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: 5,938 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 | from comfy_execution.graph_utils import GraphBuilder, is_link
from comfy_execution.graph import ExecutionBlocker
from .tools import VariantSupport
NUM_FLOW_SOCKETS = 5
@VariantSupport()
class TestWhileLoopOpen:
def __init__(self):
pass
@classmethod
def INPUT_TYPES(cls):
inputs = {
"required": {
"condition": ("BOOLEAN", {"default": True}),
},
"optional": {
},
}
for i in range(NUM_FLOW_SOCKETS):
inputs["optional"][f"initial_value{i}"] = ("*",)
return inputs
RETURN_TYPES = tuple(["FLOW_CONTROL"] + ["*"] * NUM_FLOW_SOCKETS)
RETURN_NAMES = tuple(["FLOW_CONTROL"] + [f"value{i}" for i in range(NUM_FLOW_SOCKETS)])
FUNCTION = "while_loop_open"
CATEGORY = "Testing/Flow"
def while_loop_open(self, condition, **kwargs):
values = []
for i in range(NUM_FLOW_SOCKETS):
values.append(kwargs.get(f"initial_value{i}", None))
return tuple(["stub"] + values)
@VariantSupport()
class TestWhileLoopClose:
def __init__(self):
pass
@classmethod
def INPUT_TYPES(cls):
inputs = {
"required": {
"flow_control": ("FLOW_CONTROL", {"rawLink": True}),
"condition": ("BOOLEAN", {"forceInput": True}),
},
"optional": {
},
"hidden": {
"dynprompt": "DYNPROMPT",
"unique_id": "UNIQUE_ID",
}
}
for i in range(NUM_FLOW_SOCKETS):
inputs["optional"][f"initial_value{i}"] = ("*",)
return inputs
RETURN_TYPES = tuple(["*"] * NUM_FLOW_SOCKETS)
RETURN_NAMES = tuple([f"value{i}" for i in range(NUM_FLOW_SOCKETS)])
FUNCTION = "while_loop_close"
CATEGORY = "Testing/Flow"
def explore_dependencies(self, node_id, dynprompt, upstream):
node_info = dynprompt.get_node(node_id)
if "inputs" not in node_info:
return
for k, v in node_info["inputs"].items():
if is_link(v):
parent_id = v[0]
if parent_id not in upstream:
upstream[parent_id] = []
self.explore_dependencies(parent_id, dynprompt, upstream)
upstream[parent_id].append(node_id)
def collect_contained(self, node_id, upstream, contained):
if node_id not in upstream:
return
for child_id in upstream[node_id]:
if child_id not in contained:
contained[child_id] = True
self.collect_contained(child_id, upstream, contained)
def while_loop_close(self, flow_control, condition, dynprompt=None, unique_id=None, **kwargs):
assert dynprompt is not None
if not condition:
# We're done with the loop
values = []
for i in range(NUM_FLOW_SOCKETS):
values.append(kwargs.get(f"initial_value{i}", None))
return tuple(values)
# We want to loop
upstream = {}
# Get the list of all nodes between the open and close nodes
self.explore_dependencies(unique_id, dynprompt, upstream)
contained = {}
open_node = flow_control[0]
self.collect_contained(open_node, upstream, contained)
contained[unique_id] = True
contained[open_node] = True
# We'll use the default prefix, but to avoid having node names grow exponentially in size,
# we'll use "Recurse" for the name of the recursively-generated copy of this node.
graph = GraphBuilder()
for node_id in contained:
original_node = dynprompt.get_node(node_id)
node = graph.node(original_node["class_type"], "Recurse" if node_id == unique_id else node_id)
node.set_override_display_id(node_id)
for node_id in contained:
original_node = dynprompt.get_node(node_id)
node = graph.lookup_node("Recurse" if node_id == unique_id else node_id)
assert node is not None
for k, v in original_node["inputs"].items():
if is_link(v) and v[0] in contained:
parent = graph.lookup_node(v[0])
assert parent is not None
node.set_input(k, parent.out(v[1]))
else:
node.set_input(k, v)
new_open = graph.lookup_node(open_node)
assert new_open is not None
for i in range(NUM_FLOW_SOCKETS):
key = f"initial_value{i}"
new_open.set_input(key, kwargs.get(key, None))
my_clone = graph.lookup_node("Recurse")
assert my_clone is not None
result = map(lambda x: my_clone.out(x), range(NUM_FLOW_SOCKETS))
return {
"result": tuple(result),
"expand": graph.finalize(),
}
@VariantSupport()
class TestExecutionBlockerNode:
def __init__(self):
pass
@classmethod
def INPUT_TYPES(cls):
inputs = {
"required": {
"input": ("*",),
"block": ("BOOLEAN",),
"verbose": ("BOOLEAN", {"default": False}),
},
}
return inputs
RETURN_TYPES = ("*",)
RETURN_NAMES = ("output",)
FUNCTION = "execution_blocker"
CATEGORY = "Testing/Flow"
def execution_blocker(self, input, block, verbose):
if block:
return (ExecutionBlocker("Blocked Execution" if verbose else None),)
return (input,)
FLOW_CONTROL_NODE_CLASS_MAPPINGS = {
"TestWhileLoopOpen": TestWhileLoopOpen,
"TestWhileLoopClose": TestWhileLoopClose,
"TestExecutionBlocker": TestExecutionBlockerNode,
}
FLOW_CONTROL_NODE_DISPLAY_NAME_MAPPINGS = {
"TestWhileLoopOpen": "While Loop Open",
"TestWhileLoopClose": "While Loop Close",
"TestExecutionBlocker": "Execution Blocker",
}
|