Model Card: Math Homework Solver

Model Details

• Base Model: Llama 3.1 (8 Billion parameters)
• Fine-tuning Method: PEFT (Parameter-Efficient Fine-Tuning) with QLoRA
• Quantization: 4-bit quantization for reduced memory usage
• Training Framework: Unsloth, optimized for efficient fine-tuning of large language models
• Training Environment: Google Colab (free tier), NVIDIA T4 GPU (16GB VRAM), 12GB RAM
• Dataset Used: justsomerandomdude264/ScienceQA-Dataset, 560 selected rows
• Git Repo: The git repo on my github account is justsomerandomdude264/Homework_Solver_LLM

Capabilities

Getting Started

1. Clone the repo
   git clone https://huggingface.co/justsomerandomdude264/Science_Homework_Solver-Llama3.18B
2. Run inference
   1. This method is recommended as its reliable and accurate:
   python

   1from unsloth import FastLanguageModel 2import torch 3 4# Define Your Question 5question = "A tank is filled with water to a height of 12.5 cm. The apparent depth of a needle lying at the bottom of the tank is measured by a microscope to be 9.4 cm. What is the refractive index of water? If water is replaced by a liquid of refractive index 1.63 up to the same height, by what distance would the microscope have to be moved to focus on the needle again?" # Example Question, You can change it with one of your own 6# Load the model 7model, tokenizer = FastLanguageModel.from_pretrained( 8 model_name = "Science_Homework_Solver_Llama318B/model_adapters", # The dir where the repo is cloned or "\\" for root 9 max_seq_length = 2048, 10 dtype = None, 11 load_in_4bit = True, 12 ) 13# Set the model in inference model 14FastLanguageModel.for_inference(model) 15# QA template 16qa_template = """Question: {} 17Answer: {}""" 18# Tokenize inputs 19inputs = tokenizer( 20[ 21 qa_template.format( 22 question, # Question 23 "", # Answer - left blank for generation 24 ) 25], return_tensors = "pt").to("cuda") 26# Stream the answer/output of the model 27from transformers import TextStreamer 28text_streamer = TextStreamer(tokenizer) 29_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens = 512)
   1. Another way to run inference is to use the merged adapters (not recommend as it gives inaccurate/different answers):
   python

   1from transformers import LlamaForCausalLM, AutoTokenizer 2# Load the model 3model = LlamaForCausalLM.from_pretrained( 4 "justsomerandomdude264/Science_Homework_Solver_Llama318B", 5 device_map="auto" 6) 7 8# Load the tokenizer 9tokenizer = AutoTokenizer.from_pretrained("justsomerandomdude264/Science_Homework_Solver_Llama318B") 10# Set the inputs up 11qa_template = """Question: {} 12Answer: {}""" 13 14inputs = tokenizer( 15[ 16 qa_template.format( 17 "A tank is filled with water to a height of 12.5 cm. The apparent depth of a needle lying at the bottom of the tank is measured by a microscope to be 9.4 cm. What is the refractive index of water? If water is replaced by a liquid of refractive index 1.63 up to the same height, by what distance would the microscope have to be moved to focus on the needle again?", # instruction 18 "", # output - leave this blank for generation! 19 ) 20], return_tensors = "pt").to("cuda") 21 22# Do a forward pass 23outputs = model.generate(**inputs, max_new_tokens = 128, use_cache = True) 24raw_output = str(tokenizer.batch_decode(outputs)) 25 26# Formtting the string 27# Removing the list brackets and splitting the string by newline characters 28formatted_string = raw_output.strip("[]").replace("<|begin_of_text|>", "").replace("<|eot_id|>", "").strip("''").split("\\n") 29# Print the lines one by one 30for line in formatted_string: 31 print(line)

Citation

BibTeX Citation

APA Citation