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: 5,648 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 | import random
import server
from enum import Enum
class SGmode(Enum):
FIX = 1
INCR = 2
DECR = 3
RAND = 4
class SeedGenerator:
def __init__(self, base_value, action):
self.base_value = base_value
if action == "fixed" or action == "increment" or action == "decrement" or action == "randomize":
self.action = SGmode.FIX
elif action == 'increment for each node':
self.action = SGmode.INCR
elif action == 'decrement for each node':
self.action = SGmode.DECR
elif action == 'randomize for each node':
self.action = SGmode.RAND
def next(self):
seed = self.base_value
if self.action == SGmode.INCR:
self.base_value += 1
if self.base_value > 1125899906842624:
self.base_value = 0
elif self.action == SGmode.DECR:
self.base_value -= 1
if self.base_value < 0:
self.base_value = 1125899906842624
elif self.action == SGmode.RAND:
self.base_value = random.randint(0, 1125899906842624)
return seed
def control_seed(v, action, seed_is_global):
action = v['inputs']['action'] if seed_is_global else action
value = v['inputs']['value'] if seed_is_global else v['inputs']['seed_num']
if action == 'increment' or action == 'increment for each node':
value = value + 1
if value > 1125899906842624:
value = 0
elif action == 'decrement' or action == 'decrement for each node':
value = value - 1
if value < 0:
value = 1125899906842624
elif action == 'randomize' or action == 'randomize for each node':
value = random.randint(0, 1125899906842624)
if seed_is_global:
v['inputs']['value'] = value
return value
def prompt_seed_update(json_data):
try:
seed_widget_map = json_data['extra_data']['extra_pnginfo']['workflow']['seed_widgets']
except:
return None
workflow = json_data['extra_data']['extra_pnginfo']['workflow']
seed_widget_map = workflow['seed_widgets']
value = None
mode = None
node = None
action = None
seed_is_global = False
for k, v in json_data['prompt'].items():
if 'class_type' not in v:
continue
cls = v['class_type']
if cls == 'easy globalSeed':
mode = v['inputs']['mode']
action = v['inputs']['action']
value = v['inputs']['value']
node = k, v
seed_is_global = True
# control before generated
if mode is not None and mode and seed_is_global:
value = control_seed(node[1], action, seed_is_global)
if seed_is_global:
if value is not None:
seed_generator = SeedGenerator(value, action)
for k, v in json_data['prompt'].items():
for k2, v2 in v['inputs'].items():
if isinstance(v2, str) and '$GlobalSeed.value$' in v2:
v['inputs'][k2] = v2.replace('$GlobalSeed.value$', str(value))
if k not in seed_widget_map:
continue
if 'seed_num' in v['inputs']:
if isinstance(v['inputs']['seed_num'], int):
v['inputs']['seed_num'] = seed_generator.next()
if 'seed' in v['inputs']:
if isinstance(v['inputs']['seed'], int):
v['inputs']['seed'] = seed_generator.next()
if 'noise_seed' in v['inputs']:
if isinstance(v['inputs']['noise_seed'], int):
v['inputs']['noise_seed'] = seed_generator.next()
for k2, v2 in v['inputs'].items():
if isinstance(v2, str) and '$GlobalSeed.value$' in v2:
v['inputs'][k2] = v2.replace('$GlobalSeed.value$', str(value))
# control after generated
if mode is not None and not mode:
control_seed(node[1], action, seed_is_global)
return value is not None
def workflow_seed_update(json_data):
nodes = json_data['extra_data']['extra_pnginfo']['workflow']['nodes']
seed_widget_map = json_data['extra_data']['extra_pnginfo']['workflow']['seed_widgets']
prompt = json_data['prompt']
updated_seed_map = {}
value = None
for node in nodes:
node_id = str(node['id'])
if node_id in prompt:
if node['type'] == 'easy globalSeed':
value = prompt[node_id]['inputs']['value']
length = len(node['widgets_values'])
node['widgets_values'][length-1] = node['widgets_values'][0]
node['widgets_values'][0] = value
elif node_id in seed_widget_map:
widget_idx = seed_widget_map[node_id]
if 'seed_num' in prompt[node_id]['inputs']:
seed = prompt[node_id]['inputs']['seed_num']
elif 'noise_seed' in prompt[node_id]['inputs']:
seed = prompt[node_id]['inputs']['noise_seed']
else:
seed = prompt[node_id]['inputs']['seed']
node['widgets_values'][widget_idx] = seed
updated_seed_map[node_id] = seed
server.PromptServer.instance.send_sync("easyuse-global-seed", {"id": node_id, "value": value, "seed_map": updated_seed_map})
def onprompt(json_data):
is_changed = prompt_seed_update(json_data)
if is_changed:
workflow_seed_update(json_data)
return json_data
server.PromptServer.instance.add_on_prompt_handler(onprompt) |