Instructions to use vidfom/Ltx-3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use vidfom/Ltx-3 with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="vidfom/Ltx-3", filename="ComfyUI/models/text_encoders/gemma-3-12b-it-qat-UD-Q4_K_XL.gguf", )
llm.create_chat_completion( messages = "No input example has been defined for this model task." )
- Notebooks
- Google Colab
- Kaggle
- Local Apps
- llama.cpp
How to use vidfom/Ltx-3 with llama.cpp:
Install from brew
brew install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf vidfom/Ltx-3:UD-Q4_K_XL # Run inference directly in the terminal: llama-cli -hf vidfom/Ltx-3:UD-Q4_K_XL
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf vidfom/Ltx-3:UD-Q4_K_XL # Run inference directly in the terminal: llama-cli -hf vidfom/Ltx-3:UD-Q4_K_XL
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 vidfom/Ltx-3:UD-Q4_K_XL # Run inference directly in the terminal: ./llama-cli -hf vidfom/Ltx-3:UD-Q4_K_XL
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 vidfom/Ltx-3:UD-Q4_K_XL # Run inference directly in the terminal: ./build/bin/llama-cli -hf vidfom/Ltx-3:UD-Q4_K_XL
Use Docker
docker model run hf.co/vidfom/Ltx-3:UD-Q4_K_XL
- LM Studio
- Jan
- Ollama
How to use vidfom/Ltx-3 with Ollama:
ollama run hf.co/vidfom/Ltx-3:UD-Q4_K_XL
- Unsloth Studio new
How to use vidfom/Ltx-3 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 vidfom/Ltx-3 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 vidfom/Ltx-3 to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for vidfom/Ltx-3 to start chatting
- Docker Model Runner
How to use vidfom/Ltx-3 with Docker Model Runner:
docker model run hf.co/vidfom/Ltx-3:UD-Q4_K_XL
- Lemonade
How to use vidfom/Ltx-3 with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull vidfom/Ltx-3:UD-Q4_K_XL
Run and chat with the model
lemonade run user.Ltx-3-UD-Q4_K_XL
List all available models
lemonade list
| import {app} from "../../../scripts/app.js"; | |
| import {api} from "../../../scripts/api.js"; | |
| let watched_nodes = {} | |
| let resolve = undefined | |
| let testURL = api.apiURL("/VHS_test") | |
| let errors = [] | |
| api.addEventListener("executed", async function ({detail}) { | |
| if (watched_nodes && watched_nodes[detail?.node]) { | |
| if (detail?.output?.unfinished_batch) { | |
| return | |
| } | |
| let requestBody = {tests: watched_nodes[detail.node], output: detail.output} | |
| try { | |
| let req = await fetch(api.apiURL("/VHS_test"), | |
| {method: "POST", body: JSON.stringify(requestBody)}); | |
| let testResult = await req.json() | |
| if (testResult.length != 0) { | |
| errors.push(testResult) | |
| } | |
| } catch(e) { | |
| errors.push(e) | |
| } | |
| if (!(watched_nodes.length -= 1)) { | |
| resolve() | |
| } | |
| } | |
| }); | |
| const workflowService = app.extensionManager.workflow | |
| async function runTest(file) { | |
| if (!file?.name?.endsWith(".json")) { | |
| return false | |
| } | |
| let workflow = JSON.parse(await file.text()) | |
| await app.loadGraphData(workflow) | |
| //NOTE: API is not used so workflow data is actually processed | |
| watched_nodes = workflow.tests | |
| errors = [] | |
| let p = new Promise((r) => resolve = r) | |
| await app.queuePrompt() | |
| //block until execution completes | |
| await p | |
| watched_nodes = {} | |
| if (errors.length > 0) { | |
| app.ui.dialog.show("Failed " + errors.length + " tests:\n" + errors) | |
| return true | |
| } | |
| await workflowService.closeWorkflow(workflowService.activeWorkflow, {warnIfUnsaved: false}) | |
| return false | |
| } | |
| let iconOverride = document.createElement("style") | |
| iconOverride.innerHTML = `.VHSTestIcon:before {content: '🧪';}` | |
| document.body.append(iconOverride) | |
| let testSidebar = {id: 'VHStest', title: 'VHS Test', icon: 'VHSTestIcon', type: 'custom', | |
| render: (e) => { | |
| e.innerHTML = `Select a folder containing tests | |
| <input> | |
| Or select a single test | |
| <input> | |
| ` | |
| const folderInput = e.children[0] | |
| const fileInput = e.children[1] | |
| Object.assign(folderInput, { | |
| type: "file", | |
| webkitdirectory: true, | |
| onchange: async function() { | |
| const startTime = Date.now() | |
| let failedTests = false | |
| for(const file of this.files) { | |
| failedTests ||= await runTest(file) | |
| } | |
| this.value="" | |
| if (!failedTests) { | |
| console.log("All tests passed in " + ((Date.now() - startTime)/1000) + "s") | |
| } | |
| }, | |
| }); | |
| Object.assign(fileInput, { | |
| type: "file", | |
| accept: ".json", | |
| onchange: async function() { | |
| if (this.files.length) { | |
| if(!(await runTest(this.files[0]))) { | |
| console.log("Test complete") | |
| } | |
| this.value="" | |
| } | |
| }, | |
| }); | |
| }} | |
| app.extensionManager.registerSidebarTab(testSidebar) | |