Training in progress, step 500

.gitattributes

@@ -26,5 +26,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
-wandb/run-20220503_171959-a6039xud/run-a6039xud.wandb filter=lfs diff=lfs merge=lfs -text
-wandb/run-20220503_171959-a6039xud/files/output.log filter=lfs diff=lfs merge=lfs -text
+wandb/run-20220504_142129-1tmxz74i/run-1tmxz74i.wandb filter=lfs diff=lfs merge=lfs -text

config.json

@@ -182,7 +182,7 @@
     "forced_eos_token_id": null,
     "gradient_checkpointing": false,
     "hidden_act": "gelu",
-    "hidden_dropout": 0.06862889720223829,
+    "hidden_dropout": 0.035938233699532036,
     "hidden_size": 1024,
     "id2label": {
       "0": "LABEL_0",

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4d0b1b51e27450f198f132a91c3b3413103c76e8e42117405a269af001d2f25d
+oid sha256:3169ffb69bc1bf8ff1d3e2d588851055b41795e3c440682f891ae9cfb4eb653a
 size 2353867057

run_xtreme_s.py

@@ -1,948 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-# Copyright 2022 The HuggingFace Inc. 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
-
-""" Fine-tuning a 🤗 Transformers pretrained speech model on the XTREME-S benchmark tasks"""
-
-import json
-import logging
-import os
-import re
-import sys
-from collections import OrderedDict, defaultdict
-from dataclasses import dataclass, field
-from typing import Dict, List, Optional, Union
-
-import datasets
-import numpy as np
-import torch
-from datasets import DatasetDict, load_dataset, load_metric
-
-import transformers
-from transformers import (
-    AutoConfig,
-    AutoFeatureExtractor,
-    AutoModelForAudioClassification,
-    AutoModelForCTC,
-    AutoModelForSpeechSeq2Seq,
-    AutoProcessor,
-    AutoTokenizer,
-    HfArgumentParser,
-    Seq2SeqTrainer,
-    Seq2SeqTrainingArguments,
-    SpeechEncoderDecoderModel,
-    Trainer,
-    set_seed,
-)
-from transformers.trainer_utils import get_last_checkpoint, is_main_process
-from transformers.utils import check_min_version
-from transformers.utils.versions import require_version
-
-
-# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.18.0.dev0")
-
-require_version("datasets>=1.18.0", "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt")
-
-
-logger = logging.getLogger(__name__)
-
-
-def list_field(default=None, metadata=None):
-    return field(default_factory=lambda: default, metadata=metadata)
-
-
-TASK_TO_TARGET_COLUMN_NAME = {
-    "fleurs-asr": "transcription",
-    "fleurs-lang_id": "lang_id",
-    "mls": "transcription",
-    "voxpopuli": "transcription",
-    "covost2": "translation",
-    "minds14": "intent_class",
-    "babel": "transcription",
-}
-
-
-@dataclass
-class ModelArguments:
-    """
-    Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
-    """
-
-    model_name_or_path: str = field(
-        metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"}
-    )
-    tokenizer_name_or_path: Optional[str] = field(
-        default=None,
-        metadata={"help": "Path to pretrained tokenizer or tokenizer identifier from huggingface.co/models"},
-    )
-    cache_dir: Optional[str] = field(
-        default=None,
-        metadata={
-            "help": "Where do you want to store the pretrained models and datasets downloaded from " "huggingface.co"
-        },
-    )
-    freeze_feature_encoder: bool = field(
-        default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
-    )
-    attention_dropout: float = field(
-        default=0.0, metadata={"help": "The dropout ratio for the attention probabilities."}
-    )
-    activation_dropout: float = field(
-        default=0.0, metadata={"help": "The dropout ratio for activations inside the fully connected layer."}
-    )
-    feat_proj_dropout: float = field(default=0.0, metadata={"help": "The dropout ratio for the projected features."})
-    hidden_dropout: float = field(
-        default=0.0,
-        metadata={
-            "help": "The dropout probability for all fully connected layers in the embeddings, encoder, and pooler."
-        },
-    )
-    final_dropout: float = field(
-        default=0.0,
-        metadata={"help": "The dropout probability for the final projection layer."},
-    )
-    mask_time_prob: float = field(
-        default=0.05,
-        metadata={
-            "help": "Probability of each feature vector along the time axis to be chosen as the start of the vector"
-            "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
-            "vectors will be masked along the time axis."
-        },
-    )
-    mask_time_length: int = field(
-        default=10,
-        metadata={"help": "Length of vector span to mask along the time axis."},
-    )
-    mask_feature_prob: float = field(
-        default=0.0,
-        metadata={
-            "help": "Probability of each feature vector along the feature axis to be chosen as the start of the vector"
-            "span to be masked. Approximately ``mask_feature_prob * sequence_length // mask_feature_length`` feature bins will be masked along the time axis."
-        },
-    )
-    mask_feature_length: int = field(
-        default=10,
-        metadata={"help": "Length of vector span to mask along the feature axis."},
-    )
-    layerdrop: float = field(default=0.0, metadata={"help": "The LayerDrop probability."})
-    ctc_zero_infinity: bool = field(
-        default=False,
-        metadata={"help": "Whether to zero infinite losses and the associated gradients of `torch.nn.CTCLoss`."},
-    )
-    ctc_loss_reduction: Optional[str] = field(
-        default="mean", metadata={"help": "The way the ctc loss should be reduced. Should be one of 'mean' or 'sum'."}
-    )
-
-
-@dataclass
-class DataTrainingArguments:
-    """
-    Arguments pertaining to what data we are going to input our model for training and eval.
-
-    Using `HfArgumentParser` we can turn this class
-    into argparse arguments to be able to specify them on
-    the command line.
-    """
-
-    dataset_name: str = field(
-        default="google/xtreme_s",
-        metadata={"help": "The name of the dataset to use (via the datasets library). Defaults to 'google/xtreme_s'"},
-    )
-    task: str = field(
-        default=None,
-        metadata={
-            "help": "The task name of the benchmark to use (via the datasets library). Should be on of: "
-            "'fleurs-asr', 'mls', 'voxpopuli', 'covost2', 'minds14', 'fleurs-lang_id', 'babel'."
-        },
-    )
-    language: str = field(
-        default="all",
-        metadata={"help": "The language id as defined in the datasets config name or `all` for all languages."},
-    )
-    language_group: str = field(
-        default=None,
-        metadata={
-            "help": "The language group to select a subset of languages to train on. "
-            "This option is only used the 'fleurs-asr' task. Should be one of: "
-            "'western_european_we', 'eastern_european_ee', 'central_asia_middle_north_african_cmn', "
-            "'sub_saharan_african_ssa', 'south_asian_sa', 'south_east_asian_sea', 'chinese_japanase_korean_cjk'."
-        },
-    )
-    train_split_name: str = field(
-        default="train",
-        metadata={
-            "help": "The name of the training dataset split to use (via the datasets library). Defaults to 'train'"
-        },
-    )
-    eval_split_name: str = field(
-        default="validation",
-        metadata={
-            "help": "The name of the evaluation dataset split to use (via the datasets library). "
-            "Defaults to 'validation'"
-        },
-    )
-    predict_split_name: str = field(
-        default="test",
-        metadata={
-            "help": "The name of the prediction dataset split to use (via the datasets library). " "Defaults to 'test'"
-        },
-    )
-    audio_column_name: str = field(
-        default="audio",
-        metadata={"help": "The name of the dataset column containing the audio data. Defaults to 'audio'"},
-    )
-    target_column_name: str = field(
-        default=None,
-        metadata={
-            "help": "The name of the dataset column containing the target data "
-            "(transcription/translation/label). If None, the name will be inferred from the task. Defaults to None."
-        },
-    )
-    overwrite_cache: bool = field(
-        default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
-    )
-    preprocessing_num_workers: Optional[int] = field(
-        default=None,
-        metadata={"help": "The number of processes to use for the preprocessing."},
-    )
-    max_train_samples: Optional[int] = field(
-        default=None,
-        metadata={
-            "help": "For debugging purposes or quicker training, truncate the number of training examples to this "
-            "value if set."
-        },
-    )
-    max_eval_samples: Optional[int] = field(
-        default=None,
-        metadata={
-            "help": "For debugging purposes or quicker training, truncate the number of validation examples to this "
-            "value if set."
-        },
-    )
-    max_predict_samples: Optional[int] = field(
-        default=None,
-        metadata={
-            "help": "For debugging purposes or quicker training, truncate the number of prediction examples to this "
-            "value if set."
-        },
-    )
-    chars_to_ignore: Optional[List[str]] = list_field(
-        default=', ? . ! - ; : " “ % ‘ ” �'.split(" "),
-        metadata={"help": "A list of characters to remove from the transcripts."},
-    )
-    max_duration_in_seconds: float = field(
-        default=30.0,
-        metadata={
-            "help": "Filter audio files that are longer than `max_duration_in_seconds` seconds to 'max_duration_in_seconds`"
-        },
-    )
-    min_duration_in_seconds: float = field(
-        default=0.0, metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"}
-    )
-    preprocessing_only: bool = field(
-        default=False,
-        metadata={
-            "help": "Whether to only do data preprocessing and skip training. "
-            "This is especially useful when data preprocessing errors out in distributed training due to timeout. "
-            "In this case, one should run the preprocessing in a non-distributed setup with `preprocessing_only=True` "
-            "so that the cached datasets can consequently be loaded in distributed training"
-        },
-    )
-    use_auth_token: bool = field(
-        default=False,
-        metadata={
-            "help": "If :obj:`True`, will use the token generated when running"
-            ":obj:`transformers-cli login` as HTTP bearer authorization for remote files."
-        },
-    )
-    unk_token: str = field(
-        default="[UNK]",
-        metadata={"help": "The unk token for the tokenizer"},
-    )
-    pad_token: str = field(
-        default="[PAD]",
-        metadata={"help": "The padding token for the tokenizer"},
-    )
-    word_delimiter_token: str = field(
-        default="|",
-        metadata={"help": "The word delimiter token for the tokenizer"},
-    )
-    phoneme_language: Optional[str] = field(
-        default=None,
-        metadata={
-            "help": "The target language that should be used be"
-            " passed to the tokenizer for tokenization. Note that"
-            " this is only relevant if the model classifies the"
-            " input audio to a sequence of phoneme sequences."
-        },
-    )
-    per_lang_metrics: bool = field(
-        default=True,
-        metadata={
-            "help": "If `True`, compute the test metrics separately for each language, and average the results. "
-            "If `False` compute the average test metrics in a single pass for all languages at once."
-        },
-    )
-
-
-@dataclass
-class SpeechDataCollatorWithPadding:
-
-    processor: AutoProcessor
-    decoder_start_token_id: Optional[int] = None
-    padding: Union[bool, str] = "longest"
-    pad_labels: Optional[int] = True
-    pad_to_multiple_of: Optional[int] = None
-    pad_to_multiple_of_labels: Optional[int] = None
-
-    def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
-        # split inputs and labels since they have to be of different lenghts and need
-        # different padding methods
-        input_features = [{"input_values": feature["input_values"]} for feature in features]
-
-        batch = self.processor.pad(
-            input_features,
-            padding=self.padding,
-            pad_to_multiple_of=self.pad_to_multiple_of,
-            return_tensors="pt",
-        )
-
-        if self.pad_labels:
-            label_features = [{"input_ids": feature["labels"]} for feature in features]
-            with self.processor.as_target_processor():
-                labels_batch = self.processor.pad(
-                    label_features,
-                    padding=self.padding,
-                    pad_to_multiple_of=self.pad_to_multiple_of_labels,
-                    return_tensors="pt",
-                )
-
-            # replace padding with -100 to ignore loss correctly
-            labels = labels_batch["input_ids"].masked_fill(labels_batch.attention_mask.ne(1), -100)
-
-            # if bos token is appended in previous tokenization step,
-            # cut bos token here as it's append later anyways
-            if (
-                self.decoder_start_token_id is not None
-                and (labels[:, 0] == self.decoder_start_token_id).all().cpu().item()
-            ):
-                labels = labels[:, 1:]
-
-            batch["labels"] = labels
-        else:
-            batch["labels"] = torch.tensor([feature["labels"] for feature in features])
-
-        return batch
-
-
-def create_vocabulary_from_data(
-    datasets: DatasetDict,
-    word_delimiter_token: Optional[str] = None,
-    unk_token: Optional[str] = None,
-    pad_token: Optional[str] = None,
-):
-    # Given training and test labels create vocabulary
-    def extract_all_chars(batch):
-        all_text = " ".join(batch["target_text"])
-        vocab = list(set(all_text))
-        return {"vocab": [vocab], "all_text": [all_text]}
-
-    vocabs = datasets.map(
-        extract_all_chars,
-        batched=True,
-        batch_size=-1,
-        keep_in_memory=True,
-        remove_columns=datasets["train"].column_names,
-    )
-
-    # take union of all unique characters in each dataset
-    vocab_set = (
-        (set(vocabs["train"]["vocab"][0]) if "train" in vocabs else set())
-        | (set(vocabs["eval"]["vocab"][0]) if "eval" in vocabs else set())
-        | (set(vocabs["predict"]["vocab"][0]) if "predict" in vocabs else set())
-    )
-
-    vocab_dict = {v: k for k, v in enumerate(sorted(list(vocab_set)))}
-
-    # replace white space with delimiter token
-    if word_delimiter_token is not None:
-        vocab_dict[word_delimiter_token] = vocab_dict[" "]
-        del vocab_dict[" "]
-
-    # add unk and pad token
-    if unk_token is not None:
-        vocab_dict[unk_token] = len(vocab_dict)
-
-    if pad_token is not None:
-        vocab_dict[pad_token] = len(vocab_dict)
-
-    return vocab_dict
-
-
-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, Seq2SeqTrainingArguments))
-    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()
-
-    # Detecting last checkpoint.
-    last_checkpoint = None
-    if os.path.isdir(training_args.output_dir) and training_args.do_train and not training_args.overwrite_output_dir:
-        last_checkpoint = get_last_checkpoint(training_args.output_dir)
-        if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0:
-            raise ValueError(
-                f"Output directory ({training_args.output_dir}) already exists and is not empty. "
-                "Use --overwrite_output_dir to overcome."
-            )
-        elif last_checkpoint is not None:
-            logger.info(
-                f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
-                "the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
-            )
-
-    # Setup logging
-    logging.basicConfig(
-        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
-        datefmt="%m/%d/%Y %H:%M:%S",
-        handlers=[logging.StreamHandler(sys.stdout)],
-    )
-    logger.setLevel(logging.INFO if is_main_process(training_args.local_rank) else logging.WARN)
-
-    # Log on each process the small summary:
-    logger.warning(
-        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
-        f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
-    )
-    # Set the verbosity to info of the Transformers logger (on main process only):
-    if is_main_process(training_args.local_rank):
-        transformers.utils.logging.set_verbosity_info()
-    logger.info("Training/evaluation parameters %s", training_args)
-
-    # Set seed before initializing model.
-    set_seed(training_args.seed)
-
-    # 1. First, let's load the dataset
-    raw_datasets = DatasetDict()
-    task_name = data_args.task
-    lang_id = data_args.language
-
-    if task_name is None:
-        raise ValueError(
-            "Set --task should be set to '<xtreme_s_task>' " "(e.g. 'fleurs-asr', 'mls', 'covost2', 'minds14') "
-        )
-    if lang_id is None:
-        raise ValueError(
-            "Set --language should be set to the language id of the sub dataset "
-            "config to be used (e.g. 'pl', 'en.tr', 'fr-FR') or 'all'"
-            " for multi-lingual fine-tuning."
-        )
-    if data_args.language_group is not None:
-        if data_args.task != "fleurs-asr":
-            raise ValueError("--language_group should only be used with --task=fleurs-asr")
-        if data_args.language != "all":
-            raise ValueError("--language_group should only be used with --language=all")
-
-    if data_args.target_column_name is None:
-        target_column_name = TASK_TO_TARGET_COLUMN_NAME[task_name]
-    else:
-        target_column_name = data_args.target_column_name
-
-    # here we differentiate between tasks with text as the target and classification tasks
-    is_text_target = target_column_name in ("transcription", "translation")
-
-    config_name = ".".join([task_name.split("-")[0], lang_id])
-
-    if training_args.do_train:
-        raw_datasets["train"] = load_dataset(
-            data_args.dataset_name,
-            config_name,
-            split=data_args.train_split_name,
-            use_auth_token=data_args.use_auth_token,
-            cache_dir=model_args.cache_dir,
-        )
-
-        if data_args.audio_column_name not in raw_datasets["train"].column_names:
-            raise ValueError(
-                f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.dataset_name}'. "
-                "Make sure to set `--audio_column_name` to the correct audio column - one of "
-                f"{', '.join(raw_datasets['train'].column_names)}."
-            )
-
-        if target_column_name not in raw_datasets["train"].column_names:
-            raise ValueError(
-                f"--target_column_name {target_column_name} not found in dataset '{data_args.dataset_name}'. "
-                "Make sure to set `--target_column_name` to the correct text column - one of "
-                f"{', '.join(raw_datasets['train'].column_names)}."
-            )
-
-        if data_args.max_train_samples is not None:
-            raw_datasets["train"] = raw_datasets["train"].select(range(data_args.max_train_samples))
-
-    if training_args.do_eval:
-        raw_datasets["eval"] = load_dataset(
-            data_args.dataset_name,
-            config_name,
-            split=data_args.eval_split_name,
-            use_auth_token=data_args.use_auth_token,
-            cache_dir=model_args.cache_dir,
-        )
-
-        if data_args.max_eval_samples is not None:
-            raw_datasets["eval"] = raw_datasets["eval"].select(range(data_args.max_eval_samples))
-
-    if training_args.do_predict:
-        raw_datasets["predict"] = load_dataset(
-            data_args.dataset_name,
-            config_name,
-            split=data_args.predict_split_name,
-            use_auth_token=data_args.use_auth_token,
-            cache_dir=model_args.cache_dir,
-        )
-
-        if data_args.max_predict_samples is not None:
-            raw_datasets["predict"] = raw_datasets["predict"].select(range(data_args.max_predict_samples))
-
-    lang_list = next(iter(raw_datasets.values())).features["lang_id"].names
-    if not is_text_target:
-        label_list = next(iter(raw_datasets.values())).features[target_column_name].names
-        num_labels = len(label_list)
-
-    num_workers = data_args.preprocessing_num_workers
-
-    lang_group = data_args.language_group
-    if lang_group is not None:
-        with training_args.main_process_first(desc="language group filter"):
-            lang_group_id = next(iter(raw_datasets.values())).features["lang_group_id"].str2int(lang_group)
-            raw_datasets = raw_datasets.filter(
-                lambda lang_group: lang_group == lang_group_id,
-                num_proc=num_workers,
-                input_columns=["lang_group_id"],
-            )
-
-    # 2. We remove some special characters from the datasets
-    # that make training complicated and do not help in transcribing the speech
-    # E.g. characters, such as `,` and `.` do not really have an acoustic characteristic
-    # that could be easily picked up by the model
-    chars_to_ignore_regex = (
-        f'[{"".join(data_args.chars_to_ignore)}]' if data_args.chars_to_ignore is not None else None
-    )
-
-    def remove_special_characters(batch):
-        if chars_to_ignore_regex is not None:
-            batch["target_text"] = re.sub(chars_to_ignore_regex, "", batch[target_column_name]).lower()
-        else:
-            batch["target_text"] = batch[target_column_name].lower()
-        return batch
-
-    if is_text_target:
-        with training_args.main_process_first(desc="dataset map special characters removal"):
-            raw_datasets = raw_datasets.map(
-                remove_special_characters,
-                remove_columns=[target_column_name],
-                desc="remove special characters from datasets",
-            )
-
-        # save special tokens for tokenizer
-        word_delimiter_token = data_args.word_delimiter_token
-        unk_token = data_args.unk_token
-        pad_token = data_args.pad_token
-
-
-    encoder_id = "facebook/wav2vec2-xls-r-300m"
-    decoder_id = "facebook/bart-large"
-
-    model = SpeechEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_id, decoder_id, encoder_add_adapter=True)
-    model.config.encoder.feat_proj_dropout = 0.0
-    model.config.encoder.final_dropout = 0.0
-    model.config.encoder.mask_time_prob = 0.1
-    model.config.decoder_start_token_id = model.decoder.config.bos_token_id
-    model.config.pad_token_id = model.decoder.config.pad_token_id
-    model.config.eos_token_id = model.decoder.config.eos_token_id
-    model.config.max_length = 40
-    model.config.num_beams = 1
-    model.config.encoder.layerdrop = 0.0
-    model.config.use_cache = False
-    model.config.processor_class = "Wav2Vec2Processor"
-
-    model.save_pretrained(model_args.model_name_or_path)
-
-    feature_etxractor = AutoFeatureExtractor.from_pretrained(encoder_id)
-    feature_etxractor.save_pretrained(model_args.model_name_or_path)
-    tokenizer = AutoTokenizer.from_pretrained(decoder_id)
-    tokenizer.save_pretrained(model_args.model_name_or_path)
-    
-    # 3. Next, let's load the config as we might need it to create
-    # the tokenizer
-    config = AutoConfig.from_pretrained(
-        model_args.model_name_or_path, cache_dir=model_args.cache_dir, use_auth_token=data_args.use_auth_token
-    )
-
-    if is_text_target:
-        # 4. (Optional, for ASR and translation) If no tokenizer file is defined,
-        # we create the vocabulary of the model by extracting all unique characters from
-        # the training and evaluation datasets
-        # We need to make sure that only first rank saves vocabulary
-        # make sure all processes wait until vocab is created
-        tokenizer_name_or_path = model_args.tokenizer_name_or_path
-        tokenizer_kwargs = {}
-        if tokenizer_name_or_path is None:
-            # save vocab in training output dir
-            tokenizer_name_or_path = training_args.output_dir
-
-            vocab_file = os.path.join(tokenizer_name_or_path, "vocab.json")
-
-            with training_args.main_process_first():
-                if training_args.overwrite_output_dir and os.path.isfile(vocab_file):
-                    os.remove(vocab_file)
-
-            with training_args.main_process_first(desc="dataset map vocabulary creation"):
-                if not os.path.isfile(vocab_file):
-                    os.makedirs(tokenizer_name_or_path, exist_ok=True)
-                    vocab_dict = create_vocabulary_from_data(
-                        raw_datasets,
-                        word_delimiter_token=word_delimiter_token,
-                        unk_token=unk_token,
-                        pad_token=pad_token,
-                    )
-
-                    # save vocab dict to be loaded into tokenizer
-                    with open(vocab_file, "w") as file:
-                        json.dump(vocab_dict, file)
-
-            # if tokenizer has just been created
-            # it is defined by `tokenizer_class` if present in config else by `model_type`
-            if not config.is_encoder_decoder:
-                tokenizer_kwargs = {
-                    "config": config if config.tokenizer_class is not None else None,
-                    "tokenizer_type": config.model_type if config.tokenizer_class is None else None,
-                    "unk_token": unk_token,
-                    "pad_token": pad_token,
-                    "word_delimiter_token": word_delimiter_token,
-                }
-            else:
-                tokenizer_kwargs = {}
-
-    # 5. Now we can instantiate the feature extractor, tokenizer and model
-    # Note for distributed training, the .from_pretrained methods guarantee that only
-    # one local process can concurrently download model & vocab.
-
-    # load feature_extractor and tokenizer
-    if is_text_target:
-        tokenizer = AutoTokenizer.from_pretrained(
-            tokenizer_name_or_path,
-            use_auth_token=data_args.use_auth_token,
-            **tokenizer_kwargs,
-        )
-    feature_extractor = AutoFeatureExtractor.from_pretrained(
-        model_args.model_name_or_path, cache_dir=model_args.cache_dir, use_auth_token=data_args.use_auth_token
-    )
-
-    # adapt config
-    # (speech translation requires pre-configured seq2seq models)
-    if task_name != "covost2":
-        config.update(
-            {
-                "feat_proj_dropout": model_args.feat_proj_dropout,
-                "attention_dropout": model_args.attention_dropout,
-                "hidden_dropout": model_args.hidden_dropout,
-                "final_dropout": model_args.final_dropout,
-                "mask_time_prob": model_args.mask_time_prob,
-                "mask_time_length": model_args.mask_time_length,
-                "mask_feature_prob": model_args.mask_feature_prob,
-                "mask_feature_length": model_args.mask_feature_length,
-                "gradient_checkpointing": training_args.gradient_checkpointing,
-                "layerdrop": model_args.layerdrop,
-                "ctc_zero_infinity": model_args.ctc_zero_infinity,
-                "ctc_loss_reduction": model_args.ctc_loss_reduction,
-                "activation_dropout": model_args.activation_dropout,
-            }
-        )
-        if training_args.do_train:
-            if is_text_target:
-                config.pad_token_id = tokenizer.pad_token_id
-                config.vocab_size = len(tokenizer)
-            else:
-                label_to_id = {v: i for i, v in enumerate(label_list)}
-                config.label2id = label_to_id
-                config.id2label = {id: label for label, id in label_to_id.items()}
-                config.num_labels = num_labels
-    else:
-        config.encoder.update({"hidden_dropout": model_args.hidden_dropout})
-
-    # create model
-    if target_column_name == "transcription":
-        model = AutoModelForCTC.from_pretrained(
-            model_args.model_name_or_path,
-            cache_dir=model_args.cache_dir,
-            config=config,
-            use_auth_token=data_args.use_auth_token,
-        )
-    elif config.is_encoder_decoder:
-        model = AutoModelForSpeechSeq2Seq.from_pretrained(
-            model_args.model_name_or_path,
-            cache_dir=model_args.cache_dir,
-            config=config,
-            use_auth_token=data_args.use_auth_token,
-        )
-        if model.config.decoder_start_token_id is None:
-            raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined")
-    else:
-        model = AutoModelForAudioClassification.from_pretrained(
-            model_args.model_name_or_path,
-            cache_dir=model_args.cache_dir,
-            config=config,
-            use_auth_token=data_args.use_auth_token,
-        )
-
-    # freeze encoder
-    if model_args.freeze_feature_encoder:
-        model.freeze_feature_encoder()
-
-    # 6. Now we preprocess the datasets including loading the audio, resampling and normalization
-    # Thankfully, `datasets` takes care of automatically loading and resampling the audio,
-    # so that we just need to set the correct target sampling rate and normalize the input
-    # via the `feature_extractor`
-
-    # make sure that dataset decodes audio with correct sampling rate
-    dataset_sampling_rate = next(iter(raw_datasets.values())).features[data_args.audio_column_name].sampling_rate
-    if dataset_sampling_rate != feature_extractor.sampling_rate:
-        raw_datasets = raw_datasets.cast_column(
-            data_args.audio_column_name, datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
-        )
-
-    # derive max & min input length for sample rate & max duration
-    max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
-    min_input_length = data_args.min_duration_in_seconds * feature_extractor.sampling_rate
-    audio_column_name = data_args.audio_column_name
-
-    # `phoneme_language` is only relevant if the model is fine-tuned on phoneme classification
-    phoneme_language = data_args.phoneme_language
-
-    # Preprocessing the datasets.
-    # We need to read the audio files as arrays and tokenize the targets.
-    def prepare_dataset(batch):
-        # load audio
-        sample = batch[audio_column_name]
-
-        inputs = feature_extractor(sample["array"], sampling_rate=sample["sampling_rate"])
-        batch["input_values"] = inputs.input_values[0]
-        batch["length"] = len(batch["input_values"])
-
-        # encode targets
-        additional_kwargs = {}
-        if phoneme_language is not None:
-            additional_kwargs["phonemizer_lang"] = phoneme_language
-
-        if is_text_target:
-            batch["labels"] = tokenizer(batch["target_text"], **additional_kwargs).input_ids
-        else:
-            batch["labels"] = batch[target_column_name]
-
-        batch["lang"] = batch["lang_id"]
-
-        return batch
-
-    with training_args.main_process_first(desc="dataset map preprocessing"):
-        vectorized_datasets = raw_datasets.map(
-            prepare_dataset,
-            remove_columns=next(iter(raw_datasets.values())).column_names,
-            num_proc=num_workers,
-            desc="preprocess datasets",
-        )
-
-        if training_args.do_train:
-
-            def is_audio_in_length_range(length):
-                return length > min_input_length and length < max_input_length
-
-            # filter data that is shorter than min_input_length
-            vectorized_datasets["train"] = vectorized_datasets["train"].filter(
-                is_audio_in_length_range,
-                num_proc=num_workers,
-                input_columns=["length"],
-            )
-
-    # 7. Next, we can prepare for the training step.
-    # Let's use the appropriate XTREME-S evaluation metric,
-    # instantiate a data collator and the trainer
-
-    # Define evaluation metrics during training, *i.e.* word error rate, character error rate
-    eval_metric = load_metric("xtreme_s", task_name)
-
-    # for large datasets it is advised to run the preprocessing on a
-    # single machine first with ``args.preprocessing_only`` since there will mostly likely
-    # be a timeout when running the script in distributed mode.
-    # In a second step ``args.preprocessing_only`` can then be set to `False` to load the
-    # cached dataset
-    if data_args.preprocessing_only:
-        logger.info(f"Data preprocessing finished. Files cached at {vectorized_datasets.cache_files}")
-        return
-
-    def asr_logits_argmax(logits, labels):
-        return logits.argmax(dim=-1)
-
-    def compute_asr_metric(pred):
-        pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id
-
-        pred_str = tokenizer.batch_decode(pred.predictions)
-        # we do not want to group tokens when computing the metrics
-        label_str = tokenizer.batch_decode(pred.label_ids, group_tokens=False)
-
-        metric = eval_metric.compute(predictions=pred_str, references=label_str)
-        return metric
-
-    def compute_classification_metric(pred):
-        pred_ids = np.argmax(pred.predictions, axis=1)
-        metric = eval_metric.compute(predictions=pred_ids, references=pred.label_ids)
-        return metric
-
-    # Now save everything to be able to create a single processor later
-    if is_main_process(training_args.local_rank):
-        # save feature extractor, tokenizer and config
-        feature_extractor.save_pretrained(training_args.output_dir)
-        if is_text_target:
-            tokenizer.save_pretrained(training_args.output_dir)
-        config.save_pretrained(training_args.output_dir)
-    # wait until configs are saved in the main process before loading the processor
-    if training_args.local_rank != -1:
-        torch.distributed.barrier()
-
-    if is_text_target:
-        processor = AutoProcessor.from_pretrained(training_args.output_dir)
-    else:
-        processor = AutoFeatureExtractor.from_pretrained(training_args.output_dir)
-
-    # Instantiate custom data collator
-    data_collator = SpeechDataCollatorWithPadding(processor=processor, pad_labels=is_text_target)
-
-    # Initialize Trainer
-    if target_column_name == "translation":
-        trainer = Seq2SeqTrainer(
-            model=model,
-            data_collator=data_collator,
-            args=training_args,
-            preprocess_logits_for_metrics=asr_logits_argmax if training_args.predict_with_generate else None,
-            compute_metrics=compute_asr_metric if training_args.predict_with_generate else None,
-            train_dataset=vectorized_datasets["train"] if training_args.do_train else None,
-            eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
-            tokenizer=feature_extractor,
-        )
-    else:
-        trainer = Trainer(
-            model=model,
-            data_collator=data_collator,
-            args=training_args,
-            preprocess_logits_for_metrics=asr_logits_argmax if is_text_target else None,
-            compute_metrics=compute_asr_metric if is_text_target else compute_classification_metric,
-            train_dataset=vectorized_datasets["train"] if training_args.do_train else None,
-            eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
-            tokenizer=feature_extractor,
-        )
-
-    # 8. Finally, we can start training
-
-    # Training
-    if training_args.do_train:
-
-        # use last checkpoint if exist
-        if last_checkpoint is not None:
-            checkpoint = last_checkpoint
-        elif os.path.isdir(model_args.model_name_or_path):
-            checkpoint = model_args.model_name_or_path
-        else:
-            checkpoint = None
-
-        train_result = trainer.train(resume_from_checkpoint=checkpoint)
-        trainer.save_model()
-
-        metrics = train_result.metrics
-        max_train_samples = (
-            data_args.max_train_samples
-            if data_args.max_train_samples is not None
-            else len(vectorized_datasets["train"])
-        )
-        metrics["train_samples"] = min(max_train_samples, len(vectorized_datasets["train"]))
-
-        trainer.log_metrics("train", metrics)
-        trainer.save_metrics("train", metrics)
-        trainer.save_state()
-
-    # Evaluation on the test set
-    results = {}
-    if training_args.do_predict:
-        logger.info(f"*** Evaluating on the `{data_args.predict_split_name}` set ***")
-        if data_args.per_lang_metrics:
-            # separate the `test` dataset into language-specific subsets and compute metrics for each of them
-            metrics = {}
-            average_metrics = defaultdict(list)
-            for lang_id in range(len(lang_list)):
-                lang_name = lang_list[lang_id]
-                with training_args.main_process_first(desc="per-language dataset filter"):
-                    lang_dataset = vectorized_datasets["predict"].filter(
-                        lambda lang: lang == lang_id,
-                        num_proc=num_workers,
-                        input_columns=["lang"],
-                    )
-                lang_metrics = trainer.evaluate(lang_dataset)
-                redundant_metrics = ["eval_runtime", "eval_samples_per_second", "eval_steps_per_second", "eval_epoch"]
-                for metric_name, value in lang_metrics.items():
-                    average_metrics[metric_name].append(value)
-                    if metric_name not in redundant_metrics:
-                        metrics[f"{metric_name}_{lang_name}"] = value
-            for metric_name, value in average_metrics.items():
-                metrics[metric_name] = np.mean(value)
-        else:
-            metrics = trainer.evaluate(vectorized_datasets["predict"])
-        max_predict_samples = (
-            data_args.max_predict_samples
-            if data_args.max_predict_samples is not None
-            else len(vectorized_datasets["predict"])
-        )
-        metrics["predict_samples"] = min(max_predict_samples, len(vectorized_datasets["predict"]))
-
-        # make sure that the `predict` metrics end up in the log history for the model card
-        trainer.log(OrderedDict(sorted(metrics.items())))
-
-        trainer.log_metrics("predict", metrics)
-        trainer.save_metrics("predict", metrics)
-
-    # Write model card and (optionally) push to hub
-    kwargs = {
-        "finetuned_from": model_args.model_name_or_path,
-        "tasks": task_name,
-        "tags": [task_name, data_args.dataset_name],
-        "dataset_args": f"Config: {config_name}, Training split: {data_args.train_split_name}, Eval split: {data_args.eval_split_name}, Predict split: {data_args.predict_split_name}",
-        "dataset": f"{data_args.dataset_name.upper()} - {config_name.upper()}",
-        "language": data_args.language,
-    }
-
-    if training_args.push_to_hub:
-        trainer.push_to_hub(**kwargs)
-    else:
-        trainer.create_model_card(**kwargs)
-
-    return results
-
-
-if __name__ == "__main__":
-    main()
-

run_xtreme_s.py

@@ -0,0 +1 @@
+/home/sanchit_huggingface_co/run_xtreme_s.py

runs/May03_17-15-22_sanchit--v100/events.out.tfevents.1651598399.sanchit--v100.42111.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1fa91b570736d0c591447238dcb171cfb6539cad40d9ce858fec696f3afac2b0
-size 718842
+oid sha256:06ebac1e97e241e13d6c8a536ad57e3ef5b58c510633a0824a10536aae4662a4
+size 797661

wandb/run-20220503_171959-a6039xud/files/output.log → runs/May04_08-29-27_sanchit--v100/1651653030.564084/events.out.tfevents.1651653030.sanchit--v100.48541.1

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c54d8426faeee1b1757c840aed0c33fd5844c004f6e486f877f6deea1c7413fd
-size 10696139
+oid sha256:9f818ffc5324af0a19b67e9e869654bb112292ade31d00f48263ec9cf177206c
+size 5184

wandb/run-20220503_171959-a6039xud/run-a6039xud.wandb → runs/May04_08-29-27_sanchit--v100/events.out.tfevents.1651653030.sanchit--v100.48541.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:84cdd08f802f3f87244c71bf27046ed6bcab4d0121e9ce4ff18535e211468109
-size 492654377
+oid sha256:6120cfa0b8dd7cc6afe852da88002229769bc8f3efa93f730fa0661088c42d36
+size 88290

runs/May04_13-30-37_sanchit--v100/1651674088.8879716/events.out.tfevents.1651674088.sanchit--v100.50375.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aef34e54eaac03ee713590a0eb81b34a1eb839a803375bb6e9e3a6e439991c2b
+size 5184

runs/May04_13-30-37_sanchit--v100/events.out.tfevents.1651674088.sanchit--v100.50375.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b7bf3b5b1646e5c45627c0e0db5e11e84bd3dad703289f12a20b99abd0ab7cc
+size 88291

sweep.yaml

@@ -31,9 +31,9 @@ parameters:
   num_train_epochs:
     value: 3
   per_device_train_batch_size:
-    value: 4
+    value: 8
   per_device_eval_batch_size:
-    value: 4
+    value: 8
   gradient_accumulation_steps:
     value: 8
   generation_max_length:

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:94e05d5641aa58db52aee32293e970801889668d13a72d205ffa6fe324861d4b
+oid sha256:54311328073baaf4c9576b62de98f28bcc986e980c3c849a9409edd618d1e27f
 size 3247

wandb/debug-cli.log

@@ -1,6 +1,6 @@
-2022-05-03 17:09:46 INFO Running runs: []
-2022-05-03 17:09:46 INFO Agent received command: run
-2022-05-03 17:09:46 INFO Agent starting run with config:
+2022-05-04 13:11:52 INFO Running runs: []
+2022-05-04 13:11:53 INFO Agent received command: run
+2022-05-04 13:11:53 INFO Agent starting run with config:
 	eval_split_name: test
 	eval_steps: 500
 	evaluation_strategy: steps
@@ -8,103 +8,25 @@
 	generation_num_beams: 1
 	gradient_accumulation_steps: 8
 	greater_is_better: True
-	hidden_dropout: 0.036619638921206475
+	hidden_dropout: 0.18004101365999406
 	language: fr.en
-	learning_rate: 0.00024391819705381628
+	learning_rate: 0.0002757119755681108
 	logging_steps: 1
 	max_duration_in_seconds: 20
 	metric_for_best_model: bleu
 	model_name_or_path: ./
 	num_train_epochs: 3
-	output_dir: ./output_dir
-	per_device_eval_batch_size: 4
-	per_device_train_batch_size: 4
-	save_steps: 500
-	task: covost2
-	warmup_steps: 500
-2022-05-03 17:09:46 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.036619638921206475 --language=fr.en --learning_rate=0.00024391819705381628 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./output_dir --per_device_eval_batch_size=4 --per_device_train_batch_size=4 --save_steps=500 --task=covost2 --warmup_steps=500
-2022-05-03 17:09:51 INFO Running runs: ['vz5ppd75']
-2022-05-03 17:10:26 INFO Cleaning up finished run: vz5ppd75
-2022-05-03 17:10:28 INFO Agent received command: run
-2022-05-03 17:10:28 INFO Agent starting run with config:
-	eval_split_name: test
-	eval_steps: 500
-	evaluation_strategy: steps
-	generation_max_length: 40
-	generation_num_beams: 1
-	gradient_accumulation_steps: 8
-	greater_is_better: True
-	hidden_dropout: 0.1875094322808032
-	language: fr.en
-	learning_rate: 0.00024438201183496223
-	logging_steps: 1
-	max_duration_in_seconds: 20
-	metric_for_best_model: bleu
-	model_name_or_path: ./
-	num_train_epochs: 3
-	output_dir: ./output_dir
-	per_device_eval_batch_size: 4
-	per_device_train_batch_size: 4
-	save_steps: 500
-	task: covost2
-	warmup_steps: 500
-2022-05-03 17:10:36 INFO Running runs: []
-2022-05-03 17:10:36 INFO Agent received command: run
-2022-05-03 17:10:36 INFO Agent starting run with config:
-	eval_split_name: test
-	eval_steps: 500
-	evaluation_strategy: steps
-	generation_max_length: 40
-	generation_num_beams: 1
-	gradient_accumulation_steps: 8
-	greater_is_better: True
-	hidden_dropout: 0.055722391000930585
-	language: fr.en
-	learning_rate: 0.0006457481677728278
-	logging_steps: 1
-	max_duration_in_seconds: 20
-	metric_for_best_model: bleu
-	model_name_or_path: ./
-	num_train_epochs: 3
-	output_dir: ./output_dir
-	per_device_eval_batch_size: 4
-	per_device_train_batch_size: 4
-	save_steps: 500
-	task: covost2
-	warmup_steps: 500
-2022-05-03 17:10:36 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.055722391000930585 --language=fr.en --learning_rate=0.0006457481677728278 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./output_dir --per_device_eval_batch_size=4 --per_device_train_batch_size=4 --save_steps=500 --task=covost2 --warmup_steps=500
-2022-05-03 17:10:41 INFO Running runs: ['ldsojzle']
-2022-05-03 17:11:07 INFO Cleaning up finished run: ldsojzle
-2022-05-03 17:11:07 INFO Agent received command: run
-2022-05-03 17:11:07 INFO Agent starting run with config:
-	eval_split_name: test
-	eval_steps: 500
-	evaluation_strategy: steps
-	generation_max_length: 40
-	generation_num_beams: 1
-	gradient_accumulation_steps: 8
-	greater_is_better: True
-	hidden_dropout: 0.056807662149569525
-	language: fr.en
-	learning_rate: 0.0005558468401613797
-	logging_steps: 1
-	max_duration_in_seconds: 20
-	metric_for_best_model: bleu
-	model_name_or_path: ./
-	num_train_epochs: 3
-	output_dir: ./output_dir
-	per_device_eval_batch_size: 4
-	per_device_train_batch_size: 4
+	output_dir: ./
+	per_device_eval_batch_size: 8
+	per_device_train_batch_size: 8
 	save_steps: 500
 	task: covost2
 	warmup_steps: 500
-2022-05-03 17:11:07 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.056807662149569525 --language=fr.en --learning_rate=0.0005558468401613797 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./output_dir --per_device_eval_batch_size=4 --per_device_train_batch_size=4 --save_steps=500 --task=covost2 --warmup_steps=500
-2022-05-03 17:11:12 INFO Running runs: ['qv3vjr6j']
-2022-05-03 17:10:28 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.1875094322808032 --language=fr.en --learning_rate=0.00024438201183496223 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./output_dir --per_device_eval_batch_size=4 --per_device_train_batch_size=4 --save_steps=500 --task=covost2 --warmup_steps=500
-2022-05-03 17:11:29 INFO Running runs: ['irggvkgd']
-2022-05-03 17:11:37 INFO Cleaning up finished run: qv3vjr6j
-2022-05-03 17:11:37 INFO Agent received command: run
-2022-05-03 17:11:37 INFO Agent starting run with config:
+2022-05-04 13:11:53 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.18004101365999406 --language=fr.en --learning_rate=0.0002757119755681108 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./ --per_device_eval_batch_size=8 --per_device_train_batch_size=8 --save_steps=500 --task=covost2 --warmup_steps=500
+2022-05-04 13:11:58 INFO Running runs: ['qk3ze7ok']
+2022-05-04 13:12:13 INFO Running runs: []
+2022-05-04 13:12:13 INFO Agent received command: run
+2022-05-04 13:12:13 INFO Agent starting run with config:
 	eval_split_name: test
 	eval_steps: 500
 	evaluation_strategy: steps
@@ -112,24 +34,25 @@
 	generation_num_beams: 1
 	gradient_accumulation_steps: 8
 	greater_is_better: True
-	hidden_dropout: 0.03413483050532159
+	hidden_dropout: 0.04999238095195753
 	language: fr.en
-	learning_rate: 0.00022086866790135088
+	learning_rate: 0.0007702133913256148
 	logging_steps: 1
 	max_duration_in_seconds: 20
 	metric_for_best_model: bleu
 	model_name_or_path: ./
 	num_train_epochs: 3
-	output_dir: ./output_dir
-	per_device_eval_batch_size: 4
-	per_device_train_batch_size: 4
+	output_dir: ./
+	per_device_eval_batch_size: 8
+	per_device_train_batch_size: 8
 	save_steps: 500
 	task: covost2
 	warmup_steps: 500
-2022-05-03 17:11:37 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.03413483050532159 --language=fr.en --learning_rate=0.00022086866790135088 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./output_dir --per_device_eval_batch_size=4 --per_device_train_batch_size=4 --save_steps=500 --task=covost2 --warmup_steps=500
-2022-05-03 17:15:19 INFO Running runs: []
-2022-05-03 17:15:19 INFO Agent received command: run
-2022-05-03 17:15:19 INFO Agent starting run with config:
+2022-05-04 13:12:13 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.04999238095195753 --language=fr.en --learning_rate=0.0007702133913256148 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./ --per_device_eval_batch_size=8 --per_device_train_batch_size=8 --save_steps=500 --task=covost2 --warmup_steps=500
+2022-05-04 13:12:18 INFO Running runs: ['o7jpar4x']
+2022-05-04 13:30:33 INFO Running runs: []
+2022-05-04 13:30:33 INFO Agent received command: run
+2022-05-04 13:30:33 INFO Agent starting run with config:
 	eval_split_name: test
 	eval_steps: 500
 	evaluation_strategy: steps
@@ -137,21 +60,19 @@
 	generation_num_beams: 1
 	gradient_accumulation_steps: 8
 	greater_is_better: True
-	hidden_dropout: 0.06862889720223829
+	hidden_dropout: 0.035938233699532036
 	language: fr.en
-	learning_rate: 0.0004848089062550082
+	learning_rate: 0.0003284999261672522
 	logging_steps: 1
 	max_duration_in_seconds: 20
 	metric_for_best_model: bleu
 	model_name_or_path: ./
 	num_train_epochs: 3
 	output_dir: ./
-	per_device_eval_batch_size: 4
-	per_device_train_batch_size: 4
+	per_device_eval_batch_size: 8
+	per_device_train_batch_size: 8
 	save_steps: 500
 	task: covost2
 	warmup_steps: 500
-2022-05-03 17:15:19 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.06862889720223829 --language=fr.en --learning_rate=0.0004848089062550082 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./ --per_device_eval_batch_size=4 --per_device_train_batch_size=4 --save_steps=500 --task=covost2 --warmup_steps=500
-2022-05-03 17:15:24 INFO Running runs: ['a6039xud']
-2022-05-03 23:57:28 ERROR 500 response executing GraphQL.
-2022-05-03 23:57:28 ERROR {"error":"context deadline exceeded"}
+2022-05-04 13:30:33 INFO About to run command: python3 run_xtreme_s.py --overwrite_output_dir --freeze_feature_encoder --gradient_checkpointing --predict_with_generate --fp16 --group_by_length --do_train --do_eval --load_best_model_at_end --push_to_hub --use_auth_token --eval_split_name=test --eval_steps=500 --evaluation_strategy=steps --generation_max_length=40 --generation_num_beams=1 --gradient_accumulation_steps=8 --greater_is_better=True --hidden_dropout=0.035938233699532036 --language=fr.en --learning_rate=0.0003284999261672522 --logging_steps=1 --max_duration_in_seconds=20 --metric_for_best_model=bleu --model_name_or_path=./ --num_train_epochs=3 --output_dir=./ --per_device_eval_batch_size=8 --per_device_train_batch_size=8 --save_steps=500 --task=covost2 --warmup_steps=500
+2022-05-04 13:30:38 INFO Running runs: ['1tmxz74i']

wandb/debug-internal.log

@@ -1 +1 @@
-run-20220503_171959-a6039xud/logs/debug-internal.log
+run-20220504_142129-1tmxz74i/logs/debug-internal.log

wandb/debug.log

@@ -1 +1 @@
-run-20220503_171959-a6039xud/logs/debug.log
+run-20220504_142129-1tmxz74i/logs/debug.log

wandb/latest-run

@@ -1 +1 @@
-run-20220503_171959-a6039xud
+run-20220504_142129-1tmxz74i

wandb/{run-20220503_171959-a6039xud → run-20220504_142129-1tmxz74i}/files/config.yaml

@@ -8643,7 +8643,7 @@ _wandb:
       6:
       - 1
     python_version: 3.9.5
-    start_time: 1651598400
+    start_time: 1651674089
     t:
       1:
       - 1
@@ -8914,7 +8914,7 @@ encoder:
     forced_eos_token_id: null
     gradient_checkpointing: false
     hidden_act: gelu
-    hidden_dropout: 0.06862889720223829
+    hidden_dropout: 0.035938233699532036
     hidden_size: 1024
     id2label:
       '0': LABEL_0
@@ -9009,7 +9009,7 @@ eval_accumulation_steps:
   value: None
 eval_batch_size:
   desc: null
-  value: 4
+  value: 8
 eval_delay:
   desc: null
   value: 0
@@ -9069,7 +9069,7 @@ half_precision_backend:
   value: amp
 hidden_dropout:
   desc: null
-  value: 0.06862889720223829
+  value: 0.035938233699532036
 hub_model_id:
   desc: null
   value: None
@@ -9115,7 +9115,7 @@ language:
   value: fr.en
 learning_rate:
   desc: null
-  value: 0.0004848089062550082
+  value: 0.0003284999261672522
 length_column_name:
   desc: null
   value: length
@@ -9139,7 +9139,7 @@ log_on_each_node:
   value: true
 logging_dir:
   desc: null
-  value: ./runs/May03_17-15-22_sanchit--v100
+  value: ./runs/May04_13-30-37_sanchit--v100
 logging_first_step:
   desc: null
   value: false
@@ -9226,10 +9226,10 @@ past_index:
   value: -1
 per_device_eval_batch_size:
   desc: null
-  value: 4
+  value: 8
 per_device_train_batch_size:
   desc: null
-  value: 4
+  value: 8
 per_gpu_eval_batch_size:
   desc: null
   value: None
@@ -9358,7 +9358,7 @@ tpu_num_cores:
   value: None
 train_batch_size:
   desc: null
-  value: 4
+  value: 8
 transformers_version:
   desc: null
   value: null

wandb/{run-20220503_171959-a6039xud → run-20220504_142129-1tmxz74i}/files/wandb-metadata.json

@@ -1,8 +1,8 @@
 {
     "os": "Linux-5.11.0-1028-gcp-x86_64-with-glibc2.33",
     "python": "3.9.5",
-    "heartbeatAt": "2022-05-03T17:20:04.283739",
-    "startedAt": "2022-05-03T17:19:59.978222",
+    "heartbeatAt": "2022-05-04T14:21:35.228898",
+    "startedAt": "2022-05-04T14:21:29.131934",
     "docker": null,
     "gpu": "Tesla V100-SXM2-16GB",
     "gpu_count": 2,
@@ -27,17 +27,17 @@
         "--generation_num_beams=1",
         "--gradient_accumulation_steps=8",
         "--greater_is_better=True",
-        "--hidden_dropout=0.06862889720223829",
+        "--hidden_dropout=0.035938233699532036",
         "--language=fr.en",
-        "--learning_rate=0.0004848089062550082",
+        "--learning_rate=0.0003284999261672522",
         "--logging_steps=1",
         "--max_duration_in_seconds=20",
         "--metric_for_best_model=bleu",
         "--model_name_or_path=./",
         "--num_train_epochs=3",
         "--output_dir=./",
-        "--per_device_eval_batch_size=4",
-        "--per_device_train_batch_size=4",
+        "--per_device_eval_batch_size=8",
+        "--per_device_train_batch_size=8",
         "--save_steps=500",
         "--task=covost2",
         "--warmup_steps=500"
@@ -47,7 +47,7 @@
     "codePath": "run_xtreme_s.py",
     "git": {
         "remote": "https://huggingface.co/sanchit-gandhi/xtreme_s_xlsr_2_bart_covost2_fr_en",
-        "commit": "c787ad4df48cedb3c1492e0593442fbabf32819f"
+        "commit": "269164c4f37d0592db1babe9bf5fc77464fa8c97"
     },
     "email": "[email protected]",
     "root": "/home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en",

wandb/{run-20220503_171959-a6039xud → run-20220504_142129-1tmxz74i}/logs/debug.log

@@ -1,28 +1,28 @@
-2022-05-03 17:19:59,980 INFO    MainThread:42111 [wandb_setup.py:_flush():75] Loading settings from /home/sanchit_huggingface_co/.config/wandb/settings
-2022-05-03 17:19:59,980 INFO    MainThread:42111 [wandb_setup.py:_flush():75] Loading settings from wandb/settings
-2022-05-03 17:19:59,980 INFO    MainThread:42111 [wandb_setup.py:_flush():75] Loading settings from environment variables: {'entity': 'sanchit-gandhi', 'project': 'xtreme_s_xlsr_2_bart_covost2_fr_en', 'sweep_id': '39ci3gkf', 'root_dir': '/home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en', 'run_id': 'a6039xud', 'sweep_param_path': '/home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en/wandb/sweep-39ci3gkf/config-a6039xud.yaml'}
-2022-05-03 17:19:59,980 INFO    MainThread:42111 [wandb_setup.py:_flush():75] Inferring run settings from compute environment: {'program_relpath': 'run_xtreme_s.py', 'program': '/home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en/run_xtreme_s.py'}
-2022-05-03 17:19:59,980 INFO    MainThread:42111 [wandb_init.py:_log_setup():386] Logging user logs to /home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en/wandb/run-20220503_171959-a6039xud/logs/debug.log
-2022-05-03 17:19:59,980 INFO    MainThread:42111 [wandb_init.py:_log_setup():387] Logging internal logs to /home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en/wandb/run-20220503_171959-a6039xud/logs/debug-internal.log
-2022-05-03 17:19:59,980 INFO    MainThread:42111 [wandb_init.py:init():420] calling init triggers
-2022-05-03 17:19:59,980 INFO    MainThread:42111 [wandb_init.py:init():425] wandb.init called with sweep_config: {'eval_split_name': 'test', 'eval_steps': 500, 'evaluation_strategy': 'steps', 'generation_max_length': 40, 'generation_num_beams': 1, 'gradient_accumulation_steps': 8, 'greater_is_better': True, 'hidden_dropout': 0.06862889720223829, 'language': 'fr.en', 'learning_rate': 0.0004848089062550082, 'logging_steps': 1, 'max_duration_in_seconds': 20, 'metric_for_best_model': 'bleu', 'model_name_or_path': './', 'num_train_epochs': 3, 'output_dir': './', 'per_device_eval_batch_size': 4, 'per_device_train_batch_size': 4, 'save_steps': 500, 'task': 'covost2', 'warmup_steps': 500}
+2022-05-04 14:21:29,133 INFO    MainThread:50375 [wandb_setup.py:_flush():75] Loading settings from /home/sanchit_huggingface_co/.config/wandb/settings
+2022-05-04 14:21:29,133 INFO    MainThread:50375 [wandb_setup.py:_flush():75] Loading settings from wandb/settings
+2022-05-04 14:21:29,134 INFO    MainThread:50375 [wandb_setup.py:_flush():75] Loading settings from environment variables: {'entity': 'sanchit-gandhi', 'project': 'xtreme_s_xlsr_2_bart_covost2_fr_en', 'sweep_id': 'pvyx3mpp', 'root_dir': '/home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en', 'run_id': '1tmxz74i', 'sweep_param_path': '/home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en/wandb/sweep-pvyx3mpp/config-1tmxz74i.yaml'}
+2022-05-04 14:21:29,134 INFO    MainThread:50375 [wandb_setup.py:_flush():75] Inferring run settings from compute environment: {'program_relpath': 'run_xtreme_s.py', 'program': '/home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en/run_xtreme_s.py'}
+2022-05-04 14:21:29,134 INFO    MainThread:50375 [wandb_init.py:_log_setup():386] Logging user logs to /home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en/wandb/run-20220504_142129-1tmxz74i/logs/debug.log
+2022-05-04 14:21:29,134 INFO    MainThread:50375 [wandb_init.py:_log_setup():387] Logging internal logs to /home/sanchit_huggingface_co/xtreme_s_xlsr_2_bart_covost2_fr_en/wandb/run-20220504_142129-1tmxz74i/logs/debug-internal.log
+2022-05-04 14:21:29,134 INFO    MainThread:50375 [wandb_init.py:init():420] calling init triggers
+2022-05-04 14:21:29,134 INFO    MainThread:50375 [wandb_init.py:init():425] wandb.init called with sweep_config: {'eval_split_name': 'test', 'eval_steps': 500, 'evaluation_strategy': 'steps', 'generation_max_length': 40, 'generation_num_beams': 1, 'gradient_accumulation_steps': 8, 'greater_is_better': True, 'hidden_dropout': 0.035938233699532036, 'language': 'fr.en', 'learning_rate': 0.0003284999261672522, 'logging_steps': 1, 'max_duration_in_seconds': 20, 'metric_for_best_model': 'bleu', 'model_name_or_path': './', 'num_train_epochs': 3, 'output_dir': './', 'per_device_eval_batch_size': 8, 'per_device_train_batch_size': 8, 'save_steps': 500, 'task': 'covost2', 'warmup_steps': 500}
 config: {}
-2022-05-03 17:19:59,981 INFO    MainThread:42111 [wandb_init.py:init():471] starting backend
-2022-05-03 17:19:59,981 INFO    MainThread:42111 [backend.py:_multiprocessing_setup():99] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
-2022-05-03 17:20:00,059 INFO    MainThread:42111 [backend.py:ensure_launched():219] starting backend process...
-2022-05-03 17:20:00,134 INFO    MainThread:42111 [backend.py:ensure_launched():224] started backend process with pid: 42356
-2022-05-03 17:20:00,137 INFO    MainThread:42111 [wandb_init.py:init():480] backend started and connected
-2022-05-03 17:20:00,140 INFO    MainThread:42111 [wandb_run.py:_config_callback():966] config_cb None None {'eval_split_name': 'test', 'eval_steps': 500, 'evaluation_strategy': 'steps', 'generation_max_length': 40, 'generation_num_beams': 1, 'gradient_accumulation_steps': 8, 'greater_is_better': True, 'hidden_dropout': 0.06862889720223829, 'language': 'fr.en', 'learning_rate': 0.0004848089062550082, 'logging_steps': 1, 'max_duration_in_seconds': 20, 'metric_for_best_model': 'bleu', 'model_name_or_path': './', 'num_train_epochs': 3, 'output_dir': './', 'per_device_eval_batch_size': 4, 'per_device_train_batch_size': 4, 'save_steps': 500, 'task': 'covost2', 'warmup_steps': 500}
-2022-05-03 17:20:00,155 INFO    MainThread:42111 [wandb_init.py:init():550] updated telemetry
-2022-05-03 17:20:00,359 INFO    MainThread:42111 [wandb_init.py:init():581] communicating current version
-2022-05-03 17:20:01,247 INFO    MainThread:42111 [wandb_init.py:init():586] got version response upgrade_message: "wandb version 0.12.16 is available!  To upgrade, please run:\n $ pip install wandb --upgrade"
+2022-05-04 14:21:29,134 INFO    MainThread:50375 [wandb_init.py:init():471] starting backend
+2022-05-04 14:21:29,134 INFO    MainThread:50375 [backend.py:_multiprocessing_setup():99] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2022-05-04 14:21:29,204 INFO    MainThread:50375 [backend.py:ensure_launched():219] starting backend process...
+2022-05-04 14:21:29,272 INFO    MainThread:50375 [backend.py:ensure_launched():224] started backend process with pid: 50991
+2022-05-04 14:21:29,275 INFO    MainThread:50375 [wandb_init.py:init():480] backend started and connected
+2022-05-04 14:21:29,278 INFO    MainThread:50375 [wandb_run.py:_config_callback():966] config_cb None None {'eval_split_name': 'test', 'eval_steps': 500, 'evaluation_strategy': 'steps', 'generation_max_length': 40, 'generation_num_beams': 1, 'gradient_accumulation_steps': 8, 'greater_is_better': True, 'hidden_dropout': 0.035938233699532036, 'language': 'fr.en', 'learning_rate': 0.0003284999261672522, 'logging_steps': 1, 'max_duration_in_seconds': 20, 'metric_for_best_model': 'bleu', 'model_name_or_path': './', 'num_train_epochs': 3, 'output_dir': './', 'per_device_eval_batch_size': 8, 'per_device_train_batch_size': 8, 'save_steps': 500, 'task': 'covost2', 'warmup_steps': 500}
+2022-05-04 14:21:29,292 INFO    MainThread:50375 [wandb_init.py:init():550] updated telemetry
+2022-05-04 14:21:29,453 INFO    MainThread:50375 [wandb_init.py:init():581] communicating current version
+2022-05-04 14:21:30,582 INFO    MainThread:50375 [wandb_init.py:init():586] got version response upgrade_message: "wandb version 0.12.16 is available!  To upgrade, please run:\n $ pip install wandb --upgrade"
 
-2022-05-03 17:20:01,248 INFO    MainThread:42111 [wandb_init.py:init():596] communicating run to backend with 30 second timeout
-2022-05-03 17:20:01,757 INFO    MainThread:42111 [wandb_init.py:init():624] starting run threads in backend
-2022-05-03 17:20:04,342 INFO    MainThread:42111 [wandb_run.py:_console_start():1827] atexit reg
-2022-05-03 17:20:04,342 INFO    MainThread:42111 [wandb_run.py:_redirect():1701] redirect: SettingsConsole.REDIRECT
-2022-05-03 17:20:04,343 INFO    MainThread:42111 [wandb_run.py:_redirect():1706] Redirecting console.
-2022-05-03 17:20:04,345 INFO    MainThread:42111 [wandb_run.py:_redirect():1762] Redirects installed.
-2022-05-03 17:20:04,345 INFO    MainThread:42111 [wandb_init.py:init():651] run started, returning control to user process
-2022-05-03 17:20:04,348 INFO    MainThread:42111 [wandb_run.py:_config_callback():966] config_cb None None {'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': 'torch.float32', 'use_bfloat16': False, 'pruned_heads': {}, 'tie_word_embeddings': False, 'is_encoder_decoder': True, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'max_length': 40, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'chunk_size_feed_forward': 0, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'architectures': ['SpeechEncoderDecoderModel'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': None, 'pad_token_id': 1, 'eos_token_id': 2, 'sep_token_id': None, 'decoder_start_token_id': 0, 'task_specific_params': None, 'problem_type': None, '_name_or_path': './', 'transformers_version': None, 'decoder': {'vocab_size': 50265, 'max_position_embeddings': 1024, 'd_model': 1024, 'encoder_ffn_dim': 4096, 'encoder_layers': 12, 'encoder_attention_heads': 16, 'decoder_ffn_dim': 4096, 'decoder_layers': 12, 'decoder_attention_heads': 16, 'dropout': 0.1, 'attention_dropout': 0.1, 'activation_dropout': 0.1, 'activation_function': 'gelu', 'init_std': 0.02, 'encoder_layerdrop': 0.0, 'decoder_layerdrop': 0.0, 'classifier_dropout': 0.0, 'use_cache': True, 'num_hidden_layers': 12, 'scale_embedding': False, 'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': None, 'use_bfloat16': False, 'pruned_heads': {}, 'tie_word_embeddings': True, 'is_encoder_decoder': False, 'is_decoder': True, 'cross_attention_hidden_size': None, 'add_cross_attention': True, 'tie_encoder_decoder': False, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': True, 'num_beams': 4, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 3, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'chunk_size_feed_forward': 0, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': 0, 'forced_eos_token_id': 2, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'architectures': ['BartModel'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1', 2: 'LABEL_2'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1, 'LABEL_2': 2}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 0, 'pad_token_id': 1, 'eos_token_id': 2, 'sep_token_id': None, 'decoder_start_token_id': 2, 'task_specific_params': {'summarization': {'length_penalty': 1.0, 'max_length': 128, 'min_length': 12, 'num_beams': 4}, 'summarization_cnn': {'length_penalty': 2.0, 'max_length': 142, 'min_length': 56, 'num_beams': 4}, 'summarization_xsum': {'length_penalty': 1.0, 'max_length': 62, 'min_length': 11, 'num_beams': 6}}, 'problem_type': None, '_name_or_path': 'facebook/bart-large', 'transformers_version': '4.19.0.dev0', 'add_bias_logits': False, 'add_final_layer_norm': False, 'classif_dropout': 0.1, 'gradient_checkpointing': False, 'normalize_before': False, 'model_type': 'bart'}, 'encoder': {'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': 'float32', 'use_bfloat16': False, 'pruned_heads': {}, 'tie_word_embeddings': True, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'chunk_size_feed_forward': 0, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'architectures': ['Wav2Vec2ForPreTraining'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 1, 'pad_token_id': 0, 'eos_token_id': 2, 'sep_token_id': None, 'decoder_start_token_id': None, 'task_specific_params': None, 'problem_type': None, '_name_or_path': 'facebook/wav2vec2-xls-r-300m', 'transformers_version': '4.19.0.dev0', 'feat_extract_dropout': 0.0, 'gradient_checkpointing': False, 'num_feat_extract_layers': 7, 'hidden_size': 1024, 'feat_extract_norm': 'layer', 'feat_extract_activation': 'gelu', 'conv_dim': [512, 512, 512, 512, 512, 512, 512], 'conv_stride': [5, 2, 2, 2, 2, 2, 2], 'conv_kernel': [10, 3, 3, 3, 3, 2, 2], 'conv_bias': True, 'num_conv_pos_embeddings': 128, 'num_conv_pos_embedding_groups': 16, 'num_hidden_layers': 24, 'intermediate_size': 4096, 'hidden_act': 'gelu', 'num_attention_heads': 16, 'hidden_dropout': 0.06862889720223829, 'attention_dropout': 0.1, 'activation_dropout': 0.0, 'feat_proj_dropout': 0.0, 'final_dropout': 0.0, 'layerdrop': 0.0, 'layer_norm_eps': 1e-05, 'initializer_range': 0.02, 'vocab_size': 32, 'do_stable_layer_norm': True, 'use_weighted_layer_sum': False, 'apply_spec_augment': True, 'mask_time_prob': 0.1, 'mask_time_length': 10, 'mask_time_min_masks': 2, 'mask_feature_prob': 0.0, 'mask_feature_length': 10, 'mask_feature_min_masks': 0, 'num_codevectors_per_group': 320, 'num_codevector_groups': 2, 'contrastive_logits_temperature': 0.1, 'feat_quantizer_dropout': 0.0, 'num_negatives': 100, 'codevector_dim': 768, 'proj_codevector_dim': 768, 'diversity_loss_weight': 0.1, 'ctc_loss_reduction': 'sum', 'ctc_zero_infinity': False, 'add_adapter': True, 'adapter_kernel_size': 3, 'adapter_stride': 2, 'num_adapter_layers': 3, 'output_hidden_size': 1024, 'classifier_proj_size': 256, 'tdnn_dim': [512, 512, 512, 512, 1500], 'tdnn_kernel': [5, 3, 3, 1, 1], 'tdnn_dilation': [1, 2, 3, 1, 1], 'xvector_output_dim': 512, 'model_type': 'wav2vec2'}, 'model_type': 'speech-encoder-decoder', 'processor_class': 'Wav2Vec2Processor', 'use_cache': False, 'overwrite_output_dir': True, 'do_train': True, 'do_eval': True, 'do_predict': False, 'prediction_loss_only': False, 'per_gpu_train_batch_size': 'None', 'per_gpu_eval_batch_size': 'None', 'eval_accumulation_steps': 'None', 'eval_delay': 0, 'weight_decay': 0.0, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'max_steps': -1, 'lr_scheduler_type': 'linear', 'warmup_ratio': 0.0, 'log_level': -1, 'log_level_replica': -1, 'log_on_each_node': True, 'logging_dir': './runs/May03_17-15-22_sanchit--v100', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_total_limit': 'None', 'save_on_each_node': False, 'no_cuda': False, 'seed': 42, 'data_seed': 'None', 'bf16': False, 'fp16': True, 'fp16_opt_level': 'O1', 'half_precision_backend': 'amp', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': 'None', 'local_rank': -1, 'xpu_backend': 'None', 'tpu_num_cores': 'None', 'tpu_metrics_debug': False, 'debug': '[]', 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'past_index': -1, 'run_name': './', 'disable_tqdm': False, 'remove_unused_columns': True, 'label_names': 'None', 'load_best_model_at_end': True, 'ignore_data_skip': False, 'sharded_ddp': '[]', 'deepspeed': 'None', 'label_smoothing_factor': 0.0, 'optim': 'adamw_hf', 'adafactor': False, 'group_by_length': True, 'length_column_name': 'length', 'report_to': "['tensorboard', 'wandb', 'codecarbon']", 'ddp_find_unused_parameters': 'None', 'ddp_bucket_cap_mb': 'None', 'dataloader_pin_memory': True, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': True, 'resume_from_checkpoint': 'None', 'hub_model_id': 'None', 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': False, 'gradient_checkpointing': True, 'include_inputs_for_metrics': False, 'fp16_backend': 'auto', 'push_to_hub_model_id': 'None', 'push_to_hub_organization': 'None', 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', '_n_gpu': 1, 'mp_parameters': '', 'sortish_sampler': False, 'predict_with_generate': True, 'train_batch_size': 4, 'eval_batch_size': 4}
-2022-05-03 17:20:04,351 INFO    MainThread:42111 [wandb_watch.py:watch():43] Watching
+2022-05-04 14:21:30,583 INFO    MainThread:50375 [wandb_init.py:init():596] communicating run to backend with 30 second timeout
+2022-05-04 14:21:30,658 INFO    MainThread:50375 [wandb_init.py:init():624] starting run threads in backend
+2022-05-04 14:21:35,281 INFO    MainThread:50375 [wandb_run.py:_console_start():1827] atexit reg
+2022-05-04 14:21:35,281 INFO    MainThread:50375 [wandb_run.py:_redirect():1701] redirect: SettingsConsole.REDIRECT
+2022-05-04 14:21:35,282 INFO    MainThread:50375 [wandb_run.py:_redirect():1706] Redirecting console.
+2022-05-04 14:21:35,284 INFO    MainThread:50375 [wandb_run.py:_redirect():1762] Redirects installed.
+2022-05-04 14:21:35,284 INFO    MainThread:50375 [wandb_init.py:init():651] run started, returning control to user process
+2022-05-04 14:21:35,288 INFO    MainThread:50375 [wandb_run.py:_config_callback():966] config_cb None None {'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': 'torch.float32', 'use_bfloat16': False, 'pruned_heads': {}, 'tie_word_embeddings': False, 'is_encoder_decoder': True, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'max_length': 40, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'chunk_size_feed_forward': 0, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'architectures': ['SpeechEncoderDecoderModel'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': None, 'pad_token_id': 1, 'eos_token_id': 2, 'sep_token_id': None, 'decoder_start_token_id': 0, 'task_specific_params': None, 'problem_type': None, '_name_or_path': './', 'transformers_version': None, 'decoder': {'vocab_size': 50265, 'max_position_embeddings': 1024, 'd_model': 1024, 'encoder_ffn_dim': 4096, 'encoder_layers': 12, 'encoder_attention_heads': 16, 'decoder_ffn_dim': 4096, 'decoder_layers': 12, 'decoder_attention_heads': 16, 'dropout': 0.1, 'attention_dropout': 0.1, 'activation_dropout': 0.1, 'activation_function': 'gelu', 'init_std': 0.02, 'encoder_layerdrop': 0.0, 'decoder_layerdrop': 0.0, 'classifier_dropout': 0.0, 'use_cache': True, 'num_hidden_layers': 12, 'scale_embedding': False, 'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': None, 'use_bfloat16': False, 'pruned_heads': {}, 'tie_word_embeddings': True, 'is_encoder_decoder': False, 'is_decoder': True, 'cross_attention_hidden_size': None, 'add_cross_attention': True, 'tie_encoder_decoder': False, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': True, 'num_beams': 4, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 3, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'chunk_size_feed_forward': 0, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': 0, 'forced_eos_token_id': 2, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'architectures': ['BartModel'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1', 2: 'LABEL_2'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1, 'LABEL_2': 2}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 0, 'pad_token_id': 1, 'eos_token_id': 2, 'sep_token_id': None, 'decoder_start_token_id': 2, 'task_specific_params': {'summarization': {'length_penalty': 1.0, 'max_length': 128, 'min_length': 12, 'num_beams': 4}, 'summarization_cnn': {'length_penalty': 2.0, 'max_length': 142, 'min_length': 56, 'num_beams': 4}, 'summarization_xsum': {'length_penalty': 1.0, 'max_length': 62, 'min_length': 11, 'num_beams': 6}}, 'problem_type': None, '_name_or_path': 'facebook/bart-large', 'transformers_version': '4.19.0.dev0', 'add_bias_logits': False, 'add_final_layer_norm': False, 'classif_dropout': 0.1, 'gradient_checkpointing': False, 'normalize_before': False, 'model_type': 'bart'}, 'encoder': {'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': 'float32', 'use_bfloat16': False, 'pruned_heads': {}, 'tie_word_embeddings': True, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'chunk_size_feed_forward': 0, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'architectures': ['Wav2Vec2ForPreTraining'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 1, 'pad_token_id': 0, 'eos_token_id': 2, 'sep_token_id': None, 'decoder_start_token_id': None, 'task_specific_params': None, 'problem_type': None, '_name_or_path': 'facebook/wav2vec2-xls-r-300m', 'transformers_version': '4.19.0.dev0', 'feat_extract_dropout': 0.0, 'gradient_checkpointing': False, 'num_feat_extract_layers': 7, 'hidden_size': 1024, 'feat_extract_norm': 'layer', 'feat_extract_activation': 'gelu', 'conv_dim': [512, 512, 512, 512, 512, 512, 512], 'conv_stride': [5, 2, 2, 2, 2, 2, 2], 'conv_kernel': [10, 3, 3, 3, 3, 2, 2], 'conv_bias': True, 'num_conv_pos_embeddings': 128, 'num_conv_pos_embedding_groups': 16, 'num_hidden_layers': 24, 'intermediate_size': 4096, 'hidden_act': 'gelu', 'num_attention_heads': 16, 'hidden_dropout': 0.035938233699532036, 'attention_dropout': 0.1, 'activation_dropout': 0.0, 'feat_proj_dropout': 0.0, 'final_dropout': 0.0, 'layerdrop': 0.0, 'layer_norm_eps': 1e-05, 'initializer_range': 0.02, 'vocab_size': 32, 'do_stable_layer_norm': True, 'use_weighted_layer_sum': False, 'apply_spec_augment': True, 'mask_time_prob': 0.1, 'mask_time_length': 10, 'mask_time_min_masks': 2, 'mask_feature_prob': 0.0, 'mask_feature_length': 10, 'mask_feature_min_masks': 0, 'num_codevectors_per_group': 320, 'num_codevector_groups': 2, 'contrastive_logits_temperature': 0.1, 'feat_quantizer_dropout': 0.0, 'num_negatives': 100, 'codevector_dim': 768, 'proj_codevector_dim': 768, 'diversity_loss_weight': 0.1, 'ctc_loss_reduction': 'sum', 'ctc_zero_infinity': False, 'add_adapter': True, 'adapter_kernel_size': 3, 'adapter_stride': 2, 'num_adapter_layers': 3, 'output_hidden_size': 1024, 'classifier_proj_size': 256, 'tdnn_dim': [512, 512, 512, 512, 1500], 'tdnn_kernel': [5, 3, 3, 1, 1], 'tdnn_dilation': [1, 2, 3, 1, 1], 'xvector_output_dim': 512, 'model_type': 'wav2vec2'}, 'model_type': 'speech-encoder-decoder', 'processor_class': 'Wav2Vec2Processor', 'use_cache': False, 'overwrite_output_dir': True, 'do_train': True, 'do_eval': True, 'do_predict': False, 'prediction_loss_only': False, 'per_gpu_train_batch_size': 'None', 'per_gpu_eval_batch_size': 'None', 'eval_accumulation_steps': 'None', 'eval_delay': 0, 'weight_decay': 0.0, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'max_steps': -1, 'lr_scheduler_type': 'linear', 'warmup_ratio': 0.0, 'log_level': -1, 'log_level_replica': -1, 'log_on_each_node': True, 'logging_dir': './runs/May04_13-30-37_sanchit--v100', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_total_limit': 'None', 'save_on_each_node': False, 'no_cuda': False, 'seed': 42, 'data_seed': 'None', 'bf16': False, 'fp16': True, 'fp16_opt_level': 'O1', 'half_precision_backend': 'amp', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': 'None', 'local_rank': -1, 'xpu_backend': 'None', 'tpu_num_cores': 'None', 'tpu_metrics_debug': False, 'debug': '[]', 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'past_index': -1, 'run_name': './', 'disable_tqdm': False, 'remove_unused_columns': True, 'label_names': 'None', 'load_best_model_at_end': True, 'ignore_data_skip': False, 'sharded_ddp': '[]', 'deepspeed': 'None', 'label_smoothing_factor': 0.0, 'optim': 'adamw_hf', 'adafactor': False, 'group_by_length': True, 'length_column_name': 'length', 'report_to': "['tensorboard', 'wandb', 'codecarbon']", 'ddp_find_unused_parameters': 'None', 'ddp_bucket_cap_mb': 'None', 'dataloader_pin_memory': True, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': True, 'resume_from_checkpoint': 'None', 'hub_model_id': 'None', 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': False, 'gradient_checkpointing': True, 'include_inputs_for_metrics': False, 'fp16_backend': 'auto', 'push_to_hub_model_id': 'None', 'push_to_hub_organization': 'None', 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', '_n_gpu': 1, 'mp_parameters': '', 'sortish_sampler': False, 'predict_with_generate': True, 'train_batch_size': 8, 'eval_batch_size': 8}
+2022-05-04 14:21:35,292 INFO    MainThread:50375 [wandb_watch.py:watch():43] Watching

wandb/run-20220504_142129-1tmxz74i/run-1tmxz74i.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65c51ab08a6f7680732f10e9ba9643f597a3738005b0fe08b87027fd18bdfa5f
+size 52596447

wandb/{sweep-y3ak427l/config-irggvkgd.yaml → sweep-pvyx3mpp/config-1tmxz74i.yaml}

@@ -15,11 +15,11 @@ gradient_accumulation_steps:
 greater_is_better:
   value: true
 hidden_dropout:
-  value: 0.1875094322808032
+  value: 0.035938233699532036
 language:
   value: fr.en
 learning_rate:
-  value: 0.00024438201183496223
+  value: 0.0003284999261672522
 logging_steps:
   value: 1
 max_duration_in_seconds:
@@ -31,11 +31,11 @@ model_name_or_path:
 num_train_epochs:
   value: 3
 output_dir:
-  value: ./output_dir
+  value: ./
 per_device_eval_batch_size:
-  value: 4
+  value: 8
 per_device_train_batch_size:
-  value: 4
+  value: 8
 save_steps:
   value: 500
 task:

wandb/{sweep-39ci3gkf/config-a6039xud.yaml → sweep-pvyx3mpp/config-o7jpar4x.yaml}

@@ -15,11 +15,11 @@ gradient_accumulation_steps:
 greater_is_better:
   value: true
 hidden_dropout:
-  value: 0.06862889720223829
+  value: 0.04999238095195753
 language:
   value: fr.en
 learning_rate:
-  value: 0.0004848089062550082
+  value: 0.0007702133913256148
 logging_steps:
   value: 1
 max_duration_in_seconds:
@@ -33,9 +33,9 @@ num_train_epochs:
 output_dir:
   value: ./
 per_device_eval_batch_size:
-  value: 4
+  value: 8
 per_device_train_batch_size:
-  value: 4
+  value: 8
 save_steps:
   value: 500
 task:

wandb/{sweep-y3ak427l/config-ldsojzle.yaml → sweep-pvyx3mpp/config-qk3ze7ok.yaml}

@@ -15,11 +15,11 @@ gradient_accumulation_steps:
 greater_is_better:
   value: true
 hidden_dropout:
-  value: 0.055722391000930585
+  value: 0.18004101365999406
 language:
   value: fr.en
 learning_rate:
-  value: 0.0006457481677728278
+  value: 0.0002757119755681108
 logging_steps:
   value: 1
 max_duration_in_seconds:
@@ -31,11 +31,11 @@ model_name_or_path:
 num_train_epochs:
   value: 3
 output_dir:
-  value: ./output_dir
+  value: ./
 per_device_eval_batch_size:
-  value: 4
+  value: 8
 per_device_train_batch_size:
-  value: 4
+  value: 8
 save_steps:
   value: 500
 task:

wandb/sweep-y3ak427l/config-qv3vjr6j.yaml

@@ -1,44 +0,0 @@
-wandb_version: 1
-
-eval_split_name:
-  value: test
-eval_steps:
-  value: 500
-evaluation_strategy:
-  value: steps
-generation_max_length:
-  value: 40
-generation_num_beams:
-  value: 1
-gradient_accumulation_steps:
-  value: 8
-greater_is_better:
-  value: true
-hidden_dropout:
-  value: 0.056807662149569525
-language:
-  value: fr.en
-learning_rate:
-  value: 0.0005558468401613797
-logging_steps:
-  value: 1
-max_duration_in_seconds:
-  value: 20
-metric_for_best_model:
-  value: bleu
-model_name_or_path:
-  value: ./
-num_train_epochs:
-  value: 3
-output_dir:
-  value: ./output_dir
-per_device_eval_batch_size:
-  value: 4
-per_device_train_batch_size:
-  value: 4
-save_steps:
-  value: 500
-task:
-  value: covost2
-warmup_steps:
-  value: 500

wandb/sweep-y3ak427l/config-vz5ppd75.yaml

@@ -1,44 +0,0 @@
-wandb_version: 1
-
-eval_split_name:
-  value: test
-eval_steps:
-  value: 500
-evaluation_strategy:
-  value: steps
-generation_max_length:
-  value: 40
-generation_num_beams:
-  value: 1
-gradient_accumulation_steps:
-  value: 8
-greater_is_better:
-  value: true
-hidden_dropout:
-  value: 0.036619638921206475
-language:
-  value: fr.en
-learning_rate:
-  value: 0.00024391819705381628
-logging_steps:
-  value: 1
-max_duration_in_seconds:
-  value: 20
-metric_for_best_model:
-  value: bleu
-model_name_or_path:
-  value: ./
-num_train_epochs:
-  value: 3
-output_dir:
-  value: ./output_dir
-per_device_eval_batch_size:
-  value: 4
-per_device_train_batch_size:
-  value: 4
-save_steps:
-  value: 500
-task:
-  value: covost2
-warmup_steps:
-  value: 500

wandb/sweep-y3ak427l/config-xur584bd.yaml

@@ -1,44 +0,0 @@
-wandb_version: 1
-
-eval_split_name:
-  value: test
-eval_steps:
-  value: 500
-evaluation_strategy:
-  value: steps
-generation_max_length:
-  value: 40
-generation_num_beams:
-  value: 1
-gradient_accumulation_steps:
-  value: 8
-greater_is_better:
-  value: true
-hidden_dropout:
-  value: 0.03413483050532159
-language:
-  value: fr.en
-learning_rate:
-  value: 0.00022086866790135088
-logging_steps:
-  value: 1
-max_duration_in_seconds:
-  value: 20
-metric_for_best_model:
-  value: bleu
-model_name_or_path:
-  value: ./
-num_train_epochs:
-  value: 3
-output_dir:
-  value: ./output_dir
-per_device_eval_batch_size:
-  value: 4
-per_device_train_batch_size:
-  value: 4
-save_steps:
-  value: 500
-task:
-  value: covost2
-warmup_steps:
-  value: 500