Instructions to use lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1 with Transformers:

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

pipe = pipeline("text-generation", model="lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1")
model = AutoModelForCausalLM.from_pretrained("lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1", 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 lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1

SGLang

How to use lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1 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 "lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1" \
    --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": "lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1",
		"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 "lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1" \
        --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": "lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1 with Docker Model Runner:
```
docker model run hf.co/lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1
```

lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1

Full-parameter RLM RLVR checkpoint.

Base model: Qwen/Qwen3-4B-Instruct-2507
Source checkpoint: /home/ubuntu/prime_run/outputs/rlm-rlvr-qwen3-4b-depth1-llmonly-fullft-lr5e-6-s150-bal35f40v1-ncclfixed-allrollouts/weights/step_150
Step: 150
Prompt variant: sanjaya_text_depth1_llm_only_v1
Runtime: depth-1 LLM-only RLM harness, plain Gemini subcalls, recursive child RLMs disabled.

Downloads last month: 17

Safetensors

Model size

4B params

Tensor type

BF16

Model tree for lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1

Base model

Qwen/Qwen3-4B-Instruct-2507

Finetuned

(2168)

this model

Collection including lsteno/Qwen3-4B-Instruct-2507-RLM-RLVR-FullFT-lr5e-6-depth1-v1

Qwen 3 4B RLM RLVR

Collection

Everything from my experiments on training RLMs. • 14 items • Updated 26 days ago