---
language:
- ru
- kbd
license: mit
base_model: facebook/m2m100_1.2B
tags:
- generated_from_trainer
datasets:
- anzorq/ru-kbd
model-index:
- name: m2m100_1.2B_ft_ru-kbd_50K
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# m2m100_418M_ft_ru-kbd_50K

This model is a fine-tuned version of [facebook/m2m100_1.2B](https://huggingface.co/facebook/m2m100_1.2B) on the [anzorq/ru-kbd](https://huggingface.co/datasets/anzorq/ru-kbd) dataset.

## Model description

More information needed

## Intended uses & limitations

More information needed

## Eval
```
  predict_bleu               =    23.3736
  predict_gen_len            =    16.8114
  predict_loss               =     0.9729
  predict_runtime            = 0:03:29.00
  predict_samples            =       1034
  predict_samples_per_second =      4.947
  predict_steps_per_second   =      0.211
```

### Framework versions

- Transformers 4.34.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.14.5
- Tokenizers 0.14.0


### Inference

```bash
pip install transformers sentencepiece
```

```Python
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

model_path = "anzorq/m2m100_1.2B_ft_ru-kbd_50K"
tgt_lang="zu"

tokenizer = AutoTokenizer.from_pretrained('facebook/m2m100_1.2B')
model = AutoModelForSeq2SeqLM.from_pretrained(model_path)
model.to('cuda')

def translate(text, num_beams=4, num_return_sequences=4):
    inputs = tokenizer(text, return_tensors="pt")
    inputs.to('cuda')
    num_return_sequences = min(num_return_sequences, num_beams)

    translated_tokens = model.generate(
        **inputs,
        forced_bos_token_id=tokenizer.lang_code_to_id[tgt_lang],
        num_beams=num_beams,
        num_return_sequences=num_return_sequences
    )

    translations = [tokenizer.decode(translation, skip_special_tokens=True) for translation in translated_tokens]
    return translations
```