File size: 6,949 Bytes
4e424ea
ca753f0
4e424ea
3893fde
632fdb4
7bedcdd
4e424ea
 
 
 
 
 
 
 
ca753f0
73566e5
9220522
 
 
 
935512c
9220522
4d169d8
 
 
 
9220522
935512c
665534e
935512c
9220522
 
632fdb4
9220522
 
 
 
665534e
4e424ea
9d714b0
4e424ea
 
 
 
 
f0f4c78
4e424ea
 
 
632fdb4
 
 
 
 
 
 
9220522
9d714b0
632fdb4
 
 
 
 
 
 
 
9220522
632fdb4
 
9220522
632fdb4
9d714b0
 
632fdb4
 
 
 
9d714b0
9220522
665534e
 
 
 
 
632fdb4
 
9220522
632fdb4
 
 
 
 
 
 
9220522
632fdb4
 
 
9220522
632fdb4
9220522
632fdb4
 
9220522
 
 
 
 
 
 
632fdb4
9220522
 
 
632fdb4
9d714b0
 
632fdb4
 
 
 
9d714b0
9220522
9d714b0
632fdb4
9d714b0
632fdb4
935512c
632fdb4
ca753f0
9220522
 
632fdb4
 
9d714b0
632fdb4
9d714b0
9220522
632fdb4
 
 
 
9d714b0
632fdb4
 
 
 
9220522
632fdb4
 
f0f4c78
632fdb4
665534e
632fdb4
 
935512c
2562fab
f0f4c78
4e424ea
d701afa
4e424ea
 
f0f4c78
4e424ea
 
 
 
c81f025
4e424ea
 
 
 
 
 
 
 
 
 
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
import gradio as gr
import re 
import subprocess
import time
import select
from tqdm import tqdm
from huggingface_hub import snapshot_download

#Download model
snapshot_download(
    repo_id = "Wan-AI/Wan2.1-T2V-1.3B",
    local_dir = "./Wan2.1-T2V-1.3B"
)

def infer(prompt, progress=gr.Progress(track_tqdm=True)):
    
    # Configuration
    total_process_steps = 11          # Total INFO messages expected
    irrelevant_steps = 4              # First 4 INFO messages are ignored
    relevant_steps = total_process_steps - irrelevant_steps  # 7 overall (relevant) steps

    # Create overall process progress bar (Level 1)
    overall_bar = tqdm(total=relevant_steps, desc="Overall Process", position=1,
                       ncols=120, dynamic_ncols=False, leave=True)
    processed_steps = 0

    # Regex for video generation progress (Level 3)
    progress_pattern = re.compile(r"(\d+)%\|.*\| (\d+)/(\d+)")
    video_progress_bar = None

    # Variables for sub-step progress bar (Level 2)
    # We use a tick total of 500 ticks = 20 seconds (each tick = 40ms)
    sub_bar = None
    sub_ticks = 0
    sub_tick_total = 500
    # Flag indicating whether we're still waiting for the first relevant step.
    waiting_for_first_relevant = True

    command = [
        "python", "-u", "-m", "generate",  # using -u for unbuffered output
        "--task", "t2v-1.3B",
        "--size", "832*480",
        "--ckpt_dir", "./Wan2.1-T2V-1.3B",
        "--sample_shift", "8",
        "--sample_guide_scale", "6",
        "--prompt", prompt,
        "--save_file", "generated_video.mp4"
    ]

    process = subprocess.Popen(command,
                               stdout=subprocess.PIPE,
                               stderr=subprocess.STDOUT,
                               text=True,
                               bufsize=1)

    while True:
        # Poll stdout with a 40ms timeout.
        rlist, _, _ = select.select([process.stdout], [], [], 0.04)
        if rlist:
            line = process.stdout.readline()
            if not line:
                break
            stripped_line = line.strip()
            if not stripped_line:
                continue

            # Check if line matches video generation progress (Level 3)
            progress_match = progress_pattern.search(stripped_line)
            if progress_match:
                # Before entering video phase, cancel any active sub-step bar.
                if sub_bar is not None:
                    if sub_ticks < sub_tick_total:
                        sub_bar.update(sub_tick_total - sub_ticks)
                    sub_bar.close()
                    overall_bar.update(1)
                    overall_bar.refresh()
                    sub_bar = None
                    sub_ticks = 0
                # Enter video phase.
                current = int(progress_match.group(2))
                total = int(progress_match.group(3))
                if video_progress_bar is None:
                    video_progress_bar = tqdm(total=total, desc="Video Generation", position=0,
                                              ncols=120, dynamic_ncols=True, leave=True)
                video_progress_bar.update(current - video_progress_bar.n)
                video_progress_bar.refresh()
                # When video generation completes, update overall bar.
                if video_progress_bar.n >= video_progress_bar.total:
                    overall_bar.update(1)
                    overall_bar.refresh()
                    video_progress_bar.close()
                    video_progress_bar = None
                continue

            # Process INFO messages.
            if "INFO:" in stripped_line:
                parts = stripped_line.split("INFO:", 1)
                msg = parts[1].strip() if len(parts) > 1 else ""
                print(stripped_line)

                # For the first 4 INFO messages, we simply increment processed_steps.
                if processed_steps < irrelevant_steps:
                    processed_steps += 1
                    # If we're waiting for the first relevant step, start a waiting sub-bar if not already started.
                    if waiting_for_first_relevant and sub_bar is None:
                        sub_bar = tqdm(total=sub_tick_total, desc="Waiting for first step...", position=2,
                                       ncols=120, dynamic_ncols=False, leave=True)
                        sub_ticks = 0
                    # Continue reading logs.
                    continue
                else:
                    # Now we are in the relevant phase.
                    waiting_for_first_relevant = False
                    # If a sub-bar exists (either waiting or from a previous step), finish it.
                    if sub_bar is not None:
                        if sub_ticks < sub_tick_total:
                            sub_bar.update(sub_tick_total - sub_ticks)
                        sub_bar.close()
                        overall_bar.update(1)
                        overall_bar.refresh()
                        sub_bar = None
                        sub_ticks = 0
                    # Start a new sub-step bar with the current INFO message.
                    sub_bar = tqdm(total=sub_tick_total, desc=msg, position=2,
                                   ncols=120, dynamic_ncols=False, leave=True)
                    sub_ticks = 0
                continue
            else:
                print(stripped_line)
        else:
            # No new data within 40ms.
            # If a sub-bar is active, update it.
            if sub_bar is not None:
                sub_bar.update(1)
                sub_ticks += 1
                sub_bar.refresh()
                if sub_ticks >= sub_tick_total:
                    # 20 seconds have elapsed; finish this sub-step.
                    sub_bar.close()
                    overall_bar.update(1)
                    overall_bar.refresh()
                    sub_bar = None
                    sub_ticks = 0

        if process.poll() is not None:
            break

    # Drain remaining output.
    for line in process.stdout:
        print(line.strip())
    process.wait()
    if video_progress_bar is not None:
        video_progress_bar.close()
    if sub_bar is not None:
        sub_bar.close()
    overall_bar.close()
    
    if process.returncode == 0:
        print("Command executed successfully.")
        return "generated_video.mp4"
    else:
        print("Error executing command.")
        raise Exception("Error executing command")

with gr.Blocks() as demo:
    with gr.Column():
        gr.Markdown("# Wan 2.1")
        prompt = gr.Textbox(label="Prompt")
        submit_btn = gr.Button("Submit")
        video_res = gr.Video(label="Generated Video")

    submit_btn.click(
        fn = infer,
        inputs = [prompt],
        outputs = [video_res]
    )

demo.queue().launch(show_error=True, show_api=False, ssr_mode=False)