David Pomerenke
commited on
Commit
·
e6f1c56
1
Parent(s):
544091e
Use sacrebleu
Browse files- languagebench.py +5 -4
- pyproject.toml +1 -2
- results.json +55 -25
- uv.lock +0 -0
languagebench.py
CHANGED
|
@@ -26,16 +26,17 @@ target_languages = [
|
|
| 26 |
"fra_Latn",
|
| 27 |
"spa_Latn",
|
| 28 |
"cmn_Hans",
|
|
|
|
| 29 |
]
|
| 30 |
|
| 31 |
# setup
|
|
|
|
| 32 |
client = AsyncOpenAI(
|
| 33 |
base_url="https://openrouter.ai/api/v1",
|
| 34 |
api_key=getenv("OPENROUTER_API_KEY"),
|
| 35 |
)
|
| 36 |
-
load_dotenv()
|
| 37 |
cache = Memory(location=".cache", verbose=0).cache
|
| 38 |
-
bleu = evaluate.load("
|
| 39 |
|
| 40 |
|
| 41 |
@cache
|
|
@@ -75,11 +76,11 @@ async def main():
|
|
| 75 |
"model": model,
|
| 76 |
"original_language": original_language,
|
| 77 |
"target_language": target_language,
|
| 78 |
-
"bleu": metrics["
|
| 79 |
}
|
| 80 |
)
|
| 81 |
with open("results.json", "w") as f:
|
| 82 |
-
json.dump(results, f, indent=2)
|
| 83 |
|
| 84 |
|
| 85 |
if __name__ == "__main__":
|
|
|
|
| 26 |
"fra_Latn",
|
| 27 |
"spa_Latn",
|
| 28 |
"cmn_Hans",
|
| 29 |
+
"cmn_Hant",
|
| 30 |
]
|
| 31 |
|
| 32 |
# setup
|
| 33 |
+
load_dotenv()
|
| 34 |
client = AsyncOpenAI(
|
| 35 |
base_url="https://openrouter.ai/api/v1",
|
| 36 |
api_key=getenv("OPENROUTER_API_KEY"),
|
| 37 |
)
|
|
|
|
| 38 |
cache = Memory(location=".cache", verbose=0).cache
|
| 39 |
+
bleu = evaluate.load("sacrebleu")
|
| 40 |
|
| 41 |
|
| 42 |
@cache
|
|
|
|
| 76 |
"model": model,
|
| 77 |
"original_language": original_language,
|
| 78 |
"target_language": target_language,
|
| 79 |
+
"bleu": metrics["score"],
|
| 80 |
}
|
| 81 |
)
|
| 82 |
with open("results.json", "w") as f:
|
| 83 |
+
json.dump(results, f, indent=2, ensure_ascii=False)
|
| 84 |
|
| 85 |
|
| 86 |
if __name__ == "__main__":
|
pyproject.toml
CHANGED
|
@@ -9,8 +9,7 @@ dependencies = [
|
|
| 9 |
"joblib>=1.4.2",
|
| 10 |
"openai>=1.52.2",
|
| 11 |
"pandas>=2.2.3",
|
| 12 |
-
"plotly>=5.24.1",
|
| 13 |
"python-dotenv>=1.0.1",
|
| 14 |
-
"
|
| 15 |
"tqdm>=4.66.6",
|
| 16 |
]
|
|
|
|
| 9 |
"joblib>=1.4.2",
|
| 10 |
"openai>=1.52.2",
|
| 11 |
"pandas>=2.2.3",
|
|
|
|
| 12 |
"python-dotenv>=1.0.1",
|
| 13 |
+
"sacrebleu>=2.4.3",
|
| 14 |
"tqdm>=4.66.6",
|
| 15 |
]
|
results.json
CHANGED
|
@@ -3,150 +3,180 @@
|
|
| 3 |
"model": "openai/gpt-4o-mini",
|
| 4 |
"original_language": "eng_Latn",
|
| 5 |
"target_language": "eng_Latn",
|
| 6 |
-
"bleu":
|
| 7 |
},
|
| 8 |
{
|
| 9 |
"model": "google/gemini-flash-1.5",
|
| 10 |
"original_language": "eng_Latn",
|
| 11 |
"target_language": "eng_Latn",
|
| 12 |
-
"bleu":
|
| 13 |
},
|
| 14 |
{
|
| 15 |
"model": "anthropic/claude-3.5-sonnet",
|
| 16 |
"original_language": "eng_Latn",
|
| 17 |
"target_language": "eng_Latn",
|
| 18 |
-
"bleu":
|
| 19 |
},
|
| 20 |
{
|
| 21 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 22 |
"original_language": "eng_Latn",
|
| 23 |
"target_language": "eng_Latn",
|
| 24 |
-
"bleu":
|
| 25 |
},
|
| 26 |
{
|
| 27 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 28 |
"original_language": "eng_Latn",
|
| 29 |
"target_language": "eng_Latn",
|
| 30 |
-
"bleu":
|
| 31 |
},
|
| 32 |
{
|
| 33 |
"model": "openai/gpt-4o-mini",
|
| 34 |
"original_language": "eng_Latn",
|
| 35 |
"target_language": "deu_Latn",
|
| 36 |
-
"bleu":
|
| 37 |
},
|
| 38 |
{
|
| 39 |
"model": "google/gemini-flash-1.5",
|
| 40 |
"original_language": "eng_Latn",
|
| 41 |
"target_language": "deu_Latn",
|
| 42 |
-
"bleu":
|
| 43 |
},
|
| 44 |
{
|
| 45 |
"model": "anthropic/claude-3.5-sonnet",
|
| 46 |
"original_language": "eng_Latn",
|
| 47 |
"target_language": "deu_Latn",
|
| 48 |
-
"bleu":
|
| 49 |
},
|
| 50 |
{
|
| 51 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 52 |
"original_language": "eng_Latn",
|
| 53 |
"target_language": "deu_Latn",
|
| 54 |
-
"bleu":
|
| 55 |
},
|
| 56 |
{
|
| 57 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 58 |
"original_language": "eng_Latn",
|
| 59 |
"target_language": "deu_Latn",
|
| 60 |
-
"bleu":
|
| 61 |
},
|
| 62 |
{
|
| 63 |
"model": "openai/gpt-4o-mini",
|
| 64 |
"original_language": "eng_Latn",
|
| 65 |
"target_language": "fra_Latn",
|
| 66 |
-
"bleu":
|
| 67 |
},
|
| 68 |
{
|
| 69 |
"model": "google/gemini-flash-1.5",
|
| 70 |
"original_language": "eng_Latn",
|
| 71 |
"target_language": "fra_Latn",
|
| 72 |
-
"bleu":
|
| 73 |
},
|
| 74 |
{
|
| 75 |
"model": "anthropic/claude-3.5-sonnet",
|
| 76 |
"original_language": "eng_Latn",
|
| 77 |
"target_language": "fra_Latn",
|
| 78 |
-
"bleu":
|
| 79 |
},
|
| 80 |
{
|
| 81 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 82 |
"original_language": "eng_Latn",
|
| 83 |
"target_language": "fra_Latn",
|
| 84 |
-
"bleu":
|
| 85 |
},
|
| 86 |
{
|
| 87 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 88 |
"original_language": "eng_Latn",
|
| 89 |
"target_language": "fra_Latn",
|
| 90 |
-
"bleu":
|
| 91 |
},
|
| 92 |
{
|
| 93 |
"model": "openai/gpt-4o-mini",
|
| 94 |
"original_language": "eng_Latn",
|
| 95 |
"target_language": "spa_Latn",
|
| 96 |
-
"bleu":
|
| 97 |
},
|
| 98 |
{
|
| 99 |
"model": "google/gemini-flash-1.5",
|
| 100 |
"original_language": "eng_Latn",
|
| 101 |
"target_language": "spa_Latn",
|
| 102 |
-
"bleu":
|
| 103 |
},
|
| 104 |
{
|
| 105 |
"model": "anthropic/claude-3.5-sonnet",
|
| 106 |
"original_language": "eng_Latn",
|
| 107 |
"target_language": "spa_Latn",
|
| 108 |
-
"bleu":
|
| 109 |
},
|
| 110 |
{
|
| 111 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 112 |
"original_language": "eng_Latn",
|
| 113 |
"target_language": "spa_Latn",
|
| 114 |
-
"bleu":
|
| 115 |
},
|
| 116 |
{
|
| 117 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 118 |
"original_language": "eng_Latn",
|
| 119 |
"target_language": "spa_Latn",
|
| 120 |
-
"bleu":
|
| 121 |
},
|
| 122 |
{
|
| 123 |
"model": "openai/gpt-4o-mini",
|
| 124 |
"original_language": "eng_Latn",
|
| 125 |
"target_language": "cmn_Hans",
|
| 126 |
-
"bleu": 0.
|
| 127 |
},
|
| 128 |
{
|
| 129 |
"model": "google/gemini-flash-1.5",
|
| 130 |
"original_language": "eng_Latn",
|
| 131 |
"target_language": "cmn_Hans",
|
| 132 |
-
"bleu": 0.
|
| 133 |
},
|
| 134 |
{
|
| 135 |
"model": "anthropic/claude-3.5-sonnet",
|
| 136 |
"original_language": "eng_Latn",
|
| 137 |
"target_language": "cmn_Hans",
|
| 138 |
-
"bleu": 0.
|
| 139 |
},
|
| 140 |
{
|
| 141 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 142 |
"original_language": "eng_Latn",
|
| 143 |
"target_language": "cmn_Hans",
|
| 144 |
-
"bleu": 0.
|
| 145 |
},
|
| 146 |
{
|
| 147 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 148 |
"original_language": "eng_Latn",
|
| 149 |
"target_language": "cmn_Hans",
|
| 150 |
-
"bleu": 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 151 |
}
|
| 152 |
]
|
|
|
|
| 3 |
"model": "openai/gpt-4o-mini",
|
| 4 |
"original_language": "eng_Latn",
|
| 5 |
"target_language": "eng_Latn",
|
| 6 |
+
"bleu": 96.0187510193446
|
| 7 |
},
|
| 8 |
{
|
| 9 |
"model": "google/gemini-flash-1.5",
|
| 10 |
"original_language": "eng_Latn",
|
| 11 |
"target_language": "eng_Latn",
|
| 12 |
+
"bleu": 79.64837722618887
|
| 13 |
},
|
| 14 |
{
|
| 15 |
"model": "anthropic/claude-3.5-sonnet",
|
| 16 |
"original_language": "eng_Latn",
|
| 17 |
"target_language": "eng_Latn",
|
| 18 |
+
"bleu": 47.89694173473209
|
| 19 |
},
|
| 20 |
{
|
| 21 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 22 |
"original_language": "eng_Latn",
|
| 23 |
"target_language": "eng_Latn",
|
| 24 |
+
"bleu": 57.08253125905762
|
| 25 |
},
|
| 26 |
{
|
| 27 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 28 |
"original_language": "eng_Latn",
|
| 29 |
"target_language": "eng_Latn",
|
| 30 |
+
"bleu": 71.3986619616758
|
| 31 |
},
|
| 32 |
{
|
| 33 |
"model": "openai/gpt-4o-mini",
|
| 34 |
"original_language": "eng_Latn",
|
| 35 |
"target_language": "deu_Latn",
|
| 36 |
+
"bleu": 42.76912386979146
|
| 37 |
},
|
| 38 |
{
|
| 39 |
"model": "google/gemini-flash-1.5",
|
| 40 |
"original_language": "eng_Latn",
|
| 41 |
"target_language": "deu_Latn",
|
| 42 |
+
"bleu": 48.166702527508484
|
| 43 |
},
|
| 44 |
{
|
| 45 |
"model": "anthropic/claude-3.5-sonnet",
|
| 46 |
"original_language": "eng_Latn",
|
| 47 |
"target_language": "deu_Latn",
|
| 48 |
+
"bleu": 47.56638188073429
|
| 49 |
},
|
| 50 |
{
|
| 51 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 52 |
"original_language": "eng_Latn",
|
| 53 |
"target_language": "deu_Latn",
|
| 54 |
+
"bleu": 38.8670415108337
|
| 55 |
},
|
| 56 |
{
|
| 57 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 58 |
"original_language": "eng_Latn",
|
| 59 |
"target_language": "deu_Latn",
|
| 60 |
+
"bleu": 32.2942935571844
|
| 61 |
},
|
| 62 |
{
|
| 63 |
"model": "openai/gpt-4o-mini",
|
| 64 |
"original_language": "eng_Latn",
|
| 65 |
"target_language": "fra_Latn",
|
| 66 |
+
"bleu": 47.70220301445618
|
| 67 |
},
|
| 68 |
{
|
| 69 |
"model": "google/gemini-flash-1.5",
|
| 70 |
"original_language": "eng_Latn",
|
| 71 |
"target_language": "fra_Latn",
|
| 72 |
+
"bleu": 49.50529382461407
|
| 73 |
},
|
| 74 |
{
|
| 75 |
"model": "anthropic/claude-3.5-sonnet",
|
| 76 |
"original_language": "eng_Latn",
|
| 77 |
"target_language": "fra_Latn",
|
| 78 |
+
"bleu": 50.55719906730571
|
| 79 |
},
|
| 80 |
{
|
| 81 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 82 |
"original_language": "eng_Latn",
|
| 83 |
"target_language": "fra_Latn",
|
| 84 |
+
"bleu": 43.43766704709355
|
| 85 |
},
|
| 86 |
{
|
| 87 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 88 |
"original_language": "eng_Latn",
|
| 89 |
"target_language": "fra_Latn",
|
| 90 |
+
"bleu": 37.38013101452594
|
| 91 |
},
|
| 92 |
{
|
| 93 |
"model": "openai/gpt-4o-mini",
|
| 94 |
"original_language": "eng_Latn",
|
| 95 |
"target_language": "spa_Latn",
|
| 96 |
+
"bleu": 34.65606074843554
|
| 97 |
},
|
| 98 |
{
|
| 99 |
"model": "google/gemini-flash-1.5",
|
| 100 |
"original_language": "eng_Latn",
|
| 101 |
"target_language": "spa_Latn",
|
| 102 |
+
"bleu": 34.49205632717459
|
| 103 |
},
|
| 104 |
{
|
| 105 |
"model": "anthropic/claude-3.5-sonnet",
|
| 106 |
"original_language": "eng_Latn",
|
| 107 |
"target_language": "spa_Latn",
|
| 108 |
+
"bleu": 34.58637890527096
|
| 109 |
},
|
| 110 |
{
|
| 111 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 112 |
"original_language": "eng_Latn",
|
| 113 |
"target_language": "spa_Latn",
|
| 114 |
+
"bleu": 33.41419407814188
|
| 115 |
},
|
| 116 |
{
|
| 117 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 118 |
"original_language": "eng_Latn",
|
| 119 |
"target_language": "spa_Latn",
|
| 120 |
+
"bleu": 29.470460185415075
|
| 121 |
},
|
| 122 |
{
|
| 123 |
"model": "openai/gpt-4o-mini",
|
| 124 |
"original_language": "eng_Latn",
|
| 125 |
"target_language": "cmn_Hans",
|
| 126 |
+
"bleu": 0.7678283495493847
|
| 127 |
},
|
| 128 |
{
|
| 129 |
"model": "google/gemini-flash-1.5",
|
| 130 |
"original_language": "eng_Latn",
|
| 131 |
"target_language": "cmn_Hans",
|
| 132 |
+
"bleu": 0.3178534804335777
|
| 133 |
},
|
| 134 |
{
|
| 135 |
"model": "anthropic/claude-3.5-sonnet",
|
| 136 |
"original_language": "eng_Latn",
|
| 137 |
"target_language": "cmn_Hans",
|
| 138 |
+
"bleu": 0.8670958769249191
|
| 139 |
},
|
| 140 |
{
|
| 141 |
"model": "qwen/qwen-2.5-72b-instruct",
|
| 142 |
"original_language": "eng_Latn",
|
| 143 |
"target_language": "cmn_Hans",
|
| 144 |
+
"bleu": 0.6796400550094367
|
| 145 |
},
|
| 146 |
{
|
| 147 |
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 148 |
"original_language": "eng_Latn",
|
| 149 |
"target_language": "cmn_Hans",
|
| 150 |
+
"bleu": 0.027154305073795664
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"model": "openai/gpt-4o-mini",
|
| 154 |
+
"original_language": "eng_Latn",
|
| 155 |
+
"target_language": "cmn_Hant",
|
| 156 |
+
"bleu": 2.175042632198715
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"model": "google/gemini-flash-1.5",
|
| 160 |
+
"original_language": "eng_Latn",
|
| 161 |
+
"target_language": "cmn_Hant",
|
| 162 |
+
"bleu": 0.3480387797702917
|
| 163 |
+
},
|
| 164 |
+
{
|
| 165 |
+
"model": "anthropic/claude-3.5-sonnet",
|
| 166 |
+
"original_language": "eng_Latn",
|
| 167 |
+
"target_language": "cmn_Hant",
|
| 168 |
+
"bleu": 3.8196828383724886
|
| 169 |
+
},
|
| 170 |
+
{
|
| 171 |
+
"model": "qwen/qwen-2.5-72b-instruct",
|
| 172 |
+
"original_language": "eng_Latn",
|
| 173 |
+
"target_language": "cmn_Hant",
|
| 174 |
+
"bleu": 2.1029807575075994
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"model": "meta-llama/llama-3.1-8b-instruct",
|
| 178 |
+
"original_language": "eng_Latn",
|
| 179 |
+
"target_language": "cmn_Hant",
|
| 180 |
+
"bleu": 0.017008567925605175
|
| 181 |
}
|
| 182 |
]
|
uv.lock
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|