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
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: 3,230 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 | import asyncio
import contextlib
import os
import re
import time
from collections.abc import Callable
from io import BytesIO
from yarl import URL
from comfy.cli_args import args
from comfy.model_management import processing_interrupted
from comfy_api.latest import IO
from .common_exceptions import ProcessingInterrupted
_HAS_PCT_ESC = re.compile(r"%[0-9A-Fa-f]{2}") # any % followed by 2 hex digits
_HAS_BAD_PCT = re.compile(r"%(?![0-9A-Fa-f]{2})") # any % not followed by 2 hex digits
def is_processing_interrupted() -> bool:
"""Return True if user/runtime requested interruption."""
return processing_interrupted()
def get_node_id(node_cls: type[IO.ComfyNode]) -> str:
return node_cls.hidden.unique_id
def get_auth_header(node_cls: type[IO.ComfyNode]) -> dict[str, str]:
if node_cls.hidden.auth_token_comfy_org:
return {"Authorization": f"Bearer {node_cls.hidden.auth_token_comfy_org}"}
if node_cls.hidden.api_key_comfy_org:
return {"X-API-KEY": node_cls.hidden.api_key_comfy_org}
return {}
def default_base_url() -> str:
return getattr(args, "comfy_api_base", "https://api.comfy.org")
async def sleep_with_interrupt(
seconds: float,
node_cls: type[IO.ComfyNode] | None,
label: str | None = None,
start_ts: float | None = None,
estimated_total: int | None = None,
*,
display_callback: Callable[[type[IO.ComfyNode], str, int, int | None], None] | None = None,
):
"""
Sleep in 1s slices while:
- Checking for interruption (raises ProcessingInterrupted).
- Optionally emitting time progress via display_callback (if provided).
"""
end = time.monotonic() + seconds
while True:
if is_processing_interrupted():
raise ProcessingInterrupted("Task cancelled")
now = time.monotonic()
if start_ts is not None and label and display_callback:
with contextlib.suppress(Exception):
display_callback(node_cls, label, int(now - start_ts), estimated_total)
if now >= end:
break
await asyncio.sleep(min(1.0, end - now))
def mimetype_to_extension(mime_type: str) -> str:
"""Converts a MIME type to a file extension."""
return mime_type.split("/")[-1].lower()
def get_fs_object_size(path_or_object: str | BytesIO) -> int:
if isinstance(path_or_object, str):
return os.path.getsize(path_or_object)
return len(path_or_object.getvalue())
def to_aiohttp_url(url: str) -> URL:
"""If `url` appears to be already percent-encoded (contains at least one valid %HH
escape and no malformed '%' sequences) and contains no raw whitespace/control
characters preserve the original encoding byte-for-byte (important for signed/presigned URLs).
Otherwise, return `URL(url)` and allow yarl to normalize/quote as needed."""
if any(c.isspace() for c in url) or any(ord(c) < 0x20 for c in url):
# Avoid encoded=True if URL contains raw whitespace/control chars
return URL(url)
if _HAS_PCT_ESC.search(url) and not _HAS_BAD_PCT.search(url):
# Preserve encoding only if it appears pre-encoded AND has no invalid % sequences
return URL(url, encoded=True)
return URL(url)
|