sartajbhuvaji/OLMo-1B-Base-shakespeare

Model Overview

OLMo-1B-Base-Shakespeare is a fine-tuned version of the allenai/OLMo-1B-0724-hf model, retrained on the complete collection of novels by William Shakespeare. The model aims to generate text in the style of Shakespeare's works and has been optimized to capture the linguistic and stylistic nuances present in the original text.

Model Details

• Model Type: Base Model

• Base Model: allenai/OLMo-1B-0724-hf

• Training Dataset: Works by William Shakespeare

• GPU VRAM Requirements: 25 GB

• Intended Use Cases:

  • Creative writing assistance
  • Educational purposes for studying literary styles
  • Text generation in the style of William Shakespeare

Installation

Ensure you have the transformers library installed:

pip install transformers

Inference

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

torch.random.manual_seed(0)

model_name = 'sartajbhuvaji/OLMo-1B-Base-Shakespeare'
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    device_map="cuda",
    torch_dtype="auto",
    trust_remote_code=True,
)
model.to('cuda')

input_text = 'Hello how are you?'
input_ids = tokenizer.encode(input_text, return_tensors='pt').to('cuda')

output = model.generate(input_ids, max_length=100, num_return_sequences=1, no_repeat_ngram_size=2)
generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
print(generated_text)
'''
Hello how are you?
  SECOND GENTLEMAN. I am a gentleman.
    The Duke, my lord, and all the court are yours.

                          Enter a MESSENGER

  THIRD GENTSLE MAN. Here's a messenger. What news? What's the news,
      sir? How doth your lady? Is she well? Or is she
        hears'd, beaten, or slain? The news is, sir
'''

Fientuning Details

• Global Step: 4656
• Train Runtime: 2710.0517 sec
• Train Samples per second: 13.742
• Train Steps per second: 1.718
• Epoch: 3.0

Training Curve