Views
No views yet
1from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
2
3# Load model and tokenizer
4model_name = "johnlockejrr/marianmt-he2en-nwt"
5tokenizer = AutoTokenizer.from_pretrained(model_name)
6model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
7
8# Hebrew to English translation
9hebrew_text = "שלום עולם"
10inputs = tokenizer(hebrew_text, return_tensors="pt", padding=True)
11outputs = model.generate(**inputs, max_length=128, num_beams=4)
12english_translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
13print(english_translation)1from transformers import pipeline
2
3translator = pipeline("translation", model="johnlockejrr/marianmt-he2en-nwt")
4
5# Hebrew to English
6hebrew_text = "בראשית ברא אלהים את השמים ואת הארץ"
7result = translator(hebrew_text)
8print(result[0]['translation_text'])python inference.py --model_path ./hebrew_english_model --text "שלום עולם" --direction he2en| Hebrew | English Translation |
|---|---|
| שלום עולם | Hello world |
| בראשית ברא אלהים | In the beginning God created |
| אהבה | Love |
1training_args = Seq2SeqTrainingArguments(
2 output_dir="./hebrew_english_model",
3 eval_strategy="steps",
4 eval_steps=500,
5 save_strategy="steps",
6 save_steps=500,
7 learning_rate=2e-5,
8 per_device_train_batch_size=8,
9 per_device_eval_batch_size=8,
10 weight_decay=0.01,
11 save_total_limit=3,
12 num_train_epochs=20,
13 predict_with_generate=True,
14 fp16=True,
15 load_best_model_at_end=True,
16 metric_for_best_model="bleu",
17 greater_is_better=True,
18 early_stopping_patience=3
19)1@misc{hebrew_english_translation_2025,
2 title={Hebrew-English Translation Model},
3 author={johnlockejrr},
4 year={2025},
5 url={https://huggingface.co/johnlockejrr/marianmt-he2en-nwt}
6}