---
library_name: peft
base_model: bigscience/bloomz-3b
---
# What it is

The model generates funky marketing emails, see example below.
Implementation: A QLORA fine tuning over a [bigscience/bloomz-3b](https://huggingface.co/bigscience/bloomz-3b), utilizing the [FourthBrainGenAI/MarketMail-AI](https://huggingface.co/datasets/FourthBrainGenAI/MarketMail-AI) dataset with 17 rows.

# Inference example

```python
import torch
from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM, AutoTokenizer

peft_model_id = "borislitvak/boris-bloomz-peft-marketing"
config = PeftConfig.from_pretrained(peft_model_id)
model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=False, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)

# Load the Lora model
model = PeftModel.from_pretrained(model, peft_model_id)

from IPython.display import display, Markdown

def make_inference(product, description):
  batch = tokenizer(f"### INSTRUCTION\nBelow is a product and description, please write a marketing email for this product.\n\n### Product:\n{product}\n### Description:\n{description}\n\n### Marketing Email:\n", return_tensors='pt')

  with torch.cuda.amp.autocast():
    output_tokens = model.generate(**batch, max_new_tokens=200)

  # for Jupyter. If serving in regular Python, remove display/Markdown
  display(Markdown((tokenizer.decode(output_tokens[0], skip_special_tokens=True))))

your_product_name_here = ""
your_product_description_here = ""

make_inference(your_product_name_here, your_product_description_here)
```