wesjos/SFT-Qwen3-8B-alpaca

Model Card for Qwen3-8B-SFT-Alpaca

🧠 Model Summary

This model is a fine-tuned version of Qwen3-8B-Base, trained using instruction-following SFT (Supervised Fine-Tuning) on an Alpaca-format dataset.
It is designed to improve instruction comprehension, reasoning ability, and response alignment while maintaining fluency and coherence in multiple languages (mainly Chinese and English).

• Base model: Qwen/Qwen3-8B-Base
• Fine-tuning method: Supervised Fine-Tuning (SFT)
• Training format: Alpaca-style instruction–response pairs
• Framework: Transformers + Unsloth
• Languages: Chinese, English
• License: Apache 2.0 (same as Qwen base)

---

🚀 Model Use

✅ Direct Use

You can use this model for:

• Instruction following
• Dialogue and Q&A
• Content generation
• Code explanation or summarization
• Educational and tutoring tasks

⚙️ Example Inference

from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
from unsloth import FastLanguageModel

model_name="wesjos/SFT-Qwen3-8B-alpaca"

tokenizer = AutoTokenizer.from_pretrained(model_name,trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True,
)
alpaca_prompt = """Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
{}

### Response:
"""

inputs = tokenizer(
[
    alpaca_prompt.format(
        "请你介绍一下自己", # instruction
    )
], return_tensors = "pt").to("cuda")

outputs = model.generate(**inputs, max_new_tokens=1024, use_cache=True,temperature=0.9,do_sample=True,top_p=0.95,top_k=20)
print(tokenizer.batch_decode(outputs)[0])