Upload 12 files

discriminative/eu/fine-tuning/predict/model-transfer.sh

@@ -0,0 +1,44 @@
+#!/bin/bash
+#SBATCH --qos=regular
+#SBATCH --job-name=var-nli-pred
+#SBATCH --cpus-per-task=1
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:4
+#SBATCH --constraint=a100-sxm4
+#SBATCH --time=1-00:00:00
+#SBATCH --output=var-nli-pre.log
+#SBATCH --error=var-nli-pre.err
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+
+source /scratch/jbengoetxea/phd/.phd_venv_new/bin/activate 
+
+for seed in 23 27 33
+# 27 33
+do
+    for model in jhu-clsp/mmBERT-base
+    # jhu-clsp/mmBERT-base FacebookAI/xlm-roberta-large answerdotai/ModernBERT-large
+    do
+        for dataset in nafar_nat
+        # eu native var
+        do
+            python /scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/fine-tuning/run_xnli_eus.py \
+            --model_name_or_path "/scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/models/model-transfer/$model/$seed" \
+            --language eu \
+            --train_language en \
+            --test_data $dataset \
+            --do_predict \
+            --per_device_train_batch_size 32 \
+            --num_train_epochs 10.0 \
+            --max_seq_length 128 \
+            --output_dir /scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/results/model-transfer/$model/no-rep/$dataset/$seed \
+            --save_steps 50000 \
+            --metric_for_best_model accuracy \
+            --seed $seed \
+            --eval_steps 5000 \
+            --save_total_limit 2 
+        done
+    done
+done

discriminative/eu/fine-tuning/predict/translate-train.sh

@@ -0,0 +1,47 @@
+#!/bin/bash
+#SBATCH --partition=hitz-exclusive
+#SBATCH --account=hitz-exclusive
+#SBATCH --job-name=var-nli-pred
+#SBATCH --cpus-per-task=1
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:2
+#SBATCH --constraint=a100-sxm4
+#SBATCH --time=02:00:00
+#SBATCH --output=var-nli-pre2.log
+#SBATCH --error=var-nli-pre2.err
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+
+
+source /scratch/jbengoetxea/phd/.phd_venv_new/bin/activate
+
+for seed in 27 23 33
+do
+    for model in ixa-ehu/roberta-eus-euscrawl-large-cased FacebookAI/xlm-roberta-large 
+    # 
+    # microsoft/mdeberta-v3-base FacebookAI/xlm-roberta-large ixa-ehu/roberta-eus-euscrawl-large-cased ixa-ehu/berteus-base-cased
+    do    
+        for dataset in xnli_expanded
+        # eu native var
+        do
+            python /scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/fine-tuning/run_xnli_eus.py \
+            --model_name_or_path "/scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/models/translate-train/$model/$seed" \
+            --language eu \
+            --train_language eu \
+            --test_data $dataset \
+            --do_predict \
+            --per_device_train_batch_size 32 \
+            --num_train_epochs 10.0 \
+            --max_seq_length 128 \
+            --output_dir /scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/results/translate-train/$model/$dataset/$seed \
+            --save_steps 50000 \
+            --metric_for_best_model accuracy \
+            --seed $seed \
+            --eval_steps 5000 \
+            --save_total_limit 2 
+
+        done
+    done
+done

discriminative/eu/fine-tuning/requirements.txt

@@ -0,0 +1,8 @@
+accelerate >= 0.12.0
+datasets >= 1.8.0
+sentencepiece != 0.1.92
+scipy
+scikit-learn
+protobuf
+torch >= 1.3
+evaluate

discriminative/eu/fine-tuning/run_xnli_eus.py

@@ -0,0 +1,688 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
+# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+""" Finetuning multi-lingual models on XNLI (e.g. Bert, DistilBERT, XLM), evaluation on XNLI-eu(Basque).
+    Adapted from `transformers/examples/pytorch/text-classification/run_xnli.py`"""
+
+import logging
+import os
+import random
+import sys
+import warnings
+from dataclasses import dataclass, field
+from typing import Optional
+
+import datasets
+import evaluate
+import numpy as np
+from datasets import load_dataset
+
+import transformers
+from transformers import (
+    AutoConfig,
+    AutoModelForSequenceClassification,
+    AutoTokenizer,
+    DataCollatorWithPadding,
+    EvalPrediction,
+    HfArgumentParser,
+    Trainer,
+    TrainingArguments,
+    default_data_collator,
+    set_seed,
+)
+from transformers.trainer_utils import get_last_checkpoint
+from transformers.utils import check_min_version
+from huggingface_hub.utils import send_telemetry
+from transformers.utils.versions import require_version
+import argparse
+
+import torch
+
+# import wandb
+# wandb.login(key="8495a960a8aceb5bd2f765006a1bd883733f7366")
+# wandb.init(project="xnli-eu-zero-shot-finetuning")
+
+
+parser = argparse.ArgumentParser()
+parser.add_argument("--ev_dataset", type=str, default="eu", help="Choose a development dataset from the two partitions available on the HF of XNLIeu (eu, eu_mt)")
+parser.add_argument("--pred_dataset", type=str, default="eu", help="Choose a test dataset from the three partitions available on the HF of XNLIeu (eu, eu_mt, eu_native)")
+args, remaining_args = parser.parse_known_args()
+
+ev_dataset = args.ev_dataset
+pred_dataset = args.pred_dataset
+
+
+# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
+check_min_version("4.34.0.dev0")
+
+require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")
+
+logger = logging.getLogger(__name__)
+
+@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.
+    """
+
+    max_seq_length: Optional[int] = field(
+        default=128,
+        metadata={
+            "help": (
+                "The maximum total input sequence length after tokenization. Sequences longer "
+                "than this will be truncated, sequences shorter will be padded."
+            )
+        },
+    )
+    overwrite_cache: bool = field(
+        default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
+    )
+    pad_to_max_length: bool = field(
+        default=True,
+        metadata={
+            "help": (
+                "Whether to pad all samples to `max_seq_length`. "
+                "If False, will pad the samples dynamically when batching to the maximum length in the batch."
+            )
+        },
+    )
+    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 evaluation 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."
+            )
+        },
+    )
+
+
+@dataclass
+class ModelArguments:
+    """
+    Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
+    """
+
+    model_name_or_path: str = field(
+        default=None, metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"}
+    )
+    language: str = field(
+        default=None, metadata={"help": "Evaluation language. Also train language if `train_language` is set to None."}
+    )
+    train_language: Optional[str] = field(
+        default=None, metadata={"help": "Train language if it is different from the evaluation language."}
+    )
+    test_data: str = field(
+        default=None, metadata={"help": "test data type to be used in evaluation"}
+
+    )
+    config_name: Optional[str] = field(
+        default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"}
+    )
+    tokenizer_name: Optional[str] = field(
+        default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"}
+    )
+    cache_dir: Optional[str] = field(
+        default=None,
+        metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
+    )
+    do_lower_case: Optional[bool] = field(
+        default=False,
+        metadata={"help": "arg to indicate if tokenizer should do lower case in AutoTokenizer.from_pretrained()"},
+    )
+    use_fast_tokenizer: bool = field(
+        default=True,
+        metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
+    )
+    model_revision: str = field(
+        default="main",
+        metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."},
+    )
+    token: str = field(
+        default=None,
+        metadata={
+            "help": (
+                "The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
+                "generated when running `huggingface-cli login` (stored in `~/.huggingface`)."
+            )
+        },
+    )
+    use_auth_token: bool = field(
+        default=None,
+        metadata={
+            "help": "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token`."
+        },
+    )
+    trust_remote_code: bool = field(
+        default=False,
+        metadata={
+            "help": (
+                "Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
+                "should only be set to `True` for repositories you trust and in which you have read the code, as it will"
+                "execute code present on the Hub on your local machine."
+            )
+        },
+    )
+    ignore_mismatched_sizes: bool = field(
+        default=False,
+        metadata={"help": "Will enable to load a pretrained model whose head dimensions are different."},
+    )
+
+
+def main():
+    # See all possible arguments in src/transformers/training_args.py
+    # or by passing the --help flag to this script.
+    # We now keep distinct sets of args, for a cleaner separation of concerns.
+
+    parser = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments))
+    model_args, data_args, training_args = parser.parse_args_into_dataclasses(remaining_args)
+
+    #### Fix
+    # model_args.cache_dir = "/gaueko1/users/jbengoetxea004/borra_nazazu"
+    # os.environ['HF_HOME'] = '/gaueko1/users/jbengoetxea004/borra_nazazu'
+
+    if model_args.use_auth_token is not None:
+        warnings.warn("The `use_auth_token` argument is deprecated and will be removed in v4.34.", FutureWarning)
+        if model_args.token is not None:
+            raise ValueError("`token` and `use_auth_token` are both specified. Please set only the argument `token`.")
+        model_args.token = model_args.use_auth_token
+
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    # send_telemetry("run_xnli", model_args)
+
+
+    # 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)],
+    )
+
+    if training_args.should_log:
+        # The default of training_args.log_level is passive, so we set log level at info here to have that default.
+        transformers.utils.logging.set_verbosity_info()
+
+    log_level = training_args.get_process_log_level()
+    logger.setLevel(log_level)
+    datasets.utils.logging.set_verbosity(log_level)
+    transformers.utils.logging.set_verbosity(log_level)
+    transformers.utils.logging.enable_default_handler()
+    transformers.utils.logging.enable_explicit_format()
+
+    # 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: {training_args.parallel_mode.value == 'distributed'}, 16-bits training: {training_args.fp16}"
+    )
+    logger.info(f"Training/evaluation parameters {training_args}")
+
+    # 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."
+            )
+
+    # Set seed before initializing model.
+    set_seed(training_args.seed)
+
+    # In distributed training, the load_dataset function guarantees that only one local process can concurrently
+    # download the dataset.
+    # Downloading and loading xnli dataset from the hub.
+    if training_args.do_train:
+        if model_args.train_language is None:
+            train_dataset = load_dataset(
+                "xnli",
+                model_args.language,
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+        elif model_args.train_language == "eu":
+            train_dataset = load_dataset(
+                "HiTZ/xnli-eu",
+                model_args.train_language,
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.train_language == "eu_all":
+            train_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli_expanded_train_correct.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        else:
+            train_dataset = load_dataset(
+                "xnli",
+                model_args.train_language,
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+
+    if training_args.do_eval:
+        if model_args.train_language == "eu_all": 
+            eval_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli_expanded_eval_correct.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        else:
+            eval_dataset = load_dataset(
+                "HiTZ/xnli-eu",
+                ev_dataset,
+                split="validation",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+    
+    
+    if training_args.do_predict:
+        if model_args.test_data == "eu": 
+            predict_dataset = load_dataset(
+                "HiTZ/xnli-eu",
+                pred_dataset,
+                split="test",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "native":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-eu-native.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "var":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-eu-var.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "trans_test":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/tartalo01/users/jbengoetxea004/phd/xnli-paraphrasing/data/en/eu-native-in-english-with-labels.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "trans_test_var":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-native-var-eu-NO-REPETITION.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+
+
+        elif model_args.test_data == "biz":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-test-gipuzkera-done.tsv",
+                delimiter="\t",  
+                split="train",
+                encoding="latin-1",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "gip":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-test-gipuzkera-done.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "nafar":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-test-nafar-lapurtera-done.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "biz_nat":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-native-bizkaieraz-done.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "gip_nat":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-native-gipuzkera-done.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "nafar_nat":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-native-nafar-lapurtera-done.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+
+        elif model_args.test_data == "xnli_expanded":
+            predict_dataset = load_dataset(
+                'csv',
+                data_files="/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli_expanded_test_correct.tsv",
+                delimiter="\t",  
+                split="train",
+                cache_dir=model_args.cache_dir,
+                token=model_args.token,
+            )
+    
+
+    label_list = [0,1,2]
+
+    num_labels = len(label_list)
+
+    # Load pretrained model and tokenizer
+    # In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
+    # download model & vocab.
+    config = AutoConfig.from_pretrained(
+        model_args.config_name if model_args.config_name else model_args.model_name_or_path,
+        num_labels=num_labels,
+        id2label={str(i): label for i, label in enumerate(label_list)},
+        label2id={label: i for i, label in enumerate(label_list)},
+        finetuning_task="xnli",
+        cache_dir=model_args.cache_dir,
+        revision=model_args.model_revision,
+        token=model_args.token,
+        trust_remote_code=model_args.trust_remote_code,
+    )
+    tokenizer = AutoTokenizer.from_pretrained(
+        model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path,
+        do_lower_case=model_args.do_lower_case,
+        cache_dir=model_args.cache_dir,
+        use_fast=False,
+        revision=model_args.model_revision,
+        token=model_args.token,
+        trust_remote_code=model_args.trust_remote_code,
+    )
+    model = AutoModelForSequenceClassification.from_pretrained(
+        model_args.model_name_or_path,
+        from_tf=bool(".ckpt" in model_args.model_name_or_path),
+        config=config,
+        cache_dir=model_args.cache_dir,
+        revision=model_args.model_revision,
+        token=model_args.token,
+        trust_remote_code=model_args.trust_remote_code,
+        ignore_mismatched_sizes=model_args.ignore_mismatched_sizes,
+    )
+
+    # Preprocessing the datasets
+    # Padding strategy
+    if data_args.pad_to_max_length:
+        padding = "max_length"
+    else:
+        # We will pad later, dynamically at batch creation, to the max sequence length in each batch
+        padding = False
+
+    def preprocess_function(examples):
+        # Tokenize the texts
+        return tokenizer(
+            examples["premise"],
+            examples["hypothesis"],
+            padding=padding,
+            max_length=data_args.max_seq_length,
+            truncation=True,
+        )
+
+    if training_args.do_train:
+        if data_args.max_train_samples is not None:
+            max_train_samples = min(len(train_dataset), data_args.max_train_samples)
+            train_dataset = train_dataset.select(range(max_train_samples))
+        with training_args.main_process_first(desc="train dataset map pre-processing"):
+            train_dataset = train_dataset.map(
+                preprocess_function,
+                batched=True,
+                load_from_cache_file=not data_args.overwrite_cache,
+                desc="Running tokenizer on train dataset",
+            )
+        # Log a few random samples from the training set:
+        for index in random.sample(range(len(train_dataset)), 3):
+            logger.info(f"Sample {index} of the training set: {train_dataset[index]}.")
+
+    if training_args.do_eval:
+        if data_args.max_eval_samples is not None:
+            max_eval_samples = min(len(eval_dataset), data_args.max_eval_samples)
+            eval_dataset = eval_dataset.select(range(max_eval_samples))
+        with training_args.main_process_first(desc="validation dataset map pre-processing"):
+            eval_dataset = eval_dataset.map(
+                preprocess_function,
+                batched=True,
+                load_from_cache_file=not data_args.overwrite_cache,
+                desc="Running tokenizer on validation dataset",
+            )
+
+    if training_args.do_predict:
+        if data_args.max_predict_samples is not None:
+            max_predict_samples = min(len(predict_dataset), data_args.max_predict_samples)
+            predict_dataset = predict_dataset.select(range(max_predict_samples))
+        with training_args.main_process_first(desc="prediction dataset map pre-processing"):
+            predict_dataset = predict_dataset.map(
+                preprocess_function,
+                batched=True,
+                load_from_cache_file=not data_args.overwrite_cache,
+                desc="Running tokenizer on prediction dataset",
+            )
+    # print(predict_dataset["premise"])
+    # Get the metric function
+    metric = evaluate.load("xnli")
+
+    # You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
+    # predictions and label_ids field) and has to return a dictionary string to float.
+    def compute_metrics(p: EvalPrediction):
+        preds = p.predictions[0] if isinstance(p.predictions, tuple) else p.predictions
+        preds = np.argmax(preds, axis=1)
+        return metric.compute(predictions=preds, references=p.label_ids)
+
+    # Data collator will default to DataCollatorWithPadding, so we change it if we already did the padding.
+    if data_args.pad_to_max_length:
+        data_collator = default_data_collator
+    elif training_args.fp16:
+        data_collator = DataCollatorWithPadding(tokenizer, pad_to_multiple_of=8)
+    else:
+        data_collator = None
+
+    # Initialize our Trainer
+    trainer = Trainer(
+        model=model,
+        args=training_args,
+        train_dataset=train_dataset if training_args.do_train else None,
+        eval_dataset=eval_dataset if training_args.do_eval else None,
+        compute_metrics=compute_metrics,
+        tokenizer=tokenizer,
+        data_collator=data_collator,
+    )
+
+    # Training
+    if training_args.do_train:
+        checkpoint = None
+        if training_args.resume_from_checkpoint is not None:
+            checkpoint = training_args.resume_from_checkpoint
+        elif last_checkpoint is not None:
+            checkpoint = last_checkpoint
+        train_result = trainer.train(resume_from_checkpoint=checkpoint)
+        metrics = train_result.metrics
+        max_train_samples = (
+            data_args.max_train_samples if data_args.max_train_samples is not None else len(train_dataset)
+        )
+        metrics["train_samples"] = min(max_train_samples, len(train_dataset))
+
+        trainer.save_model()  # Saves the tokenizer too for easy upload
+
+        tokenizer.save_pretrained(training_args.output_dir)
+
+        trainer.log_metrics("train", metrics)
+        trainer.save_metrics("train", metrics)
+        trainer.save_state()
+
+    # Evaluation
+    if training_args.do_eval:
+        logger.info("*** Evaluate ***")
+        metrics = trainer.evaluate(eval_dataset=eval_dataset)
+
+        max_eval_samples = data_args.max_eval_samples if data_args.max_eval_samples is not None else len(eval_dataset)
+        metrics["eval_samples"] = min(max_eval_samples, len(eval_dataset))
+
+        trainer.log_metrics("eval", metrics)
+        trainer.save_metrics("eval", metrics)
+
+    # Prediction
+    if training_args.do_predict:
+        logger.info("*** Predict ***")
+        predictions, labels, metrics = trainer.predict(predict_dataset, metric_key_prefix="predict")
+
+        max_predict_samples = (
+            data_args.max_predict_samples if data_args.max_predict_samples is not None else len(predict_dataset)
+        )
+        metrics["predict_samples"] = min(max_predict_samples, len(predict_dataset))
+
+        trainer.log_metrics("predict", metrics)
+        trainer.save_metrics("predict", metrics)
+
+        predictions = np.argmax(predictions, axis=1)
+        output_predict_file = os.path.join(training_args.output_dir, "predictions.txt")
+        if trainer.is_world_process_zero():
+            with open(output_predict_file, "w") as writer:
+                writer.write("index\tprediction\n")
+                for index, item in enumerate(predictions):
+                    item = label_list[item]
+                    writer.write(f"{index}\t{item}\n")
+
+
+        # Save predictions with premise and hypotheses
+        no_variation = ["native", "es", "trans_test"]
+        variation = ["var", "trans_test_var"]
+
+        if model_args.test_data in no_variation: 
+            output_predict_file = os.path.join(training_args.output_dir, "predictions_with_text.txt")
+            
+            if trainer.is_world_process_zero():
+                with open(output_predict_file, "w") as writer:
+                    writer.write("index\tpremise\thypothesis\tlabel\tprediction\n")  # Write header
+                    for index, (premise, hypothesis, label, prediction_idx) in enumerate(zip(
+                        predict_dataset["premise"], predict_dataset["hypothesis"], predict_dataset["label"], predictions
+                    )):
+                        predicted_label = label_list[prediction_idx]
+                        writer.write(f"{index}\t{premise}\t{hypothesis}\t{label}\t{predicted_label}\n")
+
+        elif model_args.test_data in variation: 
+            output_predict_file = os.path.join(training_args.output_dir, "predictions_with_text.txt")
+            
+            if trainer.is_world_process_zero():
+                with open(output_predict_file, "w") as writer:
+                    writer.write("index\tann_id\tchange_type\tprobintzia\tprem_id\tpremise\thypothesis\tlabel\tprediction\n")  # Write header
+                    for index, (ann_id, change_type, probintzia, prem_id, premise, hypothesis, label, prediction_idx) in enumerate(zip(
+                        predict_dataset["ann_id"], predict_dataset["change_type"], predict_dataset["probintzia"], predict_dataset["prem_id"], predict_dataset["premise"], predict_dataset["hypothesis"], predict_dataset["label"], predictions
+                    )):
+                        predicted_label = label_list[prediction_idx]
+                        writer.write(f"{index}\t{ann_id}\t{change_type}\t{probintzia}\t{prem_id}\t{premise}\t{hypothesis}\t{label}\t{predicted_label}\n")
+
+        elif model_args.test_data == "xnli_expanded": 
+            output_predict_file = os.path.join(training_args.output_dir, "predictions_with_text.txt")
+
+            if trainer.is_world_process_zero():
+                with open(output_predict_file, "w") as writer:
+                    writer.write("index\tdata_source\tdialect\tprem_id\tpremise\thypothesis_id\thypothesis\tlabel\tprediction\n")  # Write header
+                    for index, (data_source, dialect, prem_id, premise, hypothesis_id, hypothesis, label, prediction_idx) in enumerate(zip(
+                        predict_dataset["data_source"], predict_dataset["dialect"], predict_dataset["prem_id"], predict_dataset["premise"], predict_dataset["hypothesis_id"], predict_dataset["hypothesis"], predict_dataset["label"], predictions
+                    )):
+                        predicted_label = label_list[prediction_idx]
+                        writer.write(f"{index}\t{data_source}\t{dialect}\t{prem_id}\t{premise}\t{hypothesis_id}\t{hypothesis}\t{label}\t{predicted_label}\n")
+
+
+    try: 
+        torch.distributed.destroy_process_group()
+    except:
+        pass
+    
+
+if __name__ == "__main__":
+    main()

discriminative/eu/fine-tuning/train/logs/model_transder-2gpu.err

@@ -0,0 +1,141 @@
+The following values were not passed to `accelerate launch` and had defaults used instead:
+	`--num_processes` was set to a value of `2`
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--num_machines` was set to a value of `1`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+/scratch/jbengoetxea/phd/.phd_venv/bin/python3: can't open file '/scratch/jbengoetxea/phd/XNLIvar/scripts/fine-tuning/run_xnli_eus.py': [Errno 2] No such file or directory
+/scratch/jbengoetxea/phd/.phd_venv/bin/python3: can't open file '/scratch/jbengoetxea/phd/XNLIvar/scripts/fine-tuning/run_xnli_eus.py': [Errno 2] No such file or directory
+E1204 11:46:18.152000 1475833 torch/distributed/elastic/multiprocessing/api.py:869] failed (exitcode: 2) local_rank: 0 (pid: 1476026) of binary: /scratch/jbengoetxea/phd/.phd_venv/bin/python3
+Traceback (most recent call last):
+  File "/scratch/jbengoetxea/phd/.phd_venv/bin/accelerate", line 8, in <module>
+    sys.exit(main())
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/accelerate_cli.py", line 50, in main
+    args.func(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 1204, in launch_command
+    multi_gpu_launcher(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 825, in multi_gpu_launcher
+    distrib_run.run(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/run.py", line 910, in run
+    elastic_launch(
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 138, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 269, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+fine-tuning/run_xnli_eus.py FAILED
+------------------------------------------------------------
+Failures:
+[1]:
+  time      : 2025-12-04_11:46:18
+  host      : hyperion-252.sw.ehu.es
+  rank      : 1 (local_rank: 1)
+  exitcode  : 2 (pid: 1476027)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-12-04_11:46:18
+  host      : hyperion-252.sw.ehu.es
+  rank      : 0 (local_rank: 0)
+  exitcode  : 2 (pid: 1476026)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
+srun: error: hyperion-252: task 0: Exited with exit code 1
+The following values were not passed to `accelerate launch` and had defaults used instead:
+	`--num_processes` was set to a value of `2`
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--num_machines` was set to a value of `1`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+/scratch/jbengoetxea/phd/.phd_venv/bin/python3: can't open file '/scratch/jbengoetxea/phd/XNLIvar/scripts/fine-tuning/run_xnli_eus.py': [Errno 2] No such file or directory
+/scratch/jbengoetxea/phd/.phd_venv/bin/python3: can't open file '/scratch/jbengoetxea/phd/XNLIvar/scripts/fine-tuning/run_xnli_eus.py': [Errno 2] No such file or directory
+E1204 11:46:27.775000 1476160 torch/distributed/elastic/multiprocessing/api.py:869] failed (exitcode: 2) local_rank: 0 (pid: 1476196) of binary: /scratch/jbengoetxea/phd/.phd_venv/bin/python3
+Traceback (most recent call last):
+  File "/scratch/jbengoetxea/phd/.phd_venv/bin/accelerate", line 8, in <module>
+    sys.exit(main())
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/accelerate_cli.py", line 50, in main
+    args.func(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 1204, in launch_command
+    multi_gpu_launcher(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 825, in multi_gpu_launcher
+    distrib_run.run(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/run.py", line 910, in run
+    elastic_launch(
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 138, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 269, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+fine-tuning/run_xnli_eus.py FAILED
+------------------------------------------------------------
+Failures:
+[1]:
+  time      : 2025-12-04_11:46:27
+  host      : hyperion-252.sw.ehu.es
+  rank      : 1 (local_rank: 1)
+  exitcode  : 2 (pid: 1476197)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-12-04_11:46:27
+  host      : hyperion-252.sw.ehu.es
+  rank      : 0 (local_rank: 0)
+  exitcode  : 2 (pid: 1476196)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
+srun: error: hyperion-252: task 0: Exited with exit code 1
+The following values were not passed to `accelerate launch` and had defaults used instead:
+	`--num_processes` was set to a value of `2`
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--num_machines` was set to a value of `1`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+/scratch/jbengoetxea/phd/.phd_venv/bin/python3: can't open file '/scratch/jbengoetxea/phd/XNLIvar/scripts/fine-tuning/run_xnli_eus.py': [Errno 2] No such file or directory
+/scratch/jbengoetxea/phd/.phd_venv/bin/python3: can't open file '/scratch/jbengoetxea/phd/XNLIvar/scripts/fine-tuning/run_xnli_eus.py': [Errno 2] No such file or directory
+E1204 11:46:37.018000 1476330 torch/distributed/elastic/multiprocessing/api.py:869] failed (exitcode: 2) local_rank: 0 (pid: 1476362) of binary: /scratch/jbengoetxea/phd/.phd_venv/bin/python3
+Traceback (most recent call last):
+  File "/scratch/jbengoetxea/phd/.phd_venv/bin/accelerate", line 8, in <module>
+    sys.exit(main())
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/accelerate_cli.py", line 50, in main
+    args.func(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 1204, in launch_command
+    multi_gpu_launcher(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 825, in multi_gpu_launcher
+    distrib_run.run(args)
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/run.py", line 910, in run
+    elastic_launch(
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 138, in __call__
+    return launch_agent(self._config, self._entrypoint, list(args))
+  File "/scratch/jbengoetxea/phd/.phd_venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 269, in launch_agent
+    raise ChildFailedError(
+torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
+============================================================
+fine-tuning/run_xnli_eus.py FAILED
+------------------------------------------------------------
+Failures:
+[1]:
+  time      : 2025-12-04_11:46:37
+  host      : hyperion-252.sw.ehu.es
+  rank      : 1 (local_rank: 1)
+  exitcode  : 2 (pid: 1476363)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+------------------------------------------------------------
+Root Cause (first observed failure):
+[0]:
+  time      : 2025-12-04_11:46:37
+  host      : hyperion-252.sw.ehu.es
+  rank      : 0 (local_rank: 0)
+  exitcode  : 2 (pid: 1476362)
+  error_file: <N/A>
+  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
+============================================================
+srun: error: hyperion-252: task 0: Exited with exit code 1

discriminative/eu/fine-tuning/train/model-transfer-2gpu.sh

@@ -0,0 +1,54 @@
+#!/usr/bin/env bash
+#SBATCH --qos=regular
+#SBATCH --job-name=nli-model-transfer
+#SBATCH --cpus-per-task=4
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:2
+#SBATCH --constraint=a100
+#SBATCH --output=/scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/fine-tuning/train/logs/model_transfer-2gpu.log
+#SBATCH --error=/scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/fine-tuning/train/logs/model_transder-2gpu.err
+#SBATCH --time=01-00:00:00 #ee-hh:mm:ss
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+
+
+source /scratch/jbengoetxea/phd/.phd_venv/bin/activate 
+
+export TORCHDYNAMO_DISABLE=1
+
+for seed in 23 27 33
+do
+    for model in answerdotai/ModernBERT-large
+    # microsoft/mdeberta-v3-base FacebookAI/xlm-roberta-large FacebookAI/xlm-roberta-base 
+    do
+    MASTER_PORT=9327
+    MAIN_PROCESS_IP=$(scontrol show hostnames $SLURM_JOB_NODELIST | head -n 1)
+    srun accelerate launch \
+        --mixed_precision bf16 \
+        --dynamo_backend "no" \
+        fine-tuning/run_xnli_eus.py \
+        --model_name_or_path $model \
+        --language eu \
+        --train_language en \
+        --do_train \
+        --do_eval \
+        --per_device_train_batch_size 16 \
+        --learning_rate 10e-6 \
+        --num_train_epochs 10.0 \
+        --max_seq_length 128 \
+        --bf16 \
+        --output_dir /scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/models/model-transfer/$model/$seed \
+        --save_steps 50000 \
+        --eval_strategy steps \
+        --save_strategy steps \
+        --load_best_model_at_end true \
+        --metric_for_best_model accuracy \
+        --seed $seed \
+        --eval_steps 5000 \
+        --logging_steps 25 \
+        --torch_compile false \
+        --save_total_limit 2 
+    done
+done

discriminative/eu/fine-tuning/train/model-transfer.sh

@@ -0,0 +1,62 @@
+#!/usr/bin/env bash
+#SBATCH --partition=hitz-exclusive
+#SBATCH --account=hitz-exclusive
+#SBATCH --job-name=nli-model-transfer
+#SBATCH --cpus-per-task=4
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:4
+#SBATCH --constraint=a100-sxm4
+#SBATCH --output=/scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/fine-tuning/train/logs/model_transfer.log
+#SBATCH --error=/scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/fine-tuning/train/logs/model_transder.err
+#SBATCH --time=01-00:00:00 #ee-hh:mm:ss
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+
+
+source /scratch/jbengoetxea/phd/.phd_venv_new/bin/activate 
+
+export TORCHDYNAMO_DISABLE=1
+
+for seed in 23
+# 23 27 33
+do
+    for model in jhu-clsp/mmBERT-base
+    # microsoft/mdeberta-v3-base FacebookAI/xlm-roberta-large FacebookAI/xlm-roberta-base 
+    do
+    MASTER_PORT=9327
+    MAIN_PROCESS_IP=$(scontrol show hostnames $SLURM_JOB_NODELIST | head -n 1)
+    srun accelerate launch \
+        --num_processes 4 \
+        --num_machines 1 \
+        --mixed_precision bf16 \
+        --dynamo_backend "no" \
+        --rdzv_backend static \
+        --main_process_ip $MAIN_PROCESS_IP \
+        --main_process_port $MASTER_PORT \
+        --machine_rank $SLURM_NODEID \
+        fine-tuning/run_xnli_eus.py \
+        --model_name_or_path $model \
+        --language eu \
+        --train_language en \
+        --do_train \
+        --do_eval \
+        --per_device_train_batch_size 8 \
+        --learning_rate 10e-6 \
+        --num_train_epochs 10.0 \
+        --max_seq_length 128 \
+        --output_dir /scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/models/model-transfer/$model/$seed \
+        --save_steps 50000 \
+        --eval_strategy steps \
+        --save_strategy steps \
+        --bf16 \
+        --load_best_model_at_end true \
+        --metric_for_best_model accuracy \
+        --seed $seed \
+        --logging_steps 25 \
+        --eval_steps 5000 \
+        --torch_compile false \
+        --save_total_limit 2 
+    done
+done

discriminative/eu/fine-tuning/train/translate-train.sh

@@ -0,0 +1,75 @@
+#!/bin/bash
+#SBATCH --partition=hitz-exclusive
+#SBATCH --account=hitz-exclusive
+#SBATCH --job-name=var-nli-tra
+#SBATCH --cpus-per-task=1
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --time=10:00:00
+#SBATCH --mem=32GB
+#SBATCH --gres=gpu:2
+#SBATCH --output=translate-train.log
+#SBATCH --error=translate-train.err
+
+source /scratch/jbengoetxea/phd/.phd_venv_new/bin/activate
+
+export TORCHDYNAMO_DISABLE=1
+
+for seed in 23 27 33
+do
+    for model in jhu-clsp/mmBERT-base
+    do
+    MASTER_PORT=9327
+    MAIN_PROCESS_IP=$(scontrol show hostnames $SLURM_JOB_NODELIST | head -n 1)
+    srun accelerate launch \
+        --num_machines 1 \
+        --mixed_precision bf16 \
+        --dynamo_backend "no" \
+        --rdzv_backend static \
+        --main_process_ip $MAIN_PROCESS_IP \
+        --main_process_port $MASTER_PORT \
+        --machine_rank $SLURM_NODEID \
+        fine-tuning/run_xnli_eus.py \
+        --model_name_or_path $model \
+        --language eu \
+        --train_language eu \
+        --do_train \
+        --do_eval \
+        --per_device_train_batch_size 32 \
+        --learning_rate 5e-5 \
+        --num_train_epochs 10.0 \
+        --max_seq_length 128 \
+        --output_dir /scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/models/translate-train/$model/$seed \
+        --save_steps 50000 \
+        --load_best_model_at_end 1 \
+        --metric_for_best_model accuracy \
+        --seed $seed \
+        --eval_strategy steps \
+        --bf16 \
+        --eval_steps 5000 \
+        --save_total_limit 2 
+    done
+
+    # for model in ixa-ehu/roberta-eus-euscrawl-large-cased FacebookAI/xlm-roberta-large
+    # # microsoft/mdeberta-v3-base FacebookAI/xlm-roberta-large ixa-ehu/roberta-eus-euscrawl-large-cased
+    # do
+    #     python fine-tuning/run_xnli_eus.py \
+    #     --model_name_or_path $model \
+    #     --language eu \
+    #     --train_language eu_all \
+    #     --do_train \
+    #     --do_eval \
+    #     --per_device_train_batch_size 32 \
+    #     --learning_rate 10e-6 \
+    #     --num_train_epochs 10.0 \
+    #     --max_seq_length 128 \
+    #     --output_dir /scratch/jbengoetxea/phd/XNLIvar/scripts/discriminative/eu/models/translate-train/eu_all_corrected/$model/$seed \
+    #     --save_steps 50000 \
+    #     --metric_for_best_model accuracy \
+    #     --seed $seed \
+    #     --eval_strategy steps \
+    #     --eval_steps 5000 \
+    #     --save_total_limit 2
+    # done
+
+done