Instructions to use ewald1976/Corridor-G-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ewald1976/Corridor-G-12B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="ewald1976/Corridor-G-12B") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("ewald1976/Corridor-G-12B") model = AutoModelForCausalLM.from_pretrained("ewald1976/Corridor-G-12B") 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
- vLLM
How to use ewald1976/Corridor-G-12B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "ewald1976/Corridor-G-12B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "ewald1976/Corridor-G-12B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/ewald1976/Corridor-G-12B
- SGLang
How to use ewald1976/Corridor-G-12B 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 "ewald1976/Corridor-G-12B" \ --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": "ewald1976/Corridor-G-12B", "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 "ewald1976/Corridor-G-12B" \ --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": "ewald1976/Corridor-G-12B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use ewald1976/Corridor-G-12B with Docker Model Runner:
docker model run hf.co/ewald1976/Corridor-G-12B
Corridor-G-12B
Corridor-G-12B is a creative writing fine-tune of Mistral Nemo 12B focused on procedural science fiction, institutional realism, and the quiet lives of ordinary people in extraordinary places.
The goal was not to create heroes.
The goal was to create places.
What it writes
Corridor-G excels at stories where systems, routines, and environments shape daily life.
Maintenance workers.
Night-shift guards.
Transit operators.
Clerks.
Technicians.
Visitors trying to find a hotel pool.
The station exists before the scene begins and continues after it ends.
Strange things may happen, but they are rarely explained. The universe is vast, indifferent, and often busy with other concerns.
Style
Corridor-G favors:
- restrained prose
- observational narration
- practical dialogue
- institutional procedures
- small human interactions
- unresolved mysteries
- continuity of place
It tends to avoid dramatic revelations, heroic speeches, and artificial urgency.
A damaged airlock seal may be more important than an invasion.
A conversation at a reception desk may matter more than a battle.
Training
Corridor-G was trained on a curated mixture of synthetic procedural science-fiction samples and literary material.
Additional training data:
nbeerbower/gutenberg-moderne-dpo
The objective was to retain the procedural and institutional character of earlier Corridor models while introducing more natural dialogue, continuity, and human presence.
Influences
Corridor-G draws inspiration from traditions of science fiction where environments, systems, and people are given equal weight.
The result is not space opera.
It is infrastructure fiction.
Not the captain's log.
The maintenance report.
Not the war.
The station during the war.
Not the chosen hero.
The person cleaning the observation deck.
Typical Corridor-G Scene
A visitor asks for directions.
The requested facility is closed.
An alternative is suggested.
The visitor checks their budget.
The corridor remains open.
The story continues elsewhere.
What this model is not
Corridor-G is not designed for:
- high-drama roleplay
- power fantasies
- constant escalation
- cinematic action scenes
- heroic protagonists
If you are looking for galaxy-ending threats, Corridor-G may disappoint you.
If you are interested in how people live inside the galaxy before the threat arrives, it may be exactly what you want.
Example
"You are enjoying the entertainment too much," he said to her. "You have him in the second stage already. That is not permitted in this establishment."
The woman's expression did not change.
"He was already in the second stage when he sat down," she said. "I merely continued what was begun."
Voss considered this. It was a plausible claim; he had no reason to disbelieve it.
"Then I must speak to your employer about the condition of the patrons who come to you," he said. "Come with me."
He turned and walked toward the door at the back of the bar that led to the office where the establishment's manager conducted his business. The Orion woman followed, after a moment's hesitation, and the Andorian sat very still at the table, watching them go with an expression that was no longer entirely happy.
Notes
Corridor-G is intended for creative writing, worldbuilding, slice-of-life science fiction, procedural narratives, and roleplay scenarios that benefit from grounded environments and believable institutional behavior.
The model performs best when allowed to observe rather than explain, and when stories are permitted to end without complete resolution.
- Downloads last month
- -
