jimmeylove/week6Mli

Sarcastic Reddit AI - Fine-tuned Llama 3.2 1B Model

This model is a fine-tuned version of meta-llama/Llama-3.2-1B-Instruct that has been trained to generate sarcastic Reddit-style responses. It was fine-tuned using LoRA (Low-Rank Adaptation) to maintain the base model's capabilities while specializing in sarcastic responses.

Model Description

• Base Model: meta-llama/Llama-3.2-1B-Instruct
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Training Data: Custom dataset of Reddit-style sarcastic responses
• Special Capabilities:
  • Generates consistently sarcastic responses regardless of input format
  • Works with both questions and statements
  • Produces complete responses that finish naturally

Intended Use

This model is intended for generating sarcastic responses in a Reddit style. It can be used for:

• Entertainment purposes
• Creative writing assistance
• Chatbot applications requiring a sarcastic personality

Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "jimmeylove/week6Mli"
base_model = "meta-llama/Llama-3.2-1B-Instruct"

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(base_model)
model = PeftModel.from_pretrained(base_model, model_name)
tokenizer = AutoTokenizer.from_pretrained(base_model)

# Format prompt
prompt = "On Reddit, someone asked: How do birds fly?\n\nA sarcastic Redditor replied:"

# Generate response
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(
    input_ids=inputs.input_ids,
    attention_mask=inputs.attention_mask,
    max_new_tokens=1000,
    temperature=1.5,
    top_p=0.95,
    do_sample=True,
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Limitations

• The model may occasionally generate non-sarcastic responses
• As with all language models, it may produce inappropriate content
• The model inherits biases from its training data and base model

Training Details

The model was fine-tuned using the following parameters:

• LoRA rank: 8
• Target modules: q_proj, v_proj, k_proj, o_proj, gate_proj, up_proj, down_proj
• Training data: 3000 examples of sarcastic Reddit responses