Instructions to use cpral/Hermes-4-405B-exl3-2bpw with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use cpral/Hermes-4-405B-exl3-2bpw with Transformers:

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

pipe = pipeline("text-generation", model="cpral/Hermes-4-405B-exl3-2bpw")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("cpral/Hermes-4-405B-exl3-2bpw")
model = AutoModelForCausalLM.from_pretrained("cpral/Hermes-4-405B-exl3-2bpw")
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 cpral/Hermes-4-405B-exl3-2bpw with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "cpral/Hermes-4-405B-exl3-2bpw"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "cpral/Hermes-4-405B-exl3-2bpw",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/cpral/Hermes-4-405B-exl3-2bpw

SGLang

How to use cpral/Hermes-4-405B-exl3-2bpw 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 "cpral/Hermes-4-405B-exl3-2bpw" \
    --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": "cpral/Hermes-4-405B-exl3-2bpw",
		"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 "cpral/Hermes-4-405B-exl3-2bpw" \
        --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": "cpral/Hermes-4-405B-exl3-2bpw",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use cpral/Hermes-4-405B-exl3-2bpw with Docker Model Runner:
```
docker model run hf.co/cpral/Hermes-4-405B-exl3-2bpw
```

Hermes-4-405B-exl3-2bpw / README.md

cpral

Update README.md

b89fe27 verified 2 months ago

1.85 kB

language:
  - en
license: llama3
tags:
  - Llama-3.1
  - instruct
  - finetune
  - reasoning
  - hybrid-mode
  - chatml
  - function calling
  - tool use
  - json mode
  - structured outputs
  - atropos
  - dataforge
  - long context
  - roleplaying
  - chat
base_model:
  - NousResearch/Hermes-4-405B
library_name: transformers
widget:
  - example_title: Hermes 4
    messages:
      - role: system
        content: >-
          You are Hermes 4, a capable, neutrally-aligned assistant. Prefer
          concise, correct answers.
      - role: user
        content: Explain what Hadamard Transform is.
model-index:
  - name: Hermes-4-Llama-3.1-405B
    results: []

Hermes 4 — Llama-3.1 405B EXL 3 2.00bpw

2.00 BPW H8 exllamav3 quant of Hermes 4 405B.

 -- A perplexity:  1.50484401
 -- B perplexity:  4.46562014
 -- A label in top-K:
      K = 1: 0.8938
      K = 2: 0.9486
      K = 3: 0.9640
      K = 4: 0.9714
      K = 5: 0.9757
 -- B label in top-K:
      K = 1: 0.6383
      K = 2: 0.7622
      K = 3: 0.8163
      K = 4: 0.8482
      K = 5: 0.8698
 -- Top-K agreement, A vs B:
      K = 1: 0.6743
      K = 2: 0.2721
      K = 3: 0.0833
      K = 4: 0.0222
      K = 5: 0.0056
 -- KL divergence (A, B):  2.27405149
 -- KL divergence (B, A):  1.05870732

command used to generate this quant

ulimit -n 100000
PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True python convert.py -i /home/ubuntu/workspace/models/Hermes-4-405B \
                  -o /home/ubuntu/workspace/models/final/hermes4-405b-2bpw \
                  -w /home/ubuntu/workspace/models/workdir \
                  -b 2.0 \
                  -hq \
                  -ss 2048 \
                  -cpi 3600 \
                  -hb 8 \
                  -d 0