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: 6,338 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 | import logging
import os
import uuid
from typing import Callable
from aiohttp import web
import folder_paths
from app.assets.api.schemas_in import ParsedUpload, UploadError
from app.assets.helpers import validate_blake3_hash
def normalize_and_validate_hash(s: str) -> str:
"""Validate and normalize a hash string.
Returns canonical 'blake3:<hex>' or raises UploadError.
"""
try:
return validate_blake3_hash(s)
except ValueError:
raise UploadError(400, "INVALID_HASH", "hash must be like 'blake3:<hex>'")
async def parse_multipart_upload(
request: web.Request,
check_hash_exists: Callable[[str], bool],
) -> ParsedUpload:
"""
Parse a multipart/form-data upload request.
Args:
request: The aiohttp request
check_hash_exists: Callable(hash_str) -> bool to check if a hash exists
Returns:
ParsedUpload with parsed fields and temp file path
Raises:
UploadError: On validation or I/O errors
"""
if not (request.content_type or "").lower().startswith("multipart/"):
raise UploadError(
415, "UNSUPPORTED_MEDIA_TYPE", "Use multipart/form-data for uploads."
)
reader = await request.multipart()
file_present = False
file_client_name: str | None = None
tags_raw: list[str] = []
provided_name: str | None = None
user_metadata_raw: str | None = None
provided_hash: str | None = None
provided_hash_exists: bool | None = None
provided_mime_type: str | None = None
provided_preview_id: str | None = None
file_written = 0
tmp_path: str | None = None
while True:
field = await reader.next()
if field is None:
break
fname = getattr(field, "name", "") or ""
if fname == "hash":
try:
s = ((await field.text()) or "").strip().lower()
except Exception:
raise UploadError(
400, "INVALID_HASH", "hash must be like 'blake3:<hex>'"
)
if s:
provided_hash = normalize_and_validate_hash(s)
try:
provided_hash_exists = check_hash_exists(provided_hash)
except Exception as e:
logging.exception(
"check_hash_exists failed for hash=%s: %s", provided_hash, e
)
raise UploadError(
500,
"HASH_CHECK_FAILED",
"Backend error while checking asset hash.",
)
elif fname == "file":
file_present = True
file_client_name = (field.filename or "").strip()
if provided_hash and provided_hash_exists is True:
# Hash exists - drain file but don't write to disk
try:
while True:
chunk = await field.read_chunk(8 * 1024 * 1024)
if not chunk:
break
file_written += len(chunk)
except Exception:
raise UploadError(
500, "UPLOAD_IO_ERROR", "Failed to receive uploaded file."
)
continue
uploads_root = os.path.join(folder_paths.get_temp_directory(), "uploads")
unique_dir = os.path.join(uploads_root, uuid.uuid4().hex)
os.makedirs(unique_dir, exist_ok=True)
tmp_path = os.path.join(unique_dir, ".upload.part")
try:
with open(tmp_path, "wb") as f:
while True:
chunk = await field.read_chunk(8 * 1024 * 1024)
if not chunk:
break
f.write(chunk)
file_written += len(chunk)
except Exception:
delete_temp_file_if_exists(tmp_path)
raise UploadError(
500, "UPLOAD_IO_ERROR", "Failed to receive and store uploaded file."
)
elif fname == "tags":
tags_raw.append((await field.text()) or "")
elif fname == "name":
provided_name = (await field.text()) or None
elif fname == "user_metadata":
user_metadata_raw = (await field.text()) or None
elif fname == "id":
raise UploadError(
400,
"UNSUPPORTED_FIELD",
"Client-provided 'id' is not supported. Asset IDs are assigned by the server.",
)
elif fname == "mime_type":
provided_mime_type = ((await field.text()) or "").strip() or None
elif fname == "preview_id":
provided_preview_id = ((await field.text()) or "").strip() or None
if not file_present and not (provided_hash and provided_hash_exists):
raise UploadError(
400, "MISSING_FILE", "Form must include a 'file' part or a known 'hash'."
)
if (
file_present
and file_written == 0
and not (provided_hash and provided_hash_exists)
):
delete_temp_file_if_exists(tmp_path)
raise UploadError(400, "EMPTY_UPLOAD", "Uploaded file is empty.")
return ParsedUpload(
file_present=file_present,
file_written=file_written,
file_client_name=file_client_name,
tmp_path=tmp_path,
tags_raw=tags_raw,
provided_name=provided_name,
user_metadata_raw=user_metadata_raw,
provided_hash=provided_hash,
provided_hash_exists=provided_hash_exists,
provided_mime_type=provided_mime_type,
provided_preview_id=provided_preview_id,
)
def delete_temp_file_if_exists(tmp_path: str | None) -> None:
"""Safely remove a temp file and its parent directory if empty."""
if tmp_path:
try:
if os.path.exists(tmp_path):
os.remove(tmp_path)
except OSError as e:
logging.debug("Failed to delete temp file %s: %s", tmp_path, e)
try:
parent = os.path.dirname(tmp_path)
if parent and os.path.isdir(parent):
os.rmdir(parent) # only succeeds if empty
except OSError:
pass
|