Instructions to use modrill/qwen3-4b-nothink-baseline-full-sft with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use modrill/qwen3-4b-nothink-baseline-full-sft with Transformers:

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

pipe = pipeline("text-generation", model="modrill/qwen3-4b-nothink-baseline-full-sft")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("modrill/qwen3-4b-nothink-baseline-full-sft")
model = AutoModelForMultimodalLM.from_pretrained("modrill/qwen3-4b-nothink-baseline-full-sft")
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 modrill/qwen3-4b-nothink-baseline-full-sft with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "modrill/qwen3-4b-nothink-baseline-full-sft"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "modrill/qwen3-4b-nothink-baseline-full-sft",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/modrill/qwen3-4b-nothink-baseline-full-sft

SGLang

How to use modrill/qwen3-4b-nothink-baseline-full-sft 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 "modrill/qwen3-4b-nothink-baseline-full-sft" \
    --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": "modrill/qwen3-4b-nothink-baseline-full-sft",
		"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 "modrill/qwen3-4b-nothink-baseline-full-sft" \
        --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": "modrill/qwen3-4b-nothink-baseline-full-sft",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use modrill/qwen3-4b-nothink-baseline-full-sft with Docker Model Runner:
```
docker model run hf.co/modrill/qwen3-4b-nothink-baseline-full-sft
```

modrill/qwen3-4b-nothink-baseline-full-sft

Full-parameter supervised fine-tuning (SFT) of Qwen/Qwen3-4B-Base on the nothink_mix dataset. This checkpoint is a full SFT model: weights are stored in a single model.safetensors file (~8.2 GB). This is not a LoRA adapter.

Related models

LoRA SFT (adapter only): modrill/qwen3-4b-nothink-baseline-lora-sft
Think baseline full SFT: modrill/qwen3-4b-think-baseline-full-sft

Training summary

Field	Value
Method	Full SFT (ZeRO-2)
Dataset	nothink_mix
Chat template	qwen3_nothink
Epochs	2
Seed	42
Cutoff length	8192
Packing	false
Per-device batch size	4
Gradient accumulation	4
Effective batch size	64
Learning rate	2e-5
Train loss	0.1898
Train steps	2988
Finished	2026-06-08 07:10 CST

Usage

Load with Hugging Face Transformers:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "modrill/qwen3-4b-nothink-baseline-full-sft"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True)

Training hyperparameters

learning_rate: 2e-05
train_batch_size: 4
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU (4 devices)
gradient_accumulation_steps: 4
total_train_batch_size: 64
optimizer: AdamW (fused)
lr_scheduler: cosine
lr_scheduler_warmup_steps: 0.03
num_epochs: 2.0

Framework versions

Transformers 5.6.0
PyTorch 2.8.0+cu128
Datasets 4.0.0
Tokenizers 0.22.2

Downloads last month: 28

Safetensors

Model size

4B params

Tensor type

BF16

Model tree for modrill/qwen3-4b-nothink-baseline-full-sft

Base model

Qwen/Qwen3-4B-Base

Finetuned

(323)

this model