AlekseyCalvin/Lyrical_Bolmo_7b_ORPO

Model Card for BYTE LYRICAL TRANSLATION MODEL Var.2 (SFT stage)

This model is a fine-tuned version of allenai/Bolmo-7B. It has been trained using TRL.

Installation

Bolmo models have been tested with transformers 4.57.3 and Python 3.11:

pip install transformers>=4.57.3

Bolmo additionally requires the xlstm package (which needs Python>=3.11):

pip install xlstm==2.0.4

Inference

You can use this byte-level variant of the LYRICAL Poetry Translation model with the standard HuggingFace transformers library:

from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda"
bolmo = AutoModelForCausalLM.from_pretrained("AlekseyCalvin/Lyrical_Bolmo_7b_SFT_Merged", trust_remote_code=True).to(device)
tokenizer = AutoTokenizer.from_pretrained("AlekseyCalvin/Lyrical_Bolmo_7b_SFT_Merged", trust_remote_code=True)
message = ["Translate the following verses: Совершить ли мне горшочек для вмещения кишочек вымещения червей красоты земли моей  "]
input_ids = tokenizer(message, return_tensors="pt")["input_ids"].to(device)
# `max_new_tokens` is the amount of bytes to generate
response = bolmo.generate(input_ids, max_new_tokens=256, do_sample=True, temperature=0.1)
print(tokenizer.decode(response[0], skip_special_tokens=True))

Training procedure

This model was trained with SFT.

Framework versions

• TRL: 0.26.2
• Sacrebleu (for BLEU Machine Translation evaluations)
• Transformers: 4.57.3
• Pytorch: 2.9.0+cu126
• Datasets: 4.0.0
• Tokenizers: 0.22.1

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}