tags, library_name, base_model, widget, license, pipeline_tag

tags	library_name	base_model	widget	license	pipeline_tag
text-generation-inference text-generation peft	transformers	meta-llama/Llama-3.2-1B-Instruct	messages role content user What is your favorite condiment?	cc-by-2.0	text-generation

Model Trained Using AutoTrain

This model is created from llama3.2 1B Instruct. email me at: nazmus.sakib.muzahid@gmail.com

Usage

from transformers import pipeline
import nltk
from nltk.tokenize import sent_tokenize

nltk.download("punkt_tab")

pipe = pipeline("text-generation", max_new_tokens=512, do_sample=False, model="mnsm92/English_To_Bengali_Translation")

def create_messages(ot: str):
    messages = [

        {
            "role": "system",
            "content": "Translate English to Bengali. Keep the meaning same"
        },
        {"role": "user", "content": ot}
    ]

    return messages

text = """
National Citizen Party (NCP) convenor Nahid Islam has said that his party will contest the upcoming 13th Jatiya Sangsad (national parliament) election in alliance with Jamaat-e-Islami and like-minded parties.

However, he emphasised that there is no ideological unity between the NCP and Jamaat or its allies, describing the arrangement instead as an electoral understanding.

Nahid Islam made the remarks at a press conference held at the NCP’s temporary central office in the capital’s Banglamotor area on Sunday night.

Earlier in the afternoon, Jamaat-e-Islami and its like-minded parties had announced at their own press conference that the NCP would be part of their 10-party electoral understanding.

Speaking at the NCP press conference, Nahid Islam outlined the background to the understanding with Jamaat and eight other parties.
"""


sentences = sent_tokenize(text)
sentences = [ create_messages(sent) for sent in sentences]
outputs = pipe(sentences, batch_size=32, truncation=True)
output_texts = ""
for i in outputs:
    output_texts += i[0]['generated_text'][-1]['content']
output_texts.replace("assistant\n\n", " ")

Drop a feedback.