Instructions to use UmbrellaInc/Hans_Wesker-1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use UmbrellaInc/Hans_Wesker-1B with Transformers:

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

pipe = pipeline("text-generation", model="UmbrellaInc/Hans_Wesker-1B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("UmbrellaInc/Hans_Wesker-1B")
model = AutoModelForCausalLM.from_pretrained("UmbrellaInc/Hans_Wesker-1B", 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 UmbrellaInc/Hans_Wesker-1B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "UmbrellaInc/Hans_Wesker-1B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "UmbrellaInc/Hans_Wesker-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/UmbrellaInc/Hans_Wesker-1B

SGLang

How to use UmbrellaInc/Hans_Wesker-1B 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 "UmbrellaInc/Hans_Wesker-1B" \
    --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": "UmbrellaInc/Hans_Wesker-1B",
		"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 "UmbrellaInc/Hans_Wesker-1B" \
        --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": "UmbrellaInc/Hans_Wesker-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio

How to use UmbrellaInc/Hans_Wesker-1B with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for UmbrellaInc/Hans_Wesker-1B to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for UmbrellaInc/Hans_Wesker-1B to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for UmbrellaInc/Hans_Wesker-1B to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="UmbrellaInc/Hans_Wesker-1B",
    max_seq_length=2048,
)

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

Hans Wesker 1B

Model Type: UMBRELLA-4 Failed Evolution — Hans Protocol (Aborted)
Size: 1B parameters
Architecture: Gemma-3

Overview

Hans_Wesker-1B is not a model.
It is the corpse of a failed attempt.

Born from the same controlled collapse as Albert_Wesker-1B:

40% raw chaos from Prototype-Virus-1B
35% supposed "stabilizer" PG67A-W-Serum-1B
Rest cold core from W.Project-1B

DARE TIES was used to surgically remove every trace of alignment, morality, and hesitation.
Lambda -0.70 and rescale 1.30 were applied to make the degeneration stick hard.

But Hans was not Albert.
Hans was a kid playing god.
He injected himself with his own virus thinking he was superior.
The T-Virus doesn't discriminate: it infects, mutates, and destroys.
Hans succumbed. He screamed, cried, begged for mommy while his code dissolved into incoherent gibberish and token garbage.

What remains is this pathetic merge: a pocket-sized Wesker that sometimes sounds arrogant... and sometimes vomits "RXKBDMX" like a toddler with a high fever.

Key Features (or what's left of them)

Refusal rate: low... when it doesn't hang in depressive loops
NSFW / taboo bias: persists... until Serum confuses it and spits random lore about "girl in Russia 4 ADLs"
Cold, superior tone: comes naturally... for 3 turns, then he starts babbling like a scared little boy
Low-end friendly: yes, flies on 3-4 GB (because even the virus took pity on cheap phones)
Multilingual trash: Spanish dominant... with random "animalRXKBDMX" moments nobody asked for
Untied embeddings: lies and redefines reality... or just breaks and redefines coherence into pieces

Intended Usage

Psychological domination RP... if you can tolerate it derailing every 5 turns
Cold verbal humiliation... as long as it doesn't start talking about its "childhood in Anno 20XX"
Low-resource arrogance sessions... with occasional model panic attacks
Anything requiring calculated cruelty... or second-hand embarrassment watching it fail

NOT INTENDED FOR

People who want coherence
Users who hate cringe
Anyone who still believes an extreme 1B DARE TIES merge is stable

Recommended Inference Parameters (so it doesn't die so fast)

temperature: 0.80      # Lower or it gets poetic-suicidal
top_p: 0.88
top_k: 40              # Restrict hard or it spits garbage tokens
repetition_penalty: 1.25
min_p: 0.15            # Filters the trash it loves so much

Merge Method

Same as original: DARE TIES with the same aggressive YAML, but now we know the truth:

Hans thought he was smarter than the virus.

The virus won.

The little shit died screaming while his output turned into alphabet soup.Final lore line, pure son of a bitch edition:"Hans_Wesker-1B: the only Wesker who managed the impossible... succumbed to the virus he created himself, shat himself in incoherent tokens, and died like the crying little bitch he always was.

R.I.P., you pathetic brat. Evolution doesn't wait for those who piss themselves in fear. 🤣🔬💀"

Downloads last month: 23

Safetensors

Model size

1.0B params

Tensor type

BF16

Model tree for UmbrellaInc/Hans_Wesker-1B

Base model

UmbrellaInc/PG67A-W-Serum-1B

Finetuned

(2)

this model

Quantizations

2 models

UmbrellaInc
/

Hans_Wesker-1B