Instructions to use reaperdoesntknow/Qwen3.5-4B-Coder with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use reaperdoesntknow/Qwen3.5-4B-Coder with Transformers:

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

pipe = pipeline("image-text-to-text", model="reaperdoesntknow/Qwen3.5-4B-Coder")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("reaperdoesntknow/Qwen3.5-4B-Coder")
model = AutoModelForMultimodalLM.from_pretrained("reaperdoesntknow/Qwen3.5-4B-Coder", 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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use reaperdoesntknow/Qwen3.5-4B-Coder with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "reaperdoesntknow/Qwen3.5-4B-Coder"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "reaperdoesntknow/Qwen3.5-4B-Coder",
		"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/reaperdoesntknow/Qwen3.5-4B-Coder

SGLang

How to use reaperdoesntknow/Qwen3.5-4B-Coder 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 "reaperdoesntknow/Qwen3.5-4B-Coder" \
    --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": "reaperdoesntknow/Qwen3.5-4B-Coder",
		"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 "reaperdoesntknow/Qwen3.5-4B-Coder" \
        --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": "reaperdoesntknow/Qwen3.5-4B-Coder",
		"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"
						}
					}
				]
			}
		]
	}'

Unsloth Desktop
Docker Model Runner
How to use reaperdoesntknow/Qwen3.5-4B-Coder with Docker Model Runner:
```
docker model run hf.co/reaperdoesntknow/Qwen3.5-4B-Coder
```

reaperdoesntknow commited on 2 days ago

Commit

0c820a5

verified ·

1 Parent(s): 2fe0455

Document model files, usage, and evaluation limits

Browse files

Files changed (1) hide show

README.md +40 -7

README.md CHANGED Viewed

@@ -10,13 +10,46 @@ language:
 - en
 ---
-# Uploaded finetuned 16-bit (FP16) model
-- **Developed by:** reaperdoesntknow
-- **License:** apache-2.0
-- **Finetuned from model :** unsloth/Qwen3.5-4B
-This qwen3_5 model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
-[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
-<!-- cix-keeper-ts:2026-11-12T14:16:39Z -->

 - en
 ---
+# Qwen3.5-4B-Coder
+A fine-tuned Qwen3.5 4B checkpoint published by **reaperdoesntknow**, based on [unsloth/Qwen3.5-4B](https://huggingface.co/unsloth/Qwen3.5-4B), and distributed as sharded Safetensors weights for Transformers.
+The repository name identifies this as a coding-oriented release. The published repository does not include a training dataset reference, training recipe, or benchmark report; the name alone should not be interpreted as evidence of improved coding performance.
+## Model details
+- **Developer:** reaperdoesntknow
+- **Base model recorded in the original card:** unsloth/Qwen3.5-4B
+- **License recorded in repository metadata:** Apache-2.0
+- **Language recorded in repository metadata:** English
+- **Architecture in config.json:** Qwen3_5ForConditionalGeneration, with text and vision configuration
+- **Files:** two Safetensors shards, a weight index, tokenizer, processor configuration, and chat template
+- **Configured text dtype:** bfloat16. This is a configuration value, not a claim that every stored tensor has that dtype. The original export template labeled the model FP16; no uniform weight-dtype assertion is made here.
+## Usage
+Use a Transformers version that supports the Qwen3.5 conditional-generation architecture. The saved configuration records Transformers 5.3.0; that is provenance, not an independently tested minimum version.
+Use the tokenizer/processor and chat template included in this repository rather than substituting an unrelated model's prompt format. Follow the [upstream model's usage documentation](https://huggingface.co/unsloth/Qwen3.5-4B) with the model identifier set to:
+`reaperdoesntknow/Qwen3.5-4B-Coder`
+This card does not claim that inference, multimodal behavior, or a particular hardware configuration has been tested for this export.
+## Training and evaluation
+The original export card identifies Unsloth and Hugging Face TRL as training tools. Dataset composition, preprocessing, hyperparameters, training duration, and checkpoint-selection criteria are not documented in the published files reviewed for this card.
+No reproducible coding benchmark, baseline comparison, or safety evaluation is supplied. The original template's generic training-speed statement is not treated as a measured result for this release.
+## Intended use and limitations
+Suitable for local experimentation and evaluation of coding assistance. Review and test generated code before use: outputs may be incorrect, insecure, or incompatible with your dependencies. Do not run generated commands or code with access to secrets or production systems without review.
+The presence of vision configuration does not establish tested image-understanding performance. Evaluate the exact checkpoint, inference stack, and intended workload before deployment.
+## Related repository
+A separately published [qwen3.5-4B-Code-GGUF repository](https://huggingface.co/reaperdoesntknow/qwen3.5-4B-Code-GGUF) provides GGUF files. Its conversion source revision is not recorded in the reviewed card, so an exact conversion relationship is not asserted here.
+<!-- cix-keeper-ts:2026-11-12T14:16:39Z -->