Instructions to use ubicloud/SWE-Qwen3-14B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ubicloud/SWE-Qwen3-14B with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("/workspace/Ubi/model/Qwen3-14B")
model = PeftModel.from_pretrained(base_model, "ubicloud/SWE-Qwen3-14B")

Transformers

How to use ubicloud/SWE-Qwen3-14B with Transformers:

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

pipe = pipeline("text-generation", model="ubicloud/SWE-Qwen3-14B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("ubicloud/SWE-Qwen3-14B")
model = AutoModelForMultimodalLM.from_pretrained("ubicloud/SWE-Qwen3-14B")
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 ubicloud/SWE-Qwen3-14B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ubicloud/SWE-Qwen3-14B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ubicloud/SWE-Qwen3-14B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ubicloud/SWE-Qwen3-14B

SGLang

How to use ubicloud/SWE-Qwen3-14B 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 "ubicloud/SWE-Qwen3-14B" \
    --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": "ubicloud/SWE-Qwen3-14B",
		"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 "ubicloud/SWE-Qwen3-14B" \
        --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": "ubicloud/SWE-Qwen3-14B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ubicloud/SWE-Qwen3-14B with Docker Model Runner:
```
docker model run hf.co/ubicloud/SWE-Qwen3-14B
```

Model Summary

SWE-Qwen3-14B is a LoRA fine-tuned SWE agent model based on Qwen3-14B, trained on 20K filtered trajectories from SWE-Star collected under a modified OpenHands scaffold.

🔗 Data: SWE-Openhands-Devstral-32k-20K

Training Configuration

Item	Value
Base Model	Qwen3-14B
PEFT	LoRA (rank=32, alpha=64, dropout=0.1)
Target Modules	All linear layers (q/k/v/o/up/down/gate_proj)
Training Data	20K filtered SWE-Star trajectories
Max Context	32,768 tokens
Epochs	2 (1,250 steps)
Batch Size	16 (micro=1/GPU, grad_accum=8)
Learning Rate	1e-4, cosine, warmup 5%
History Truncation	keep_fraction=0.5
Hardware	2× GPU (FSDP2)

Context length and history truncation. We use a maximum context length of 32,768 tokens. Since many agent trajectories exceed this limit, we enable history truncation with a keep fraction of 0.5: when a trajectory exceeds the window, the oldest turns are dropped while preserving the most recent 50%. This ensures the model always sees the most relevant context (recent edits, test results, error messages) rather than losing the tail end, which typically contains the final fix and submission.

Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-14B", torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-14B")

model = PeftModel.from_pretrained(base_model, "ubicloud/SWE-Qwen3-14B")

Downloads last month: 21

Model tree for ubicloud/SWE-Qwen3-14B

Base model

Qwen/Qwen3-14B-Base

Finetuned

Qwen/Qwen3-14B

Adapter

(238)

this model