Instructions to use darrellbest/Hemmingway-1-VL with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use darrellbest/Hemmingway-1-VL with Transformers:

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

pipe = pipeline("image-text-to-text", model="darrellbest/Hemmingway-1-VL")
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("darrellbest/Hemmingway-1-VL")
model = AutoModelForMultimodalLM.from_pretrained("darrellbest/Hemmingway-1-VL", 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 darrellbest/Hemmingway-1-VL with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "darrellbest/Hemmingway-1-VL"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "darrellbest/Hemmingway-1-VL",
		"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/darrellbest/Hemmingway-1-VL

SGLang

How to use darrellbest/Hemmingway-1-VL 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 "darrellbest/Hemmingway-1-VL" \
    --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": "darrellbest/Hemmingway-1-VL",
		"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 "darrellbest/Hemmingway-1-VL" \
        --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": "darrellbest/Hemmingway-1-VL",
		"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 darrellbest/Hemmingway-1-VL with Docker Model Runner:
```
docker model run hf.co/darrellbest/Hemmingway-1-VL
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Hemmingway-1 VL

Altworld/Hemmingway-1 (a Qwen3.8-27B fine-tune for everyday writing), converted from its text-only Qwen3_5ForCausalLM layout back into the stock Qwen/Qwen3.8-27B vision-language layout (Qwen3_5ForConditionalGeneration), with Qwen3.8-27B's own vision tower restored. Every language-model weight is Hemmingway-1's, unchanged. The layout is the one the Qwen3.8 tooling ecosystem expects (vLLM, llama.cpp mmproj, Heretic, quantizers), and you can give it images again.

All credit for the model goes to Altworld. This is a format conversion, nothing more.

What changed, and proof that nothing else did

Tensor names: model.layers.* → model.language_model.layers.* (and embed_tokens, norm). lm_head.* and the multi-token-prediction head mtp.* are kept as-is.
Added: the 333 model.visual.* tensors from Qwen/Qwen3.8-27B, plus its image/video processor configs.
Config: Qwen3.8-27B's own config.json. Hemmingway-1's text config is identical to its text_config, field for field.

Verified:

Check	Result
All 866 Hemmingway-1 tensors (incl. MTP), byte-compared	byte-identical, same dtypes
All 333 vision tensors vs Qwen3.8-27B	byte-identical
Logits, original vs this repo (transformers 5.17, bf16), on 5 prompts incl. a tool call and a tool result	bit-equal (max difference 0)
48-token greedy continuations on the same prompts	identical, incl. the `<tool_call>` output

The chat template (including tool calling) is Qwen3.8's, which Hemmingway-1 already uses unchanged.

Use

from transformers import AutoModelForImageTextToText, AutoTokenizer
tok = AutoTokenizer.from_pretrained("darrellbest/Hemmingway-1-VL")
model = AutoModelForImageTextToText.from_pretrained("darrellbest/Hemmingway-1-VL", dtype="auto", device_map="auto")

vllm serve darrellbest/Hemmingway-1-VL

Other formats, each tested: GGUF (BF16 / Q8_0 / Q6_K / Q4_K_M + mmproj), FP8 and NVFP4 for vLLM. Uncensored: darrellbest/Hemmingway-1-Heretic.

For fine-tuning, either layout works. Altworld's original text-only repo is the simplest for text/tool-call LoRA. Since the language weights are identical, an adapter trained on one applies to the other once its module names get the language_model. prefix.

Licence

Same as Hemmingway-1: CC BY-NC 4.0. Non-commercial use, with credit to Hemmingway-1 / Altworld. Commercial use needs an agreement with Altworld (luka@hemmingway.io). The vision tower comes from Qwen3.8-27B (Apache-2.0).

Downloads last month: 22

Safetensors

Model size

28B params

Tensor type

BF16

Model tree for darrellbest/Hemmingway-1-VL

Base model

Qwen/Qwen3.8-27B

Finetuned

Altworld/Hemmingway-1

Finetuned

(4)

this model

Quantizations

2 models