Spaces:
Sleeping
Sleeping
File size: 9,779 Bytes
0b4562b |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 |
"""
Hugging Face Space for STARFlow
Text-to-Image and Text-to-Video Generation
This app allows you to run STARFlow inference on Hugging Face GPU infrastructure.
"""
import os
import gradio as gr
import torch
import subprocess
import pathlib
from pathlib import Path
# Check if running on Hugging Face Spaces
HF_SPACE = os.environ.get("SPACE_ID") is not None
# Verify CUDA availability (will be True on HF Spaces with GPU hardware)
if torch.cuda.is_available():
print(f"✅ CUDA available! Device: {torch.cuda.get_device_name(0)}")
print(f" CUDA Version: {torch.version.cuda}")
print(f" PyTorch Version: {torch.__version__}")
else:
print("⚠️ CUDA not available. Make sure GPU hardware is selected in Space settings.")
def generate_image(prompt, aspect_ratio, cfg, seed, checkpoint_file, config_path):
"""Generate image from text prompt."""
if checkpoint_file is None:
return None, "Error: Please upload a checkpoint file."
# Handle Gradio file object
if hasattr(checkpoint_file, 'name'):
checkpoint_path = checkpoint_file.name
else:
checkpoint_path = str(checkpoint_file)
if not os.path.exists(checkpoint_path):
return None, f"Error: Checkpoint file not found at {checkpoint_path}."
if not config_path or not os.path.exists(config_path):
return None, "Error: Config file not found. Please ensure config file exists."
try:
# Create output directory
output_dir = Path("outputs")
output_dir.mkdir(exist_ok=True)
# Run sampling command
cmd = [
"python", "sample.py",
"--model_config_path", config_path,
"--checkpoint_path", checkpoint_path,
"--caption", prompt,
"--sample_batch_size", "1",
"--cfg", str(cfg),
"--aspect_ratio", aspect_ratio,
"--seed", str(seed),
"--save_folder", "1",
"--finetuned_vae", "none",
"--jacobi", "1",
"--jacobi_th", "0.001",
"--jacobi_block_size", "16"
]
result = subprocess.run(cmd, capture_output=True, text=True, cwd=os.getcwd())
if result.returncode != 0:
return None, f"Error: {result.stderr}"
# Find the generated image
# The sample.py script saves to logdir/model_name/...
# We need to find the most recent output
output_files = list(output_dir.glob("**/*.png")) + list(output_dir.glob("**/*.jpg"))
if output_files:
latest_file = max(output_files, key=lambda p: p.stat().st_mtime)
return str(latest_file), "Success! Image generated."
else:
return None, "Error: Generated image not found."
except Exception as e:
return None, f"Error: {str(e)}"
def generate_video(prompt, aspect_ratio, cfg, seed, target_length, checkpoint_file, config_path, input_image):
"""Generate video from text prompt."""
if checkpoint_file is None:
return None, "Error: Please upload a checkpoint file."
# Handle Gradio file object
if hasattr(checkpoint_file, 'name'):
checkpoint_path = checkpoint_file.name
else:
checkpoint_path = str(checkpoint_file)
if not os.path.exists(checkpoint_path):
return None, f"Error: Checkpoint file not found at {checkpoint_path}."
if not config_path or not os.path.exists(config_path):
return None, "Error: Config file not found. Please ensure config file exists."
# Handle input image
input_image_path = None
if input_image is not None:
if hasattr(input_image, 'name'):
input_image_path = input_image.name
else:
input_image_path = str(input_image)
try:
# Create output directory
output_dir = Path("outputs")
output_dir.mkdir(exist_ok=True)
# Run sampling command
cmd = [
"python", "sample.py",
"--model_config_path", config_path,
"--checkpoint_path", checkpoint_path,
"--caption", prompt,
"--sample_batch_size", "1",
"--cfg", str(cfg),
"--aspect_ratio", aspect_ratio,
"--seed", str(seed),
"--out_fps", "16",
"--save_folder", "1",
"--jacobi", "1",
"--jacobi_th", "0.001",
"--finetuned_vae", "none",
"--disable_learnable_denoiser", "0",
"--jacobi_block_size", "32",
"--target_length", str(target_length)
]
if input_image_path and os.path.exists(input_image_path):
cmd.extend(["--input_image", input_image_path])
else:
cmd.extend(["--input_image", "none"])
result = subprocess.run(cmd, capture_output=True, text=True, cwd=os.getcwd())
if result.returncode != 0:
return None, f"Error: {result.stderr}"
# Find the generated video
output_files = list(output_dir.glob("**/*.mp4")) + list(output_dir.glob("**/*.gif"))
if output_files:
latest_file = max(output_files, key=lambda p: p.stat().st_mtime)
return str(latest_file), "Success! Video generated."
else:
return None, "Error: Generated video not found."
except Exception as e:
return None, f"Error: {str(e)}"
# Create Gradio interface
with gr.Blocks(title="STARFlow - Text-to-Image & Video Generation") as demo:
gr.Markdown("""
# STARFlow: Scalable Transformer Auto-Regressive Flow
Generate high-quality images and videos from text prompts using STARFlow models.
**Note**: You'll need to upload model checkpoints. Check the README for model download links.
""")
with gr.Tabs():
with gr.Tab("Text-to-Image"):
with gr.Row():
with gr.Column():
image_prompt = gr.Textbox(
label="Prompt",
placeholder="a film still of a cat playing piano",
lines=3
)
image_checkpoint = gr.File(
label="Model Checkpoint (.pth file)",
file_types=[".pth"]
)
image_config = gr.Textbox(
label="Config Path",
value="configs/starflow_3B_t2i_256x256.yaml",
placeholder="configs/starflow_3B_t2i_256x256.yaml"
)
image_aspect = gr.Dropdown(
choices=["1:1", "2:3", "3:2", "16:9", "9:16", "4:5", "5:4"],
value="1:1",
label="Aspect Ratio"
)
image_cfg = gr.Slider(1.0, 10.0, value=3.6, step=0.1, label="CFG Scale")
image_seed = gr.Number(value=999, label="Seed", precision=0)
image_btn = gr.Button("Generate Image", variant="primary")
with gr.Column():
image_output = gr.Image(label="Generated Image")
image_status = gr.Textbox(label="Status", interactive=False)
image_btn.click(
fn=generate_image,
inputs=[image_prompt, image_aspect, image_cfg, image_seed, image_checkpoint, image_config],
outputs=[image_output, image_status],
show_progress=True
)
with gr.Tab("Text-to-Video"):
with gr.Row():
with gr.Column():
video_prompt = gr.Textbox(
label="Prompt",
placeholder="a corgi dog looks at the camera",
lines=3
)
video_checkpoint = gr.File(
label="Model Checkpoint (.pth file)",
file_types=[".pth"]
)
video_config = gr.Textbox(
label="Config Path",
value="configs/starflow-v_7B_t2v_caus_480p.yaml",
placeholder="configs/starflow-v_7B_t2v_caus_480p.yaml"
)
video_aspect = gr.Dropdown(
choices=["16:9", "1:1", "4:3"],
value="16:9",
label="Aspect Ratio"
)
video_cfg = gr.Slider(1.0, 10.0, value=3.5, step=0.1, label="CFG Scale")
video_seed = gr.Number(value=99, label="Seed", precision=0)
video_length = gr.Slider(81, 481, value=81, step=80, label="Target Length (frames)")
video_input_image = gr.File(
label="Input Image (optional, for image-to-video)",
file_types=["image"]
)
video_btn = gr.Button("Generate Video", variant="primary")
with gr.Column():
video_output = gr.Video(label="Generated Video")
video_status = gr.Textbox(label="Status", interactive=False)
video_btn.click(
fn=generate_video,
inputs=[video_prompt, video_aspect, video_cfg, video_seed, video_length,
video_checkpoint, video_config, video_input_image],
outputs=[video_output, video_status],
show_progress=True
)
if __name__ == "__main__":
demo.launch(server_name="0.0.0.0", server_port=7860)
|