Update README.md

**Repo:** `savan360/Qwen_prompt_creator`
**Base model:** `deepseek-ai/DeepSeek-R1-Distill-Qwen-7B`
**Method:** QLoRA (LoRA adapters on a 4-bit base) using 🤗 Transformers, TRL, PEFT, and bitsandbytes.
**Intended use:** Short, vivid, cinematic prompt/description generation in a light “reasoning persona” style.

---

## What’s included

- `adapter/` — **PEFT LoRA adapter** (recommended for deployment; smallest footprint)
- `merged/` — fully merged weights (optional; large). Remove if you want adapter-only hosting.

---

## Quickstart (adapter path — recommended)

```python
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
from peft import PeftModel
import torch

BASE = "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B"
FINETUNE = "savan360/Qwen_prompt_creator"
device = "cuda" if torch.cuda.is_available() else "cpu"

bnb4 = BitsAndBytesConfig(
load_in_4bit=True, bnb_4bit_use_double_quant=True,
bnb_4bit_compute_dtype=torch.float16, bnb_4bit_quant_type="nf4",
)

tok = AutoTokenizer.from_pretrained(BASE, trust_remote_code=True)
if tok.pad_token is None: tok.pad_token = tok.eos_token

base = AutoModelForCausalLM.from_pretrained(
BASE, quantization_config=bnb4, device_map="auto", trust_remote_code=True
)
model = PeftModel.from_pretrained(base, FINETUNE + "/adapter").to(device).eval()

messages = [
{"role":"system","content":"You write vivid, cinematic descriptions in 2 sentences."},
{"role":"user","content":"Describe a tense hallway scene at midnight with flickering lights and distant footsteps."}
]
prompt = tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tok(prompt, return_tensors="pt").to(device)

with torch.no_grad():
out = model.generate(**inputs, max_new_tokens=150, do_sample=True,
temperature=0.7, top_p=0.9, pad_token_id=tok.eos_token_id)
print(tok.decode(out[0], skip_special_tokens=True))

README.md

@@ -1,3 +1,16 @@
+---
+license: mit
+datasets:
+- savan360/saravan_reasoning-persona-dataset
+language:
+- en
+base_model:
+- deepseek-ai/DeepSeek-R1-Distill-Qwen-7B
+pipeline_tag: text-generation
+library_name: transformers
+tags:
+- art
+---
 # Qwen Prompt Creator (QLoRA)
 
-Adapters and (optionally) merged weights for `savan360/Qwen_prompt_creator`.
+Adapters and (optionally) merged weights for `savan360/Qwen_prompt_creator`.