_metrics = jax.tree_map(jnp.sum, eval_metrics)
+                eval_normalizer = eval_metrics.pop("normalizer")
+                eval_metrics = jax.tree_map(lambda x: x / eval_normalizer, eval_metrics)
+
+                # Update progress bar
+                epochs.desc = f"Step... ({cur_step} | Loss: {eval_metrics['loss']}, Acc: {eval_metrics['accuracy']})"
+
+                # Save metrics
+                if has_tensorboard and jax.process_index() == 0:
+                    write_eval_metric(summary_writer, eval_metrics, cur_step)
+
+            if cur_step % training_args.save_steps == 0 and cur_step > 0:
+                # save checkpoint after each epoch and push checkpoint to the hub
+                if jax.process_index() == 0:
+                    params = jax.device_get(jax.tree_map(lambda x: x[0], state.params))
+                    model.save_pretrained(training_args.output_dir, params=params)
+                    tokenizer.save_pretrained(training_args.output_dir)
+                    if training_args.push_to_hub:
+                        repo.push_to_hub(commit_message=f"Saving weights and logs of step {cur_step}", blocking=False)
+
+    # Eval after training
+    if training_args.do_eval:
+        num_eval_samples = len(tokenized_datasets["validation"])
+        # Avoid using jax.numpy here in case of TPU training
+        eval_samples_idx = np.arange(num_eval_samples)
+        eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size, drop_last=False)
+
+        eval_metrics = []
+        for _, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
+            samples = [tokenized_datasets["validation"][int(idx)] for idx in batch_idx]
+            model_inputs = data_collator(samples)
+
+            # Model forward
+            metrics = pad_shard_unpad(p_eval_step, static_return=True)(
+                state.params, model_inputs.data, min_device_batch=per_device_eval_batch_size
+            )
+            eval_metrics.append(metrics)
+
+        # normalize eval metrics
+        eval_metrics = get_metrics(eval_metrics)
+        eval_metrics = jax.tree_map(lambda metric: jnp.sum(metric).item(), eval_metrics)
+        eval_normalizer = eval_metrics.pop("normalizer")
+        eval_metrics = jax.tree_map(lambda x: x / eval_normalizer, eval_metrics)
+
+        try:
+            perplexity = math.exp(eval_metrics["loss"])
+        except OverflowError:
+            perplexity = float("inf")
+        eval_metrics["perplexity"] = perplexity
+
+        if jax.process_index() == 0:
+            eval_metrics = {f"eval_{metric_name}": value for metric_name, value in eval_metrics.items()}
+            path = os.path.join(training_args.output_dir, "eval_results.json")
+            with open(path, "w") as f:
+                json.dump(eval_metrics, f, indent=4, sort_keys=True)
+
+
+if __name__ == "__main__":
+    main()

examples/flax/language-modeling/run_mlm_flax.py

@@ -638,7 +638,6 @@ def main():
             config,
             seed=training_args.seed,
             dtype=getattr(jnp, model_args.dtype),
-            use_auth_token=True if model_args.use_auth_token else None,
         )
 
     # Store some constant

examples/flax/language-modeling/run_t5_mlm_flax.py

@@ -327,7 +327,7 @@ class FlaxDataCollatorForT5MLM:
     pad_token_id: int
     decoder_start_token_id: int
 
-    def __call__(self, examples: List[Dict[str, np.ndarray]]) -> Dict[str, np.ndarray]:
+    def __call__(self, examples: List[Dict[str, np.ndarray]]) -> BatchEncoding:
 
         # convert list to dict and tensorize input
         batch = BatchEncoding(
@@ -746,7 +746,6 @@ def main():
             config,
             seed=training_args.seed,
             dtype=getattr(jnp, model_args.dtype),
-            use_auth_token=True if model_args.use_auth_token else None,
         )
 
     # Data collator