NLLB English-Kimochi Translation Model

This model is a fine-tuned version of facebook/nllb-200-distilled-600M for English-to-Kimochi translation.

Model Details

• Base Model: facebook/nllb-200-distilled-600M
• Language Pair: English (eng_Latn) → Kimochi (kom_Latn)
• Dataset: kariiiiiimu/english-to-kimochi
• Training: 5 epochs

Usage

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("kariiiiiimu/nllb")
model = AutoModelForSeq2SeqLM.from_pretrained("kariiiiiimu/nllb")

# Set source language
tokenizer.src_lang = "eng_Latn"

# Translate
text = "Hello, how are you?"
inputs = tokenizer(text, return_tensors="pt")

# Get target language ID safely
target_lang_id = None
for attr in ['lang_code_to_id', 'lang_code_to_token_id']:
    if hasattr(tokenizer, attr):
        lang_mapping = getattr(tokenizer, attr)
        if "kom_Latn" in lang_mapping:
            target_lang_id = lang_mapping["kom_Latn"]
            break

if target_lang_id:
    outputs = model.generate(**inputs, forced_bos_token_id=target_lang_id)
else:
    outputs = model.generate(**inputs)
    
translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(translation)

Training Configuration

• Max Length: 64
• Batch Size: 2
• Learning Rate: 2e-05
• Epochs: 5