BasitAliii/TechQA-TinyLlama-FT

TechQA

TechQA is a fine-tuned LoRA model based on TinyLlama-1.1B-Chat-v1.0. It has been trained on a synthetic instructional Q&A dataset covering technical topics including:

• Machine Learning
• Data Structures
• Algorithms
• Python Programming
• Web Development
• Databases

The model was fine-tuned using the TechQA dataset, containing 3,000 examples in a concise question-answer format. It is suitable for generating instructional, technical, and educational responses.

---

Model Details

• Base model: TinyLlama-1.1B-Chat-v1.0
• Fine-tuning method: LoRA (Low-Rank Adaptation)
• Number of trainable parameters: ~1.1 million
• Dataset size: 3,000 examples
• Task: Instruction-following, Q&A in technical domains

---

Usage

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

# Load base model
base_model_id = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
lora_path = "YourUsername/TechQA"  # Hugging Face repo

tokenizer = AutoTokenizer.from_pretrained(base_model_id)
tokenizer.pad_token = tokenizer.eos_token

base_model = AutoModelForCausalLM.from_pretrained(
    base_model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Load LoRA weights
model = PeftModel.from_pretrained(base_model, lora_path)
model.eval()

# Example generation
instruction = "Explain regression in machine learning."
prompt = f"### Instruction:\n{instruction}\n\n### Response:\n"

inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=150)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))