EuroEval 15.6.1__py3-none-any.whl → 15.7.1__py3-none-any.whl

euroeval/prompt_templates/sentiment_classification.py

@@ -1,7 +1,7 @@
 """Templates for the Sentiment Analysis task."""
 
 from ..data_models import PromptConfig
-from ..languages import DA, DE, EN, ES, FO, FR, IS, IT, NB, NL, NN, NO, SV
+from ..languages import DA, DE, EN, ES, FI, FO, FR, IS, IT, NB, NL, NN, NO, SV
 
 SENT_TEMPLATES = {
     DA: PromptConfig(
@@ -44,6 +44,16 @@ SENT_TEMPLATES = {
         default_instruction_prompt="Documento: {text}\n\nClasifica el sentimiento del "
         "documento. Responde con {labels_str}, y nada más.",
     ),
+    FI: PromptConfig(
+        default_prompt_label_mapping=dict(
+            positive="positiivinen", neutral="neutrali", negative="negatiivinen"
+        ),
+        default_prompt_prefix="Seuraavassa on arvosteluja ja niiden tunnesävy, joka "
+        "voi olla {labels_str}.",
+        default_prompt_template="Teksti: {text}\nTunnesävy: {label}",
+        default_instruction_prompt="Teksti: {text}\n\nLuokittele arvostelun tunnesävy. "
+        "Vastaa vain {labels_str}, ei muuta.",
+    ),
     FO: PromptConfig(
         default_prompt_label_mapping=dict(
             positive="positivt", neutral="neutralt", negative="negativt"

euroeval/prompt_templates/summarization.py

@@ -1,7 +1,7 @@
 """Templates for the Summarization task."""
 
 from ..data_models import PromptConfig
-from ..languages import DA, DE, EN, ES, FR, IS, IT, NB, NL, NN, NO, SV
+from ..languages import DA, DE, EN, ES, FI, FR, IS, IT, NB, NL, NN, NO, SV
 
 # TODO: Missing Faroese
 SUMM_TEMPLATES = {
@@ -36,6 +36,14 @@ SUMM_TEMPLATES = {
         "documento anterior.",
         default_prompt_label_mapping=dict(),
     ),
+    FI: PromptConfig(
+        default_prompt_prefix="Seuraavassa on artikkeleita ja niihin liittyviä "
+        "tiivistelmiä.",
+        default_prompt_template="Uutisartikkeli: {text}\nTiivistelmä: {target_text}",
+        default_instruction_prompt="Uutisartikkeli: {text}\n\nKirjoita tiivistelmä "
+        "yllä olevasta artikkelista.",
+        default_prompt_label_mapping=dict(),
+    ),
     FR: PromptConfig(
         default_prompt_prefix="Les documents suivants sont accompagnés d'un résumé.",
         default_prompt_template="Document: {text}\nRésumé: {target_text}",

euroeval/scores.py

@@ -18,6 +18,7 @@ def log_scores(
     metric_configs: list["MetricConfig"],
     scores: list[dict[str, float]],
     model_id: str,
+    model_revision: str,
 ) -> "ScoreDict":
     """Log the scores.
 
@@ -30,13 +31,18 @@ def log_scores(
             The scores that are to be logged. This is a list of dictionaries full of
             scores.
         model_id:
-            The full Hugging Face Hub path to the pretrained transformer model.
+            The model ID of the model that was evaluated.
+        model_revision:
+            The revision of the model.
 
     Returns:
         A dictionary with keys 'raw_scores' and 'total', with 'raw_scores' being
         identical to `scores` and 'total' being a dictionary with the aggregated scores
         (means and standard errors).
     """
+    if model_revision and model_revision != "main":
+        model_id += f"@{model_revision}"
+
     logger.info(f"Finished evaluation of {model_id} on {dataset_name}.")
 
     total_dict: dict[str, float] = dict()

euroeval/task_group_utils/sequence_classification.py

@@ -132,22 +132,23 @@ def extract_labels_from_generation(
         The predicted labels.
     """
     if model_output.scores is not None:
-        return get_closest_logprobs_labels(
+        labels = get_closest_logprobs_labels(
             generation_logprobs=model_output.scores,
             dataset_config=dataset_config,
             first_label_token_mapping=first_label_token_mapping,
         )
-    else:
-        return get_closest_word_edit_labels(
-            generated_sequences=model_output.sequences, dataset_config=dataset_config
-        )
+        if labels is not None:
+            return labels
+    return get_closest_word_edit_labels(
+        generated_sequences=model_output.sequences, dataset_config=dataset_config
+    )
 
 
 def get_closest_logprobs_labels(
     generation_logprobs: list[list[list[tuple[str, float]]]],
     dataset_config: "DatasetConfig",
     first_label_token_mapping: dict[str, str] | bool,
-) -> list[str]:
+) -> list[str] | None:
     """Get the labels with the highest predicted logprob value.
 
     In case a candidate label is split into multiple tokens, we only use the first
@@ -167,7 +168,7 @@ def get_closest_logprobs_labels(
             mapping is outputted then the model will always output scores).
 
     Returns:
-        The predicted labels.
+        The predicted labels, or None if labels could not be extracted.
 
     Raises:
         InvalidBenchmark:
@@ -193,10 +194,7 @@ def get_closest_logprobs_labels(
             # We want to use the first generated label which contains a unique candidate
             # label, as the output label
             output_label: str | None = None
-            previously_generated_labels: list[str] = list()
-            for label_idx, generated_label in enumerate(generated_labels):
-                generated_label = "".join(previously_generated_labels) + generated_label
-
+            for generated_label in generated_labels:
                 # Get the candidate labels that starts with the generated label
                 if isinstance(first_label_token_mapping, dict):
                     if any(
@@ -222,31 +220,28 @@ def get_closest_logprobs_labels(
                         if candidate_label.startswith(generated_label)
                     }
 
-                # If we can uniquely determine the output label, we break the loop. If
-                # there are multiple possible labels then we store the current one, and
-                # concatenate it with the next generated label. We can only do this if
-                # the current one is the first one, however, since we're using greedy
-                # sampling. In case this happens for a label that is not the first one,
-                # we warn the user.
+                # If we can uniquely determine the output label, we break the loop.
                 if len(candidate_output_labels) == 1:
                     output_label = candidate_output_labels.pop()
                     break
+
+                # If we have multiple candidate labels, we cannot uniquely determine the
+                # output label, so we abandon extracting the labels using logprobs and
+                # fall back to using word edit distance.
                 elif len(candidate_output_labels) > 1:
-                    if label_idx == 0:
-                        previously_generated_labels.append(generated_label)
-                    else:
-                        output_label = candidate_output_labels.pop()
-                        candidate_output_labels.add(output_label)
-                        raise InvalidBenchmark(
-                            "Multiple candidate labels found for the generated label "
-                            f"{generated_label!r}: {candidate_output_labels}. Since "
-                            "this is not the first generated label, we cannot "
-                            "concatenate it with the next generated label. We are thus "
-                            f"forced to use the arbitrary {output_label!r} as the "
-                            "output label, potentially resulting in worse performance. "
-                            "Please report this issue to the EuroEval team at "
-                            "github.com/EuroEval/EuroEval/issues."
-                        )
+                    log_once(
+                        "Multiple candidate labels found for the generated label "
+                        f"{generated_label!r}: {candidate_output_labels}. This means "
+                        "that using logprobs to extract the labels is not reliable, "
+                        "and we will instead fall back to extracting the labels "
+                        "using word edit distance.",
+                        level=logging.DEBUG,
+                    )
+                    return None
+
+                # If no candidate label is found, we ignore the generated label, as it
+                # basically means that the model is just really bad at generating
+                # labels.
                 elif len(candidate_output_labels) == 0:
                     logger.debug(
                         f"No candidate label found for the generated label "

euroeval/task_group_utils/text_to_text.py

@@ -10,11 +10,7 @@ from evaluate import EvaluationModule
 from ..constants import METRIC_ATTRIBUTES_TAKING_UP_MEMORY
 from ..data_models import BenchmarkConfig, DatasetConfig, GenerativeModelOutput
 from ..exceptions import InvalidBenchmark
-from ..utils import (
-    HiddenPrints,
-    clear_memory,
-    raise_if_model_output_contains_nan_values,
-)
+from ..utils import HiddenPrints, raise_if_model_output_contains_nan_values
 
 if t.TYPE_CHECKING:
     from transformers.trainer_utils import EvalPrediction
@@ -89,20 +85,8 @@ def compute_metrics(
                     score_dict: dict[str, float] | None = metric.compute(
                         predictions=predictions, references=labels, **cfg.compute_kwargs
                     )
-
-                # Clear the cache of the BERTScorer to avoid memory leaks
-                for attribute in METRIC_ATTRIBUTES_TAKING_UP_MEMORY:
-                    if hasattr(metric, attribute):
-                        delattr(metric, attribute)
-
-                clear_memory()
                 break
             except Exception as e:
-                # Clear the cache of the BERTScorer to avoid memory leaks
-                if hasattr(metric, "cached_bertscorer"):
-                    del metric.cached_bertscorer
-                    clear_memory()
-
                 oom_error = [
                     "CUDA out of memory",
                     "CUDA error",
@@ -111,16 +95,7 @@ def compute_metrics(
                 if not any(error in str(e) for error in oom_error):
                     raise InvalidBenchmark(str(e))
 
-                if cfg.compute_kwargs.get("batch_size", 1) > 1:
-                    batch_size = cfg.compute_kwargs["batch_size"]
-                    cfg.compute_kwargs["batch_size"] = batch_size // 2
-                    logger.debug(
-                        "Out of memory error occurred during the computation of "
-                        f"the metric {cfg.pretty_name}. Reducing the batch size to "
-                        f"{cfg.compute_kwargs['batch_size']}."
-                    )
-                elif cfg.compute_kwargs.get("device", "cpu") != "cpu":
-                    cfg.compute_kwargs["batch_size"] = 32
+                if cfg.compute_kwargs.get("device", "cpu") != "cpu":
                     cfg.compute_kwargs["device"] = "cpu"
                     logger.debug(
                         "Out of memory error occurred during the computation of "
@@ -129,6 +104,14 @@ def compute_metrics(
                     )
                 else:
                     raise InvalidBenchmark(str(e))
+            finally:
+                for attribute in METRIC_ATTRIBUTES_TAKING_UP_MEMORY:
+                    if hasattr(metric, attribute):
+                        logger.debug(
+                            f"Deleting the {attribute!r} attribute of the metric "
+                            f"{cfg.pretty_name} to free up memory."
+                        )
+                        delattr(metric, attribute)
 
         # The metric returns None if we are running on multi-GPU and the current
         # process is not the main process

euroeval/tasks.py

@@ -142,7 +142,7 @@ SUMM = Task(
             huggingface_id="bertscore",
             results_key="f1",
             compute_kwargs=dict(
-                model_type="microsoft/mdeberta-v3-base", device="auto", batch_size=32
+                model_type="microsoft/mdeberta-v3-base", device="auto", batch_size=1
             ),
         ),
         MetricConfig(

euroeval/tokenization_utils.py

@@ -7,6 +7,7 @@ import typing as t
 import torch
 
 from .constants import TASK_GROUPS_USING_LOGPROBS
+from .enums import GenerativeType
 from .exceptions import InvalidModel
 from .utils import log_once
 
@@ -14,7 +15,7 @@ if t.TYPE_CHECKING:
     from transformers.tokenization_utils import PreTrainedTokenizer
     from transformers.tokenization_utils_base import PreTrainedTokenizerBase
 
-    from .data_models import DatasetConfig
+    from .data_models import DatasetConfig, ModelConfig
 
 
 logger = logging.getLogger("euroeval")
@@ -254,35 +255,50 @@ def get_end_of_chat_token_ids(tokenizer: "PreTrainedTokenizer") -> list[int] | N
 
 
 def get_first_label_token_mapping(
-    dataset_config: "DatasetConfig", tokenizer: "PreTrainedTokenizer | None"
+    dataset_config: "DatasetConfig",
+    model_config: "ModelConfig",
+    tokenizer: "PreTrainedTokenizer | None",
+    generative_type: "GenerativeType | None",
 ) -> dict[str, str] | bool:
     """Check if the model should output scores.
 
     Args:
         dataset_config:
             The dataset configuration.
+        model_config:
+            The model configuration.
         tokenizer:
             The tokenizer, or None if not available.
+        generative_type:
+            The generative type, or None if not available.
 
     Returns:
         A mapping from labels to the first token in each label, or alternatively a
         Boolean value indicating whether the model should output scores (if the mapping
         is outputted then the model will always output scores).
     """
+    if generative_type == GenerativeType.REASONING:
+        log_once(
+            f"The model {model_config.model_id!r} is a reasoning model and "
+            "thus does not support logprobs, so we do not enable it.",
+            level=logging.DEBUG,
+        )
+        return False
+
     # If we do not have any tokenizer, then we cannot check if the model should output
     # scores and we just assume it should if the dataset supports it
     output_scores = dataset_config.task.task_group in TASK_GROUPS_USING_LOGPROBS
     if tokenizer is None:
         if output_scores:
             log_once(
-                "The model will output scores, since the dataset supports it and no "
-                "tokenizer is available.",
+                f"The model {model_config.model_id!r} will output scores, since the "
+                "dataset supports it and no tokenizer is available.",
                 level=logging.DEBUG,
             )
         else:
             log_once(
-                "The model will not output scores, since the dataset does not support "
-                "it and no tokenizer is available.",
+                f"The model {model_config.model_id!r} will not output scores, since "
+                "the dataset does not support it and no tokenizer is available.",
                 level=logging.DEBUG,
             )
         return output_scores

{euroeval-15.6.1.dist-info → euroeval-15.7.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: EuroEval
-Version: 15.6.1
+Version: 15.7.1
 Summary: The robust European language model benchmark.
 Project-URL: Repository, https://github.com/EuroEval/EuroEval
 Project-URL: Issues, https://github.com/EuroEval/EuroEval/issues
@@ -32,7 +32,7 @@ Requires-Python: <4.0,>=3.10
 Requires-Dist: accelerate>=0.34.2
 Requires-Dist: bert-score>=0.3.13
 Requires-Dist: click>=8.1.3
-Requires-Dist: datasets>=2.15.0
+Requires-Dist: datasets>=3.5.0
 Requires-Dist: demjson3>=3.0.6
 Requires-Dist: evaluate>=0.4.1
 Requires-Dist: huggingface-hub>=0.30.1
@@ -239,6 +239,18 @@ A huge thank you to all the contributors who have helped make this project a suc
 <a href="https://github.com/peregilk"><img src="https://avatars.githubusercontent.com/u/9079808" width=50 alt="Contributor avatar for peregilk"/></a>
 <a href="https://github.com/Rijgersberg"><img src="https://avatars.githubusercontent.com/u/8604946" width=50 alt="Contributor avatar for Rijgersberg"/></a>
 
+
+### Contribute to EuroEval
+
+We welcome contributions to EuroEval! Whether you're fixing bugs, adding features, or
+contributing new datasets, your help makes this project better for everyone.
+
+- **General contributions**: Check out our [contribution guidelines](CONTRIBUTING.md)
+  for information on how to get started.
+- **Adding datasets**: If you're interested in adding a new dataset to EuroEval, we have
+  a [dedicated guide](NEW_DATASET_GUIDE.md) with step-by-step instructions.
+
+
 ### Special Thanks
 - Thanks to [Google](https://google.com/) for sponsoring Gemini credits as part of their
   [Google Cloud for Researchers Program](https://cloud.google.com/edu/researchers).

{euroeval-15.6.1.dist-info → euroeval-15.7.1.dist-info}/RECORD

@@ -1,37 +1,39 @@
 euroeval/__init__.py,sha256=NiT6S4II1YpnNl5KFHDNogE-rvVkOHQy5pR483eq_do,2581
 euroeval/benchmark_config_factory.py,sha256=JCjJS2pjtiuQ6tpwZ_DJFvNzwdbZu5YdJcHhFz-q6eU,12562
-euroeval/benchmarker.py,sha256=7LVFr7zL7OeJPs7WVYwekNnEmiIKPXHydcbAkW99MUk,48080
+euroeval/benchmarker.py,sha256=OnjGVblWW20wSmA7Tr2c-qE3g8FIjxW6wTJySAcGxVk,48492
 euroeval/callbacks.py,sha256=F1AJCLB8FJpxqYprwLi_PsH4Bc0x4lyR8UiTG-GlFLY,2452
 euroeval/cli.py,sha256=EMB6g6kRvxIqlfYLSoMzwLAtEd-fqXipo4A_HTkhjkA,8575
-euroeval/constants.py,sha256=t2mAT8tE3Dn2lXWHTnaFoaOIaUcdiBjJTASCt7nSdkg,1984
-euroeval/data_loading.py,sha256=7xXdoFSvEDzpw1FNR8E8YV4c9Vy86hlU5-qLm9RUejE,3318
-euroeval/data_models.py,sha256=oZLrGg1dhIIwbgtEzq4U_fu_ZbBsz35mrqsyizuZNPw,23138
+euroeval/constants.py,sha256=p6kp_R6-Tq5LBvyXyT6Sa6N3SkjEElGS2LSZRBoQaYs,1985
+euroeval/data_loading.py,sha256=L_REtxefte5Ke4xE_Cz01zkfCyKlOYhSqT5ZXXulHPc,3992
+euroeval/data_models.py,sha256=t5FwpGxiSIMe7iKae-tT7usUWki-ILzAFFm7dPJoFsk,22973
 euroeval/enums.py,sha256=L9LcNeruuhHvze9vKRogXY9vonRzoBqDzWSP6hxKQ7A,3195
 euroeval/exceptions.py,sha256=LRd7HoudupRp5-AX3L0X4hIAWCa6JVx-LViHPg7u7dg,5821
 euroeval/finetuning.py,sha256=IieAhgvxjeLHAHBief1Ay-STcCosQmrDHFTRTXFZX0Q,10743
 euroeval/generation.py,sha256=LSsskfLjIJ-c3gQxmr7eiAobPOm-5bU9vnR7uHQ7XmU,10745
+euroeval/generation_utils.py,sha256=zRsaOHcbhysbMa983BZXxfd-qMe4NYts-ZbQxfvNTK4,13310
 euroeval/human_evaluation.py,sha256=VGvw1X6Mkdf22r-THSNWXMIqyJP44yh4rW53vq-0huo,27681
 euroeval/languages.py,sha256=IQUbGMyn7pxAyM70M0FTO80m92Q4KgIU604MJhVia-Q,8513
 euroeval/model_cache.py,sha256=n39yFpZkudBCVwz1EQpZ-g5BQtlQemQ5nP3IiFKJZHg,8275
 euroeval/model_config.py,sha256=64KKHPTrpsFhFAANtBnAKkOs7PWZ50GXkXeDl4jICgs,2748
 euroeval/model_loading.py,sha256=B6dyjYO0Dg7NOcUXls8Sjwe6W0c2UqJ1OGw-RkzoSSQ,2239
-euroeval/scores.py,sha256=OL1MPVSgBySc9gMGeZBnj_j6-EvpDtEOwjO12IgeP6o,2899
+euroeval/scores.py,sha256=TovjCZD8wmGrIjA4v5oAQp18P5KVcHvakkByDh0Hstk,3059
 euroeval/speed_benchmark.py,sha256=J7VKWMf7GU_l0lRR8f0QeUr_vAaBQqTbgQ_yToHhp_0,3980
-euroeval/tasks.py,sha256=VVXFDcEM250KTGXd1pxQb8vwdia4ZJxgTUY5Kdsa-ik,7070
-euroeval/tokenization_utils.py,sha256=PNuS-FTdVrL9TWNDGlq42MvUggKwmyYM0BnC5I37IO0,11876
+euroeval/tasks.py,sha256=87gbe__K5KNIb1aBSuwGnMPmZgamJFecNNYmNgMxaVo,7069
+euroeval/tokenization_utils.py,sha256=fbMVAOkRdcpf9L2SVechPpmWYgDXgQcc-sDrYu21wFI,12487
 euroeval/types.py,sha256=E0JhLfg-ek5pdFcYJbnGRUSodHxkuR3o8XGuIrBcuRM,2485
 euroeval/utils.py,sha256=DyWhtdFlAM1TZuiYXWNPN8KxNrZGNa-J3WfS6DGwkvM,10467
 euroeval/benchmark_modules/__init__.py,sha256=TNO-sNDwlXE-LMFXfwwqjQqUy55gywSmwRBcoPUFuaU,236
 euroeval/benchmark_modules/base.py,sha256=LcG46I2O5wcvu_3T_irBY6VkUhWVPKifBhcP-ln93TA,10798
 euroeval/benchmark_modules/fresh.py,sha256=_LWmpqiNGGTA-NoVC0v3-fS1sraDS9n-pgKUzz89jVk,9919
 euroeval/benchmark_modules/hf.py,sha256=yFApLL4_ia5Kw2iat5RSI8h5RhI4OP04HlzYidlhBCs,44012
-euroeval/benchmark_modules/litellm.py,sha256=wohdi1WoeJ-JEdQLgg2q3JbZJA77XO7yGZaTRvbRU4o,47575
-euroeval/benchmark_modules/vllm.py,sha256=FTpwal5WdrVsOpkjm_RXwf6-2PrNrrP1LO6BVGYb6GE,48086
+euroeval/benchmark_modules/litellm.py,sha256=v_rbCm2FiTMqcUui_09k3E1-s5uOmbfAvSy2c7Mm0_E,42636
+euroeval/benchmark_modules/vllm.py,sha256=Q-3vtZz5XxQQImJxOiF0XDrQ4T_p0bkgdPw1Jobgu3s,39380
 euroeval/dataset_configs/__init__.py,sha256=fkD1hzW7szJLc1MdK-AY4EBFWBUX5Z8t4f9uBHQnRvU,1858
 euroeval/dataset_configs/danish.py,sha256=MTt9EcriSer0QaFQ7_6evYxh-g9OPjroWegYdFpiKag,3395
-euroeval/dataset_configs/dutch.py,sha256=N3zL0vGe4OyPgVU_AiYNNfk96jSc_JDtKrVIHbaEYCU,3536
+euroeval/dataset_configs/dutch.py,sha256=r21nxEvMmBkKqPXVW082batPsxJ9d0RB4DzngOTMJSk,3185
 euroeval/dataset_configs/english.py,sha256=yHw7D0zSNVbiSBAjR1mWX4V5FSkhqy4y-o-pnyWCLxE,2323
 euroeval/dataset_configs/faroese.py,sha256=QQgLe5gv0f3AtXe5rV65xZ98gFgyITQPDr3UwO4Bnv4,1350
+euroeval/dataset_configs/finnish.py,sha256=BIfcxdgJu4CfevHAjzwH7cYd8Xd9DGrm49lcJZcGVQM,1730
 euroeval/dataset_configs/french.py,sha256=ATsj8_9_GxFTQgmfrniPQFZ1R9hoQCI1_ieWTnscFHU,2382
 euroeval/dataset_configs/german.py,sha256=QO6PrBQY6kyZeQMU1vg6KrC_sKyj9U2ukS9nbKO19is,2560
 euroeval/dataset_configs/icelandic.py,sha256=mncl7X4yO9gBmYqXMBfm7FKU1jcKryerSgd0dqlIA_4,4198
@@ -40,20 +42,20 @@ euroeval/dataset_configs/norwegian.py,sha256=2SD5681gZFa1Ig-AEpnyStbivan_bq_Pada
 euroeval/dataset_configs/spanish.py,sha256=fc0dHWU7-g_p6kaSGA8nD1vLVQF_yqR2PkixrYyWywc,2212
 euroeval/dataset_configs/swedish.py,sha256=SOD2nKQTVwTpTvr362mDPHon42kr9vWs5C0mK02Fh-o,2811
 euroeval/prompt_templates/__init__.py,sha256=HWMZpybxs2xHPnVeJ43893conARahIVLWNXeRhXEGZw,357
-euroeval/prompt_templates/linguistic_acceptability.py,sha256=sx_WqLm7N6Thll6COUCCA0lXe9RMZ7WhoH6X498pixM,6232
-euroeval/prompt_templates/multiple_choice.py,sha256=H0CDQPs_WzgSJ7oI_FBzHs0TOF0Na2qZYJLhDC7S8tk,4710
-euroeval/prompt_templates/named_entity_recognition.py,sha256=T65oFEtVT8JRF9c7bq2nPm233rftPdEAGic0DU-toko,11835
-euroeval/prompt_templates/reading_comprehension.py,sha256=WbQoal_tjoTt7qsmSZXEWwlI77vgiANcZoZC1l1AZjc,6090
-euroeval/prompt_templates/sentiment_classification.py,sha256=LcFD89e5nMOv4u-Unj8_jHpNjKMmgKPEfz0-e39VbsM,6639
-euroeval/prompt_templates/summarization.py,sha256=eX0uUTf_5Xorm6f_TlBBNwLC9zKvR7YJkP0RSaLWgIw,4585
+euroeval/prompt_templates/linguistic_acceptability.py,sha256=FAIJKS26EVRxlLHk1C3lN0GDtd5AM0MwvaMf-NNIxfU,6677
+euroeval/prompt_templates/multiple_choice.py,sha256=6iEqiPpT-3WJN_gsyhyapnwsrcsYGdVkSkzwn-VKKxw,5101
+euroeval/prompt_templates/named_entity_recognition.py,sha256=Xd6gBJD2e1l8-We2Ujor7crRUBcbgnNeeVknBIrTMJo,12737
+euroeval/prompt_templates/reading_comprehension.py,sha256=yLqryWQAW04GULz_EyNDLOS7ZrDUeasuLFt-dtqCnYk,6585
+euroeval/prompt_templates/sentiment_classification.py,sha256=LDOwjGQ2kqhwgNyphPywQeolwNB09o-xYWc9RUbzc84,7136
+euroeval/prompt_templates/summarization.py,sha256=mcWeKNhGWmp7IG_iY64T-VOSabQg5wKddjSbJNYFDp8,4984
 euroeval/task_group_utils/__init__.py,sha256=CorGVkixkoEDOQuDsrOGlTmF1zmM0wnGHs8psWTfD28,72
 euroeval/task_group_utils/multiple_choice_classification.py,sha256=nB78TzOgd0HBvTclmjOYJid9ZVAgu8IHZsqB_n1SAZU,6178
 euroeval/task_group_utils/question_answering.py,sha256=kZBABJ_WYNTH4Xgo2jIvfx7iYvfoGt0EUObSaXRCGmk,27700
-euroeval/task_group_utils/sequence_classification.py,sha256=gqd0-l5o7vAY5QIpGSkSqwJwez3Y0r5SqOiywfPNW8A,12239
-euroeval/task_group_utils/text_to_text.py,sha256=QECnGdZ0YLjsbMc6LwXqVi4KMuITdiOjmJUNQtAAOW0,5712
+euroeval/task_group_utils/sequence_classification.py,sha256=xPz1gJioK96iv2bNoDWiC2EJkhRvRd7QZNgY8bT237c,11703
+euroeval/task_group_utils/text_to_text.py,sha256=Nu1_qRPLbboCd9Q5rxqY4fQFJ_aGXu80aWQqoTG1cYc,5047
 euroeval/task_group_utils/token_classification.py,sha256=3idWB81Fcx9UhTuk-gxMfXENrCBmiWBDUWdULXoIhpw,17863
-euroeval-15.6.1.dist-info/METADATA,sha256=4i98IBxn6yWh4ugBW-SnljmDfKEXBSfRGjZyf_dlOUs,13183
-euroeval-15.6.1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-euroeval-15.6.1.dist-info/entry_points.txt,sha256=tKQRxN0HX2mGtbZbZQdCRFUDZIecA_z4mZduueor3Ug,135
-euroeval-15.6.1.dist-info/licenses/LICENSE,sha256=oZp5fpOSQ7w-vFui8QNwrBIosrO7cnpArItdbvn52Ao,1082
-euroeval-15.6.1.dist-info/RECORD,,
+euroeval-15.7.1.dist-info/METADATA,sha256=Fj6QejwQCK0zGuP_DHSQ7sul195ivUqOUCT5AVxgLSI,13669
+euroeval-15.7.1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+euroeval-15.7.1.dist-info/entry_points.txt,sha256=tKQRxN0HX2mGtbZbZQdCRFUDZIecA_z4mZduueor3Ug,135
+euroeval-15.7.1.dist-info/licenses/LICENSE,sha256=oZp5fpOSQ7w-vFui8QNwrBIosrO7cnpArItdbvn52Ao,1082
+euroeval-15.7.1.dist-info/RECORD,,