ananya-Qwen2.5-finetuned
Qwen2.5-1.5B-Instruct fine-tuned with LoRA (merged) on Q&A pairs generated from a personal profile (college training exercise). Ask "Who is Ananya?" for the full profile, or a single field such as "What is Ananya's role?". Works with or without a system prompt.
Training
| Setting | Value |
|---|---|
| Base model | Qwen/Qwen2.5-1.5B-Instruct |
| Method | LoRA (r=16, alpha=32, dropout=0.05), all attention + MLP projections |
| Data | 616 Q&A pairs (synonym wordings, with and without system prompt) |
| Epochs / LR | 5 / 2e-4, cosine schedule |
| Precision / GPU | fp16 / Google Colab T4 |
| Loss masking | Prompt tokens masked; only answers trained |
Usage
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
repo = "Ananya2029/ananya-Qwen2.5-finetuned"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, torch_dtype=torch.float16, device_map="auto")
msgs = [{"role": "user", "content": "Who is Ananya?"}]
inputs = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt", return_dict=True).to(model.device)
out = model.generate(**inputs, max_new_tokens=300, do_sample=False)
print(tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
Limitations
Trained only on one personal profile. It is a fine-tuning exercise, not a general-purpose assistant.