Not-For-All-Audiences

nsfw

conversational

text-generation-inference

exl2

Model card Files Files and versions

xet

Community

Instructions to use VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

Transformers

How to use VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8 with Transformers:

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

pipe = pipeline("text-generation", model="VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8")
model = AutoModelForCausalLM.from_pretrained("VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8")
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 VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8

SGLang

How to use VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8 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 "VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8" \
    --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": "VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8",
		"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 "VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8" \
        --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": "VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8 with Docker Model Runner:
```
docker model run hf.co/VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8
```

Configuration Parsing Warning:In config.json: "quantization_config.bits" must be an integer

L3.3-70B-Lycosa-v0.2

This is a merge of pre-trained language models created using mergekit.

Merge Details

changes from v0.1:
- Dropped llama-3.3-70b-instruct as a pivot to further reduce positive bias. No noticeable impact on reasoning.
- Added DeepSeek-R1-Distill-Llama-70B as a target model for improved reasoning.

An RP merge with a focus on:
- model intelligence
- removing positive bias
- creativity

This model was merged using the sce merge method using deepseek-ai/DeepSeek-R1-Distill-Llama-70B as a base.

The included DeepSeek-R1-Distill-Llama-70B chat template is recommended.

<｜begin▁of▁sentence｜>system prompt here<｜User｜>user 1st message here<｜Assistant｜>assistant 1st response here<｜end▁of▁sentence｜><｜User｜>user 2nd message here<｜Assistant｜>

The llama3 chat template is no longer recommended due to the increased Deepseek-R1 influence in this v0.2 merge.

Models Merged

The following models were included in the merge:

deepseek-ai/DeepSeek-R1-Distill-Llama-70B
Sao10K/70B-L3.3-Cirrus-x1
TheDrummer/Nautilus-70B-v0.1
Doctor-Shotgun/L3.3-70B-Magnum-v4-SE
SicariusSicariiStuff/Negative_LLAMA_70B

Configuration

The following YAML configuration was used to produce this model:

models:
  # Pivot model
  - model: SicariusSicariiStuff/Negative_LLAMA_70B
  # Target models
  - model: Sao10K/70B-L3.3-Cirrus-x1
  - model: TheDrummer/Nautilus-70B-v0.1
  - model: Doctor-Shotgun/L3.3-70B-Magnum-v4-SE
  - model: deepseek-ai/DeepSeek-R1-Distill-Llama-70B
merge_method: sce
base_model: deepseek-ai/DeepSeek-R1-Distill-Llama-70B
parameters:
  select_topk: 1.0
dtype: bfloat16

Downloads last month: 3

Model tree for VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8

Base model

deepseek-ai/DeepSeek-R1-Distill-Llama-70B

Quantized

(67)

this model

Collection including VoidStare/L3.3-70B-Lycosa-v0.2-EXL2-6.5bpw-h8

EXL2 Quants

Collection

28 items • Updated Aug 29, 2025 • 1