Instructions to use Yobenboben/L3.3-ElectraEXTRA-R1-70b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Yobenboben/L3.3-ElectraEXTRA-R1-70b with Transformers:

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

pipe = pipeline("text-generation", model="Yobenboben/L3.3-ElectraEXTRA-R1-70b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Yobenboben/L3.3-ElectraEXTRA-R1-70b")
model = AutoModelForCausalLM.from_pretrained("Yobenboben/L3.3-ElectraEXTRA-R1-70b")
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 Yobenboben/L3.3-ElectraEXTRA-R1-70b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Yobenboben/L3.3-ElectraEXTRA-R1-70b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Yobenboben/L3.3-ElectraEXTRA-R1-70b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Yobenboben/L3.3-ElectraEXTRA-R1-70b

SGLang

How to use Yobenboben/L3.3-ElectraEXTRA-R1-70b 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 "Yobenboben/L3.3-ElectraEXTRA-R1-70b" \
    --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": "Yobenboben/L3.3-ElectraEXTRA-R1-70b",
		"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 "Yobenboben/L3.3-ElectraEXTRA-R1-70b" \
        --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": "Yobenboben/L3.3-ElectraEXTRA-R1-70b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Yobenboben/L3.3-ElectraEXTRA-R1-70b with Docker Model Runner:
```
docker model run hf.co/Yobenboben/L3.3-ElectraEXTRA-R1-70b
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

ElectraEXTRA

Like Electranova but with a different model, so the thinking works better in it. The writing quality is also better imo.

Settings:

Samplers: With thinking: Temp 1.05, top nsigma 0.7; w/o: Temp 1.15, top nsigma 0.7, minP 0.02, smoothing factor 0.3, smoothing curve 2

Sys. prompt: LeCeption or the one from here

Quants

Static: https://huggingface.co/mradermacher/L3.3-ElectraEXTRA-R1-70b-GGUF

Weighted/imatrix: https://huggingface.co/mradermacher/L3.3-ElectraEXTRA-R1-70b-i1-GGUF

Merge Details

Merge Method

This model was merged using the SCE merge method using Steelskull/L3.3-Electra-R1-70b as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: Sao10K/Llama-3.3-70B-Vulpecula-r1
    parameters:
      select_topk:
        - filter: self_attn
          value: 0.1
        - filter: "q_proj|k_proj|v_proj"
          value: 0.1
        - filter: "up_proj|down_proj"
          value: 0.1
        - filter: mlp
          value: 0.1
        - value: 0.1  # default for other components
  - model: Nohobby/L3.3-Prikol-70B-EXTRA
    parameters:
      select_topk:
        - filter: self_attn
          value: 0.15
        - filter: "q_proj|k_proj|v_proj"
          value: 0.1
        - filter: "up_proj|down_proj"
          value: 0.1
        - filter: mlp
          value: 0.1
        - value: 0.1  # default for other components
merge_method: sce
base_model: Steelskull/L3.3-Electra-R1-70b
dtype: float32
out_dtype: bfloat16
tokenizer:
  source: Steelskull/L3.3-Electra-R1-70b