Instructions to use SulphurAI/Sulphur-2-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SulphurAI/Sulphur-2-base with Diffusers:

pip install -U diffusers transformers accelerate

import torch
from diffusers import DiffusionPipeline

# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("SulphurAI/Sulphur-2-base", dtype=torch.bfloat16, device_map="cuda")

prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]

llama-cpp-python

How to use SulphurAI/Sulphur-2-base with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="SulphurAI/Sulphur-2-base",
	filename="prompt_enhancer/mmproj-BF16.gguf",
)

llm.create_chat_completion(
	messages = "\"A young man walking on the street\""
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use SulphurAI/Sulphur-2-base with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf SulphurAI/Sulphur-2-base:BF16
# Run inference directly in the terminal:
llama cli -hf SulphurAI/Sulphur-2-base:BF16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf SulphurAI/Sulphur-2-base:BF16
# Run inference directly in the terminal:
llama cli -hf SulphurAI/Sulphur-2-base:BF16

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf SulphurAI/Sulphur-2-base:BF16
# Run inference directly in the terminal:
./llama-cli -hf SulphurAI/Sulphur-2-base:BF16

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf SulphurAI/Sulphur-2-base:BF16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf SulphurAI/Sulphur-2-base:BF16

Use Docker

docker model run hf.co/SulphurAI/Sulphur-2-base:BF16

LM Studio
Jan
Ollama
How to use SulphurAI/Sulphur-2-base with Ollama:
```
ollama run hf.co/SulphurAI/Sulphur-2-base:BF16
```

Unsloth Studio

How to use SulphurAI/Sulphur-2-base with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for SulphurAI/Sulphur-2-base to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for SulphurAI/Sulphur-2-base to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for SulphurAI/Sulphur-2-base to start chatting

How to use SulphurAI/Sulphur-2-base with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf SulphurAI/Sulphur-2-base:BF16

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "SulphurAI/Sulphur-2-base:BF16"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use SulphurAI/Sulphur-2-base with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf SulphurAI/Sulphur-2-base:BF16

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default SulphurAI/Sulphur-2-base:BF16

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use SulphurAI/Sulphur-2-base with Docker Model Runner:
```
docker model run hf.co/SulphurAI/Sulphur-2-base:BF16
```

Lemonade

How to use SulphurAI/Sulphur-2-base with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull SulphurAI/Sulphur-2-base:BF16

Run and chat with the model

lemonade run user.Sulphur-2-base-BF16

List all available models

lemonade list

Is Sulphur-2 just slower than LTX 2.3?

#50

by Zenshuuin - opened about 22 hours ago

Discussion

Zenshuuin

about 22 hours ago

•

edited about 9 hours ago

So I've been using RuneXX's LTX 2.3 workflows (https://huggingface.co/RuneXX/LTX-2.3-Workflows). I like that these cover a number of different use cases besides just T2V I2V. There's also FLF2V, and video extension workflows. I was hoping since its based on the LTX 2.3 pipeline, I would be able to just plug Sulphur-2 into these workflows.

I tired downloading sulphur_dev_fp8mixed.safetensors and ltx-2.3-22b-distilled-lora-1.1_fro90_ceil72_condsafe.safetensors and using them in the RuneXX workflows.

I was able to generate a decent video clip, but it was MUCH slower than with the LTX 2.3 models

A clip that on ltx-2.3-22b-distilled-1.1_transformer_only_fp8_scaled.safetensors took only 5-10 minutes, took over an hour and a half using the sulphur-2 models.

So, I'm wondering if I'm missing something. Is there some tweak I can make to get the workflow running faster, or is this model just a lot slower than the ltx 2.3 model I was using?

bmgjet

about 18 hours ago

I just use the comfyui template for ltx 2.3 for t2v, i2v, flf2v.
Then switch the model and lora to the sulfur ones.
Windows 10, RTX 5090 with 64GB DDR5 System Memory

15sec duration same prompt and seed.
t2v sulfur slower at 173.48 vs 161.09
i2v sulfur faster at 162.13 vs 167.22
flf2v sulfur slower 174.14 vs 168.41

Id call that with in the margin of error.
That big a difference you are getting is like the difference between fitting in system ram and having to use the page file.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment