Spaces:
Paused
Paused
Update app_wip.py
Browse files- app_wip.py +126 -0
app_wip.py
CHANGED
|
@@ -19,3 +19,129 @@ snapshot_download(
|
|
| 19 |
repo_id='JaydenLu666/Reward-Forcing-T2V-1.3B',
|
| 20 |
local_dir='./checkpoints/Reward-Forcing-T2V-1.3B'
|
| 21 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
repo_id='JaydenLu666/Reward-Forcing-T2V-1.3B',
|
| 20 |
local_dir='./checkpoints/Reward-Forcing-T2V-1.3B'
|
| 21 |
)
|
| 22 |
+
|
| 23 |
+
import os
|
| 24 |
+
import uuid
|
| 25 |
+
import subprocess
|
| 26 |
+
from datetime import datetime
|
| 27 |
+
|
| 28 |
+
import gradio as gr
|
| 29 |
+
|
| 30 |
+
# === Chemins à adapter si besoin ===
|
| 31 |
+
CONFIG_PATH = "configs/reward_forcing.yaml"
|
| 32 |
+
CHECKPOINT_PATH = "checkpoints/Reward-Forcing-T2V-1.3B/rewardforcing.pt"
|
| 33 |
+
|
| 34 |
+
PROMPT_DIR = "prompts/gradio_inputs"
|
| 35 |
+
OUTPUT_ROOT = "videos/gradio_outputs"
|
| 36 |
+
|
| 37 |
+
os.makedirs(PROMPT_DIR, exist_ok=True)
|
| 38 |
+
os.makedirs(OUTPUT_ROOT, exist_ok=True)
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
def run_inference(prompt: str, duration: str, use_ema: bool):
|
| 42 |
+
"""
|
| 43 |
+
1. Écrit le prompt dans un fichier .txt
|
| 44 |
+
2. Lance inference.py avec ce fichier comme --data_path
|
| 45 |
+
3. Retourne le chemin de la première vidéo .mp4 générée + les logs
|
| 46 |
+
"""
|
| 47 |
+
if not prompt or not prompt.strip():
|
| 48 |
+
raise gr.Error("Veuillez entrer un prompt texte 🙂")
|
| 49 |
+
|
| 50 |
+
# 1) On mappe la durée choisie → num_output_frames
|
| 51 |
+
if duration == "5s (21 frames)":
|
| 52 |
+
num_output_frames = 21
|
| 53 |
+
else: # "30s (120 frames)"
|
| 54 |
+
num_output_frames = 120
|
| 55 |
+
|
| 56 |
+
# 2) Fichier .txt temporaire pour le prompt
|
| 57 |
+
prompt_id = uuid.uuid4().hex[:8]
|
| 58 |
+
prompt_path = os.path.join(PROMPT_DIR, f"prompt_{prompt_id}.txt")
|
| 59 |
+
|
| 60 |
+
with open(prompt_path, "w", encoding="utf-8") as f:
|
| 61 |
+
# TextDataset lit juste chaque ligne comme un prompt
|
| 62 |
+
f.write(prompt.strip() + "\n")
|
| 63 |
+
|
| 64 |
+
# 3) Dossier de sortie unique pour cette génération
|
| 65 |
+
ts = datetime.now().strftime("%Y%m%d_%H%M%S")
|
| 66 |
+
output_folder = os.path.join(OUTPUT_ROOT, f"{ts}_{prompt_id}")
|
| 67 |
+
os.makedirs(output_folder, exist_ok=True)
|
| 68 |
+
|
| 69 |
+
# 4) Commande inference.py
|
| 70 |
+
cmd = [
|
| 71 |
+
"python",
|
| 72 |
+
"inference.py",
|
| 73 |
+
"--num_output_frames", str(num_output_frames),
|
| 74 |
+
"--config_path", CONFIG_PATH,
|
| 75 |
+
"--checkpoint_path", CHECKPOINT_PATH,
|
| 76 |
+
"--output_folder", output_folder,
|
| 77 |
+
"--data_path", prompt_path,
|
| 78 |
+
"--num_samples", "1",
|
| 79 |
+
]
|
| 80 |
+
if use_ema:
|
| 81 |
+
cmd.append("--use_ema")
|
| 82 |
+
|
| 83 |
+
result = subprocess.run(
|
| 84 |
+
cmd,
|
| 85 |
+
stdout=subprocess.PIPE,
|
| 86 |
+
stderr=subprocess.STDOUT,
|
| 87 |
+
text=True,
|
| 88 |
+
)
|
| 89 |
+
|
| 90 |
+
logs = result.stdout
|
| 91 |
+
|
| 92 |
+
# 5) On récupère la première vidéo produite
|
| 93 |
+
mp4s = [f for f in os.listdir(output_folder) if f.lower().endswith(".mp4")]
|
| 94 |
+
if not mp4s:
|
| 95 |
+
raise gr.Error(
|
| 96 |
+
"Aucune vidéo trouvée dans le dossier de sortie.\n"
|
| 97 |
+
"Regarde les logs ci-dessous pour voir ce qui a coincé."
|
| 98 |
+
)
|
| 99 |
+
|
| 100 |
+
mp4s.sort()
|
| 101 |
+
video_path = os.path.join(output_folder, mp4s[0])
|
| 102 |
+
return video_path, logs
|
| 103 |
+
|
| 104 |
+
|
| 105 |
+
with gr.Blocks(title="Reward Forcing T2V Demo") as demo:
|
| 106 |
+
gr.Markdown(
|
| 107 |
+
"""
|
| 108 |
+
# 🎬 Reward Forcing – Text-to-Video
|
| 109 |
+
|
| 110 |
+
Entrez un prompt texte, on génère un fichier `.txt` en interne
|
| 111 |
+
puis on lance `inference.py` avec ce fichier comme `--data_path`.
|
| 112 |
+
"""
|
| 113 |
+
)
|
| 114 |
+
|
| 115 |
+
with gr.Row():
|
| 116 |
+
prompt_in = gr.Textbox(
|
| 117 |
+
label="Prompt",
|
| 118 |
+
placeholder="Ex: A cinematic shot of a spaceship flying above a neon city at night...",
|
| 119 |
+
lines=4,
|
| 120 |
+
)
|
| 121 |
+
|
| 122 |
+
with gr.Row():
|
| 123 |
+
duration = gr.Radio(
|
| 124 |
+
["5s (21 frames)", "30s (120 frames)"],
|
| 125 |
+
value="5s (21 frames)",
|
| 126 |
+
label="Durée",
|
| 127 |
+
)
|
| 128 |
+
use_ema = gr.Checkbox(value=True, label="Utiliser les poids EMA (--use_ema)")
|
| 129 |
+
|
| 130 |
+
generate_btn = gr.Button("🚀 Générer la vidéo", variant="primary")
|
| 131 |
+
|
| 132 |
+
with gr.Row():
|
| 133 |
+
video_out = gr.Video(label="Vidéo générée")
|
| 134 |
+
logs_out = gr.Textbox(
|
| 135 |
+
label="Logs de inference.py",
|
| 136 |
+
lines=10,
|
| 137 |
+
interactive=False,
|
| 138 |
+
)
|
| 139 |
+
|
| 140 |
+
generate_btn.click(
|
| 141 |
+
fn=run_inference,
|
| 142 |
+
inputs=[prompt_in, duration, use_ema],
|
| 143 |
+
outputs=[video_out, logs_out],
|
| 144 |
+
)
|
| 145 |
+
|
| 146 |
+
if __name__ == "__main__":
|
| 147 |
+
demo.launch()
|