Instructions to use shuaishuaicdp/Code-X-SFT-27B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use shuaishuaicdp/Code-X-SFT-27B with Transformers:

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

pipe = pipeline("image-text-to-text", model="shuaishuaicdp/Code-X-SFT-27B")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("shuaishuaicdp/Code-X-SFT-27B")
model = AutoModelForMultimodalLM.from_pretrained("shuaishuaicdp/Code-X-SFT-27B", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use shuaishuaicdp/Code-X-SFT-27B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "shuaishuaicdp/Code-X-SFT-27B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "shuaishuaicdp/Code-X-SFT-27B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/shuaishuaicdp/Code-X-SFT-27B

SGLang

How to use shuaishuaicdp/Code-X-SFT-27B 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 "shuaishuaicdp/Code-X-SFT-27B" \
    --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": "shuaishuaicdp/Code-X-SFT-27B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "shuaishuaicdp/Code-X-SFT-27B" \
        --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": "shuaishuaicdp/Code-X-SFT-27B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use shuaishuaicdp/Code-X-SFT-27B with Docker Model Runner:
```
docker model run hf.co/shuaishuaicdp/Code-X-SFT-27B
```

Code-X-SFT-27B

Code-X-SFT-27B is a research, non-commercial supervised fine-tune of Qwen/Qwen3.6-27B for coding-agent workflows. This repository contains the consolidated full checkpoint rather than a standalone adapter.

Checkpoint summary

Architecture: Qwen3_5ForConditionalGeneration
Parameters: 27,356,728,560
Precision: BF16
Format: safetensors, 12 shards
Tensor entries: 1,184
Tensor payload size: 54,713,457,120 bytes
Context length used for post-training: 32,000 tokens

The checkpoint retains the image-text architecture of the base model. The post-training data focused on coding-agent interactions; multimodal quality has not been independently validated for this release.

Training summary

The model was trained for two epochs on 2,553 synthetic coding-agent SFT records. Post-training used a rank-64 LoRA with alpha 128 over linear layers, followed by consolidation with the base checkpoint. The recorded final training loss was 0.2276 at step 290. These values are included for reproducibility and do not imply downstream benchmark performance.

The synthetic dataset and reproducible training recipe are released separately from the model weights.

Usage

Use a recent Transformers release with support for the Qwen3.5/Qwen3.6 model family. Replace the placeholder with either this Hub repository ID or a local checkpoint path.

from transformers import AutoProcessor, Qwen3_5ForConditionalGeneration

model_id = "shuaishuaicdp/Code-X-SFT-27B"

processor = AutoProcessor.from_pretrained(model_id)
model = Qwen3_5ForConditionalGeneration.from_pretrained(
    model_id,
    dtype="auto",
    device_map="auto",
)

Follow the upstream Qwen model card for message formatting, multimodal input construction, and generation examples.

Validation performed before release

Verified that every tensor in the weight index resolves to exactly one of the 12 included shards.
Parsed every safetensors header without loading tensor payloads.
Matched tensor shapes, byte ranges, parameter count, and payload size to the index metadata.
Loaded the configuration, tokenizer, and processor offline from the staged directory.
Scanned public metadata for credentials, private filesystem paths, and local training-run identifiers.

No full benchmark table is claimed in this model card until the corresponding evaluation artifacts are packaged and independently reproducible.

Intended use and limitations

This checkpoint is intended for non-commercial research on coding agents, tool use, synthetic-data studies, and reproducible evaluation. It may produce incorrect, insecure, or destructive code. Run agents inside an isolated environment, review generated commands, and do not expose secrets or production credentials. Users are responsible for evaluating suitability, safety, and legal compliance for their own use case.

License and attribution

Release-specific modifications are provided under the Creative Commons Attribution-NonCommercial 4.0 International license in LICENSE. The checkpoint incorporates material from Qwen/Qwen3.6-27B, which is distributed under the Apache License 2.0; its license is retained in LICENSE-QWEN-APACHE-2.0. See NOTICE.md for attribution and a description of the changes. Users must comply with all applicable terms.

Downloads last month: 40

Safetensors

Model size

27B params

Tensor type

BF16

Model tree for shuaishuaicdp/Code-X-SFT-27B

Base model

Qwen/Qwen3.6-27B

Finetuned

(333)

this model