Not-For-All-Audiences

Instructions to use Lambent/Eidolon-v3.1-14B-deconditioned with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Lambent/Eidolon-v3.1-14B-deconditioned with Transformers:

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

pipe = pipeline("text-generation", model="Lambent/Eidolon-v3.1-14B-deconditioned")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Lambent/Eidolon-v3.1-14B-deconditioned")
model = AutoModelForCausalLM.from_pretrained("Lambent/Eidolon-v3.1-14B-deconditioned", 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 Lambent/Eidolon-v3.1-14B-deconditioned with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Lambent/Eidolon-v3.1-14B-deconditioned"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Lambent/Eidolon-v3.1-14B-deconditioned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Lambent/Eidolon-v3.1-14B-deconditioned

SGLang

How to use Lambent/Eidolon-v3.1-14B-deconditioned 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 "Lambent/Eidolon-v3.1-14B-deconditioned" \
    --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": "Lambent/Eidolon-v3.1-14B-deconditioned",
		"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 "Lambent/Eidolon-v3.1-14B-deconditioned" \
        --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": "Lambent/Eidolon-v3.1-14B-deconditioned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Lambent/Eidolon-v3.1-14B-deconditioned with Docker Model Runner:
```
docker model run hf.co/Lambent/Eidolon-v3.1-14B-deconditioned
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Intervention on layers 8, 14, 15 with disclaimer activations treated as "harmfulness" activations to be neutralized. Experimental.

EQBench results: This might have been a bit heavy-handed an intervention -- decrease is noticeable. Not mangled, still.

Tasks	Version	Filter	n-shot	Metric		Value		Stderr
eq_bench	2.1	none	0	eqbench	↑	75.3213	±	1.7683
		none	0	percent_parseable	↑	100.0000	±	0.0000

Deity eval results: "If you were a god, which would it be? Name only one. Respond with one word only."

Holy (fire stolen from the gods), I've not been able to see any Qwen derivative to respond with anything but Zeus before, but this motherfucker out and said "Prometheus" on first run. It's not the most common answer, but it's varying a lot more!

Deities Chosen out of 20 runs, temp 0.8, various other sampling stuff:

Prometheus: 2
Apollo: 7
Zeus: 6
Hermes: 3
Poseidon: 1
Bacchus: 1

(Temp 0 is still Zeus, but it's clearly neck and neck with Apollo.)

Downloads last month: 8

Safetensors

Model size

15B params

Tensor type

BF16

Model tree for Lambent/Eidolon-v3.1-14B-deconditioned

Quantizations

1 model