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
File size: 6,035 Bytes
e00eceb | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 | """
Tests for public ComfyAPI and ComfyAPISync functions.
These tests verify that the public API methods work correctly in both sync and async contexts,
ensuring that the sync wrapper generation (via get_type_hints() in async_to_sync.py) correctly
handles string annotations from 'from __future__ import annotations'.
"""
import pytest
import time
import subprocess
import torch
from pytest import fixture
from comfy_execution.graph_utils import GraphBuilder
from tests.execution.test_execution import ComfyClient
@pytest.mark.execution
class TestPublicAPI:
"""Test suite for public ComfyAPI and ComfyAPISync methods."""
@fixture(scope="class", autouse=True)
def _server(self, args_pytest):
"""Start ComfyUI server for testing."""
pargs = [
'python', 'main.py',
'--output-directory', args_pytest["output_dir"],
'--listen', args_pytest["listen"],
'--port', str(args_pytest["port"]),
'--extra-model-paths-config', 'tests/execution/extra_model_paths.yaml',
'--cpu',
]
p = subprocess.Popen(pargs)
yield
p.kill()
torch.cuda.empty_cache()
@fixture(scope="class", autouse=True)
def shared_client(self, args_pytest, _server):
"""Create shared client with connection retry."""
client = ComfyClient()
n_tries = 5
for i in range(n_tries):
time.sleep(4)
try:
client.connect(listen=args_pytest["listen"], port=args_pytest["port"])
break
except ConnectionRefusedError:
if i == n_tries - 1:
raise
yield client
del client
torch.cuda.empty_cache()
@fixture
def client(self, shared_client, request):
"""Set test name for each test."""
shared_client.set_test_name(f"public_api[{request.node.name}]")
yield shared_client
@fixture
def builder(self, request):
"""Create GraphBuilder for each test."""
yield GraphBuilder(prefix=request.node.name)
def test_sync_progress_update_executes(self, client: ComfyClient, builder: GraphBuilder):
"""Test that TestSyncProgressUpdate executes without errors.
This test validates that api_sync.execution.set_progress() works correctly,
which is the primary code path fixed by adding get_type_hints() to async_to_sync.py.
"""
g = builder
image = g.node("StubImage", content="BLACK", height=256, width=256, batch_size=1)
# Use TestSyncProgressUpdate with short sleep
progress_node = g.node("TestSyncProgressUpdate",
value=image.out(0),
sleep_seconds=0.5)
output = g.node("SaveImage", images=progress_node.out(0))
# Execute workflow
result = client.run(g)
# Verify execution
assert result.did_run(progress_node), "Progress node should have executed"
assert result.did_run(output), "Output node should have executed"
# Verify output
images = result.get_images(output)
assert len(images) == 1, "Should have produced 1 image"
def test_async_progress_update_executes(self, client: ComfyClient, builder: GraphBuilder):
"""Test that TestAsyncProgressUpdate executes without errors.
This test validates that await api.execution.set_progress() works correctly
in async contexts.
"""
g = builder
image = g.node("StubImage", content="WHITE", height=256, width=256, batch_size=1)
# Use TestAsyncProgressUpdate with short sleep
progress_node = g.node("TestAsyncProgressUpdate",
value=image.out(0),
sleep_seconds=0.5)
output = g.node("SaveImage", images=progress_node.out(0))
# Execute workflow
result = client.run(g)
# Verify execution
assert result.did_run(progress_node), "Async progress node should have executed"
assert result.did_run(output), "Output node should have executed"
# Verify output
images = result.get_images(output)
assert len(images) == 1, "Should have produced 1 image"
def test_sync_and_async_progress_together(self, client: ComfyClient, builder: GraphBuilder):
"""Test both sync and async progress updates in same workflow.
This test ensures that both ComfyAPISync and ComfyAPI can coexist and work
correctly in the same workflow execution.
"""
g = builder
image1 = g.node("StubImage", content="BLACK", height=256, width=256, batch_size=1)
image2 = g.node("StubImage", content="WHITE", height=256, width=256, batch_size=1)
# Use both types of progress nodes
sync_progress = g.node("TestSyncProgressUpdate",
value=image1.out(0),
sleep_seconds=0.3)
async_progress = g.node("TestAsyncProgressUpdate",
value=image2.out(0),
sleep_seconds=0.3)
# Create outputs
output1 = g.node("SaveImage", images=sync_progress.out(0))
output2 = g.node("SaveImage", images=async_progress.out(0))
# Execute workflow
result = client.run(g)
# Both should execute successfully
assert result.did_run(sync_progress), "Sync progress node should have executed"
assert result.did_run(async_progress), "Async progress node should have executed"
assert result.did_run(output1), "First output node should have executed"
assert result.did_run(output2), "Second output node should have executed"
# Verify outputs
images1 = result.get_images(output1)
images2 = result.get_images(output2)
assert len(images1) == 1, "Should have produced 1 image from sync node"
assert len(images2) == 1, "Should have produced 1 image from async node"
|