Latest version of translator

translator.py

@@ -46,7 +46,7 @@ def split_into_chunks(text, tokenizer, max_tokens=128):
     return chunks
 
 
-def to_lang_code(texts, lang_code, model, tokenizer, max_tokens=128):
+def to_lang_code(texts, lang_code, model, tokenizer, max_tokens=128, sentence_joiner=" "):
     is_string = isinstance(texts, str)
     if is_string:
         texts = [texts]
@@ -66,7 +66,7 @@ def to_lang_code(texts, lang_code, model, tokenizer, max_tokens=128):
         translated_tokens = model.generate(
             **inputs,
             forced_bos_token_id=tokenizer.lang_code_to_id[lang_code],
-            max_new_tokens=512,
+            max_length=512,  # max_new_tokens=512,
             # max_length=int(len(inputs.tokens()) * 1.25)  # 25% more tokens for the translation just in case
         )
         translated_text = tokenizer.batch_decode(translated_tokens, skip_special_tokens=True)
@@ -75,7 +75,7 @@ def to_lang_code(texts, lang_code, model, tokenizer, max_tokens=128):
     outputs = []
     start = 0
     for length in lengths:
-        outputs.append(" ".join(translated_texts[start:start + length]))
+        outputs.append(sentence_joiner.join(translated_texts[start:start + length]))
         start += length
     return outputs[0] if is_string else outputs
 
@@ -90,6 +90,8 @@ def main(
     dataset_revision: Optional[str]=None,
     source_lang: Optional[str]="eng_Latn",
     target_langs: Optional[Union[list, tuple]]=("nob_Latn", "nno_Latn"),
+    sentence_joiner: Optional[str]=" ",
+    max_tokens_per_chunk: Optional[int]=128,
     batch_size: Optional[int]=24,
     output_dir: Optional[Path]=Path("./"),
 ) -> None:
@@ -101,9 +103,21 @@ def main(
     )
 
     for lang_code in target_langs:
+        lang_code_short = re.split(r"[-_ /]", lang_code)[0]
+        if dataset_config:
+            output_path = output_dir / dataset_config / lang_code_short
+        else:
+            output_path = output_dir / lang_code_short
         for split in dataset_splits:
             ds = load_dataset(dataset_name, name=dataset_config, revision=dataset_revision, split=split)
-            translate = partial(to_lang_code, lang_code=lang_code, model=model, tokenizer=tokenizer)
+            translate = partial(
+                to_lang_code, 
+                lang_code=lang_code, 
+                model=model, 
+                tokenizer=tokenizer, 
+                sentence_joiner=sentence_joiner,
+                max_tokens=max_tokens_per_chunk,
+            )
             ds = ds.map(
                 lambda batch: {
                     column: translate(batch[column])
@@ -113,12 +127,10 @@ def main(
                 batch_size=batch_size,
                 desc=f"Translating to {lang_code} ({split})",
             )
-            lang_code_short = re.split(r"[-_ /]", lang_code)[0]
-            ds.save_to_disk(output_dir / lang_code_short / split, max_shard_size="1GB")
-
+            ds.save_to_disk(output_path / split, max_shard_size="1GB")
             json_filename = f"{lang_code_short}_{split}.json.gz".lower()
             ds.to_pandas().to_json(
-                output_dir / lang_code_short / json_filename, orient='records', lines=True
+                output_path / json_filename, orient='records', lines=True
             )
 
 
@@ -134,6 +146,8 @@ if __name__ == "__main__":
     parser.add_argument('--model_revision')
     parser.add_argument('--source_lang', default="eng_Latn")
     parser.add_argument('--target_langs', default="nob_Latn,nno_Latn", help="Comma separated target languages to translate to")
+    parser.add_argument('--sentence_joiner', default=" ", help="String to join sentences split for translation")
+    parser.add_argument('--max_tokens_per_chunk', default=128, type=int, help="Max number of tokens for each chunk for translation")
     parser.add_argument('--batch_size', '-bs', default=24, type=int, help='Number of inputs per batch for prediction')
     parser.add_argument('--output_dir', '-o', default="./", type=str)
     args = parser.parse_args()
@@ -147,6 +161,8 @@ if __name__ == "__main__":
         model_revision=args.model_revision,
         source_lang=args.source_lang,
         target_langs=args.target_langs.split(","),
+        sentence_joiner=args.sentence_joiner,
+        max_tokens_per_chunk=args.max_tokens_per_chunk,
         batch_size=args.batch_size,
         output_dir=Path(args.output_dir),
     )