FlameF0X / Martico2432-Lizard-17m

This is a model made by Martico2432, I, FlameF0X, just try to add Transformers compatibility. After i manage to do this i would send the repo to Martico2432 and i would make this repo private or delete it. I am just fixing some stuff.

Lizard

A lightweight 17m parameter coding model.

Model Details

Model Description

• Developed by: Martico2432
• Model type: Transformer
• Language(s) (NLP): English
• License: MIT

Model loading

#this doesnt work btw
import sys
import os
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig

# Add the path to your local model directory
model_path = "FlameF0X/Martico2432-Lizard-17m"  # Replace with your actual path
sys.path.append(model_path)

# Import your custom classes
from configuration_lizard import LizardConfig
from modeling_lizard import LizardForCausalLM

# Register the custom model and config BEFORE loading
AutoConfig.register("lizard", LizardConfig)
AutoModelForCausalLM.register(LizardConfig, LizardForCausalLM)

try:
    model = AutoModelForCausalLM.from_pretrained(model_path)
    tokenizer = AutoTokenizer.from_pretrained(model_path)
    
    # Generate text
    prompt = "Once upon a time"
    input_ids = tokenizer(prompt, return_tensors="pt").input_ids
    generated_ids = model.generate(input_ids, max_new_tokens=50)
    result = tokenizer.decode(generated_ids[0], skip_special_tokens=True)
    print(result)
    
except Exception as e:
    print(f"An error occurred: {e}")
    import traceback
    traceback.print_exc()

Uses

• Experimentation with small models

Direct Use

• Try to get the model to give a coherent code

Out-of-Scope Use

This model won't work for any malicious goal. It's too dumb.

Bias, Risks, and Limitations

17m parameters is not a lot, and it limits a lot it's usage

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases, and limitations of the model

How to Get Started with the Model

Use the code in the repository of this model to get started

Training Details

Training Data

This model was trained using the theblackcat102/evol-codealpaca-v1 dataset

Training Procedure

Training Hyperparameters

• Training regime: fp32

Technical Specifications

Hardware

• Any GPU with 8 Gb of memory should be able to run it, CPU doesn't run

Software

• PyTorch