Instructions to use willamazon1/sdft-tau-lora-iter160 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use willamazon1/sdft-tau-lora-iter160 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="willamazon1/sdft-tau-lora-iter160")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("willamazon1/sdft-tau-lora-iter160")
model = AutoModelForCausalLM.from_pretrained("willamazon1/sdft-tau-lora-iter160", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use willamazon1/sdft-tau-lora-iter160 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "willamazon1/sdft-tau-lora-iter160"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "willamazon1/sdft-tau-lora-iter160",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/willamazon1/sdft-tau-lora-iter160

SGLang

How to use willamazon1/sdft-tau-lora-iter160 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "willamazon1/sdft-tau-lora-iter160" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "willamazon1/sdft-tau-lora-iter160",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "willamazon1/sdft-tau-lora-iter160" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "willamazon1/sdft-tau-lora-iter160",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use willamazon1/sdft-tau-lora-iter160 with Docker Model Runner:
```
docker model run hf.co/willamazon1/sdft-tau-lora-iter160
```

sdft-tau-lora-iter160

Qwen3-8B, multi-stage SFT (SDFT) checkpoint with a tau-bench RL LoRA adapter merged in. Full weights, ready to load with transformers / SGLang / vLLM — no PEFT needed.

Lineage

Stage	What
Base	`Qwen/Qwen3-8B-Base`
SDFT	multi-stage SFT chain: Math → Sea → Search → TauSFT → Tau-IF
RL	GSPO on tau-bench `retail` (train split), LoRA-only (base frozen), iteration 160

RL / LoRA configuration

adapter: rank 16, alpha 32 (scaling = alpha/r = 2.0), dropout 0.0
targets: linear_qkv, linear_proj, linear_fc1, linear_fc2 on all 36 layers (144 modules, 288 tensors)
advantage estimator GSPO, KL loss 0.01 (low_var_kl), eps-clip 0.2/0.25
Adam, lr 1e-6 constant, weight decay 0.01
TIS off (its sequence-level rejection veto zeroes essentially every LoRA sequence)
user simulator: GLM-4.7-Flash served in-cluster (matches tau-bench's LLM-user setup)

How it was exported

The adapter was merged in Megatron parameter space (W += 2.0 · B·A per LoRA'd module, into the frozen base weights carried by the same iter_0000160 torch_dist checkpoint), then converted to HuggingFace safetensors. Merging before conversion avoids having to un-fuse the GQA-interleaved QKV and the gate/up split by hand. LayerNorm weights are untouched — the LoRA delta applies to the post-LN matmuls only.

Verification performed on the export:

adapter health gate: 144/144 modules have non-zero lora_B (max |lora_B| = 9.82e-05), so the adapter is genuinely trained rather than sitting at its zero init
relative delta size ‖ΔW‖/‖W‖: min 3.76e-05, median 5.01e-05, max 1.11e-04
0 non-finite tensors across all 4 shards; 399 tensors, 15.26 GiB, index complete

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "willamazon1/sdft-tau-lora-iter160"
tok = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, dtype="bfloat16", device_map="auto")

Note this is derived from a base (non-instruct) Qwen3 checkpoint plus SFT/RL stages; use the same prompt format as the tau-bench agent it was trained with rather than assuming a generic chat template.

Downloads last month: 6

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for willamazon1/sdft-tau-lora-iter160

Base model

Qwen/Qwen3-8B-Base

Adapter

(89)

this model