Added Flores 200 (#717)

* adding relevant languages + sorting everyone
* added chrf++
* updated language list to rm redundancies, renamed task to follow usual pattern
* fix adapter plus manage languages simply - in the future we might want to have a custom enum with one key several values

Author: clefourrier

src/lighteval/metrics/metrics.py

@@ -158,6 +158,14 @@ class Metrics(Enum):
         corpus_level_fn=CorpusLevelTranslationMetric("chrf").compute,
         higher_is_better=True,
     )
+    chrf_plus = CorpusLevelMetric(
+        metric_name="chrf++",
+        sample_level_fn=GenerativePreparator().prepare,
+        category=MetricCategory.GENERATIVE,
+        use_case=MetricUseCase.TRANSLATION,
+        corpus_level_fn=CorpusLevelTranslationMetric("chrf++").compute,
+        higher_is_better=True,
+    )
     copyright = SampleLevelMetricGrouping(
         metric_name=["longest_common_prefix_length", "edit_distance", "edit_similarity"],
         sample_level_fn=StringDistance(

src/lighteval/metrics/metrics_corpus.py

@@ -104,6 +104,8 @@ def get_metric(self):
             return sacrebleu.BLEU(trg_lang=self.lang)
         elif self.metric_type == "chrf":
             return sacrebleu.CHRF()
+        elif self.metric_type == "chrf++":
+            return sacrebleu.CHRF(word_order=2)
         elif self.metric_type == "ter":
             return sacrebleu.TER(asian_support=True if self.lang != "" else False)
         else:

src/lighteval/tasks/multilingual/tasks.py

@@ -21,6 +21,7 @@
 # SOFTWARE.
 
 from functools import partial
+from itertools import combinations
 
 from langcodes import Language as LangCodeLanguage
 from langcodes import standardize_tag
@@ -30,6 +31,7 @@
     multilingual_quasi_exact_match_metric,
     multilingual_quasi_f1_score_metric,
 )
+from lighteval.metrics.metrics import Metrics
 from lighteval.metrics.normalizations import LogProbCharNorm, LogProbPMINorm, LogProbTokenNorm
 from lighteval.tasks.default_prompts import LETTER_INDICES
 from lighteval.tasks.lighteval_task import LightevalTaskConfig
@@ -52,13 +54,14 @@
 from lighteval.tasks.templates.multichoice import get_mcq_prompt_function
 from lighteval.tasks.templates.nli import get_nli_prompt_function
 from lighteval.tasks.templates.qa import get_qa_prompt_function
+from lighteval.tasks.templates.translation import get_translation_prompt_function
 from lighteval.tasks.templates.utils.formulation import (
     CFFormulation,
     HybridFormulation,
     MCFFormulation,
 )
 from lighteval.tasks.templates.utils.translation_literals import TRANSLATION_LITERALS
-from lighteval.utils.language import Language, iso_639_3_ind_to_iso_639_3_macro
+from lighteval.utils.language import Language, iso_639_3_ind_to_iso_639_3_macro, manage_duplicate_language_codes
 
 
 TASKS_TABLE = []
@@ -3903,3 +3906,250 @@
         *hindi_boolq_tasks,
     ]
 )
+
+# ------------------------------- Translation Tasks ------------------------------- #
+flores_200_languages = [
+    # "ace_Arab",
+    "ace_Latn",
+    "acm_Arab",
+    "acq_Arab",
+    "aeb_Arab",
+    "afr_Latn",
+    "ajp_Arab",
+    "aka_Latn",
+    "amh_Ethi",
+    "apc_Arab",
+    "arb_Arab",
+    # "arb_Latn",
+    "ars_Arab",
+    "ary_Arab",
+    "arz_Arab",
+    "asm_Beng",
+    "ast_Latn",
+    "awa_Deva",
+    "ayr_Latn",
+    "azb_Arab",
+    "azj_Latn",
+    "bak_Cyrl",
+    "bam_Latn",
+    "ban_Latn",
+    "bel_Cyrl",
+    "bem_Latn",
+    "ben_Beng",
+    "bho_Deva",
+    # "bjn_Arab",
+    "bjn_Latn",
+    "bod_Tibt",
+    "bos_Latn",
+    "bug_Latn",
+    "bul_Cyrl",
+    "cat_Latn",
+    "ceb_Latn",
+    "ces_Latn",
+    "cjk_Latn",
+    "ckb_Arab",
+    "crh_Latn",
+    "cym_Latn",
+    "dan_Latn",
+    "deu_Latn",
+    "dik_Latn",
+    "dyu_Latn",
+    "dzo_Tibt",
+    "ell_Grek",
+    "eng_Latn",
+    "epo_Latn",
+    "est_Latn",
+    "eus_Latn",
+    "ewe_Latn",
+    "fao_Latn",
+    "fij_Latn",
+    "fin_Latn",
+    "fon_Latn",
+    "fra_Latn",
+    "fur_Latn",
+    "fuv_Latn",
+    "gla_Latn",
+    "gle_Latn",
+    "glg_Latn",
+    "grn_Latn",
+    "guj_Gujr",
+    "hat_Latn",
+    "hau_Latn",
+    "heb_Hebr",
+    "hin_Deva",
+    "hne_Deva",
+    "hrv_Latn",
+    "hun_Latn",
+    "hye_Armn",
+    "ibo_Latn",
+    "ilo_Latn",
+    "ind_Latn",
+    "isl_Latn",
+    "ita_Latn",
+    "jav_Latn",
+    "jpn_Jpan",
+    "kab_Latn",
+    "kac_Latn",
+    "kam_Latn",
+    "kan_Knda",
+    # "kas_Arab",
+    "kas_Deva",
+    "kat_Geor",
+    # "knc_Arab",
+    "knc_Latn",
+    "kaz_Cyrl",
+    "kbp_Latn",
+    "kea_Latn",
+    "khm_Khmr",
+    "kik_Latn",
+    "kin_Latn",
+    "kir_Cyrl",
+    "kmb_Latn",
+    "kmr_Latn",
+    "kon_Latn",
+    "kor_Hang",
+    "lao_Laoo",
+    "lij_Latn",
+    "lim_Latn",
+    "lin_Latn",
+    "lit_Latn",
+    "lmo_Latn",
+    "ltg_Latn",
+    "ltz_Latn",
+    "lua_Latn",
+    "lug_Latn",
+    "luo_Latn",
+    "lus_Latn",
+    "lvs_Latn",
+    "mag_Deva",
+    "mai_Deva",
+    "mal_Mlym",
+    "mar_Deva",
+    # "min_Arab",
+    "min_Latn",
+    "mkd_Cyrl",
+    "plt_Latn",
+    "mlt_Latn",
+    "mni_Beng",
+    "khk_Cyrl",
+    "mos_Latn",
+    "mri_Latn",
+    "mya_Mymr",
+    "nld_Latn",
+    "nno_Latn",
+    "nob_Latn",
+    "npi_Deva",
+    "nso_Latn",
+    "nus_Latn",
+    "nya_Latn",
+    "oci_Latn",
+    "gaz_Latn",
+    "ory_Orya",
+    "pag_Latn",
+    "pan_Guru",
+    "pap_Latn",
+    "pes_Arab",
+    "pol_Latn",
+    "por_Latn",
+    "prs_Arab",
+    "pbt_Arab",
+    "quy_Latn",
+    "ron_Latn",
+    "run_Latn",
+    "rus_Cyrl",
+    "sag_Latn",
+    "san_Deva",
+    "sat_Olck",
+    "scn_Latn",
+    "shn_Mymr",
+    "sin_Sinh",
+    "slk_Latn",
+    "slv_Latn",
+    "smo_Latn",
+    "sna_Latn",
+    "snd_Arab",
+    "som_Latn",
+    "sot_Latn",
+    "spa_Latn",
+    "als_Latn",
+    "srd_Latn",
+    "srp_Cyrl",
+    "ssw_Latn",
+    "sun_Latn",
+    "swe_Latn",
+    "swh_Latn",
+    "szl_Latn",
+    "tam_Taml",
+    "tat_Cyrl",
+    "tel_Telu",
+    "tgk_Cyrl",
+    "tgl_Latn",
+    "tha_Thai",
+    "tir_Ethi",
+    "taq_Latn",
+    "taq_Tfng",
+    "tpi_Latn",
+    "tsn_Latn",
+    "tso_Latn",
+    "tuk_Latn",
+    "tum_Latn",
+    "tur_Latn",
+    "twi_Latn",
+    "tzm_Tfng",
+    "uig_Arab",
+    "ukr_Cyrl",
+    "umb_Latn",
+    "urd_Arab",
+    "uzn_Latn",
+    "vec_Latn",
+    "vie_Latn",
+    "war_Latn",
+    "wol_Latn",
+    "xho_Latn",
+    "ydd_Hebr",
+    "yor_Latn",
+    "yue_Hant",
+    "zho_Hans",
+    # "zho_Hant",
+    "zsm_Latn",
+    "zul_Latn",
+]
+
+
+def flores_adapter(lang1, lang2):
+    return lambda line: {
+        "source_text": line[f"sentence_{lang1}"],
+        "target_text": line[f"sentence_{lang2}"],
+    }
+
+
+flores200_tasks = [
+    LightevalTaskConfig(
+        name=f"flores200:{lang1}-{lang2}",
+        prompt_function=get_translation_prompt_function(
+            source_language=Language(manage_duplicate_language_codes(lang1.split("_")[0])),
+            target_language=Language(manage_duplicate_language_codes(lang2.split("_")[0])),
+            adapter=flores_adapter(lang1, lang2),
+            formulation=CFFormulation(),
+        ),
+        suite=("lighteval",),
+        hf_repo="facebook/flores",
+        hf_subset=f"{lang1}-{lang2}",
+        hf_avail_splits=["dev", "devtest"],
+        evaluation_splits=["devtest"],
+        few_shots_split="dev",
+        few_shots_select=None,
+        generation_size=300,
+        metric=[Metrics.chrf_plus, Metrics.bleu, Metrics.bleu_1, Metrics.bleu_4],
+        stop_sequence=["\n"],
+        trust_dataset=True,
+        version=0,
+    )
+    for (lang1, lang2) in combinations(flores_200_languages, 2)
+]
+
+TASKS_TABLE.extend(
+    [
+        *flores200_tasks,
+    ]
+)