Configuration Parsing Warning:In adapter_config.json: "peft.base_model_name_or_path" must be a string
dhruvanmurthy/Qwen3-8B-tool-use-sft
LoRA adapter for Qwen/Qwen3-8B fine-tuned for tool-use via SFT (supervised fine-tuning).
Model Details
- Base model:
Qwen/Qwen3-8B - Training stage:
SFT - LoRA rank: 64
- Task: Multi-tool selection and argument generation
- Trained with: Tinker remote GPU training
Evaluation Results
| Metric | Score |
|---|---|
| Tool Selection Accuracy | 77.4% |
| Argument Accuracy | 31.0% |
| Schema Compliance | 81.2% |
| Multi-Step Success | 38.6% |
| Avg Latency | 7125 ms |
Usage
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-8B")
model = PeftModel.from_pretrained(base, "dhruvanmurthy/Qwen3-8B-tool-use-sft")
License
MIT
- Downloads last month
- 1
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support