File size: 10,389 Bytes
dcdd99b
0557d7f
ace6ed9
 
 
21b17c3
ace6ed9
dcdd99b
0557d7f
dcdd99b
0557d7f
dcdd99b
9f13dde
0557d7f
dcdd99b
ace6ed9
 
 
 
 
 
 
 
 
 
 
 
 
 
6a4178d
 
 
 
 
 
 
 
 
ace6ed9
 
 
6a4178d
 
 
 
 
 
 
 
ace6ed9
 
dcdd99b
 
 
 
21b17c3
dcdd99b
21b17c3
dcdd99b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21b17c3
6a4178d
dcdd99b
21b17c3
 
 
 
 
0557d7f
dcdd99b
9f13dde
 
 
 
 
 
dcdd99b
 
9f13dde
dcdd99b
30343a2
 
 
 
 
 
dcdd99b
21b17c3
dcdd99b
30343a2
 
0557d7f
6a4178d
30343a2
 
 
 
 
dcdd99b
 
 
 
 
 
 
 
 
 
21b17c3
30343a2
 
 
dcdd99b
9f13dde
30343a2
 
6a4178d
dcdd99b
 
6a4178d
 
30343a2
dcdd99b
ace6ed9
dcdd99b
9f13dde
dcdd99b
30343a2
dcdd99b
6a4178d
dcdd99b
 
6a4178d
30343a2
 
ace6ed9
dcdd99b
30343a2
 
 
dcdd99b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
30343a2
0557d7f
dcdd99b
 
30343a2
 
 
 
ace6ed9
 
 
 
0557d7f
ace6ed9
0557d7f
ace6ed9
dcdd99b
21b17c3
dcdd99b
1cffa06
 
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
import os, json, traceback
from typing import Any, Dict, Tuple
import gradio as gr
from PIL import Image
import torch
from transformers import AutoProcessor, AutoTokenizer, AutoModelForCausalLM, AutoConfig

# -------- Env / params --------
MODEL_ID = os.environ.get("MODEL_ID", "inference-net/ClipTagger-12b")
HF_TOKEN = os.environ.get("HF_TOKEN")
TEMP = 0.1
MAX_NEW_TOKENS = 768  # faster demo; raise later if needed
DTYPE = torch.bfloat16 if torch.cuda.is_available() else torch.float32

# -------- Prompts (yours) --------
SYSTEM_PROMPT = (
    "You are an image annotation API trained to analyze YouTube video keyframes. "
    "You will be given instructions on the output format, what to caption, and how to perform your job. "
    "Follow those instructions. For descriptions and summaries, provide them directly and do not lead them with "
    "'This image shows' or 'This keyframe displays...', just get right into the details."
)

USER_PROMPT = """You are an image annotation API trained to analyze YouTube video keyframes. You must respond with a valid JSON object matching the exact structure below.

Your job is to extract detailed **factual elements directly visible** in the image. Do not speculate or interpret artistic intent, camera focus, or composition. Do not include phrases like "this appears to be", "this looks like", or anything about the image itself. Describe what **is physically present in the frame**, and nothing more.

Return JSON in this structure:

{
    "description": "A detailed, factual account of what is visibly happening (4 sentences max). Only mention concrete elements or actions that are clearly shown. Do not include anything about how the image is styled, shot, or composed. Do not lead the description with something like 'This image shows' or 'this keyframe is...', just get right into the details.",
    "objects": ["object1 with relevant visual details", "object2 with relevant visual details", ...],
    "actions": ["action1 with participants and context", "action2 with participants and context", ...],
    "environment": "Detailed factual description of the setting and atmosphere based on visible cues (e.g., interior of a classroom with fluorescent lighting, or outdoor forest path with snow-covered trees).",
    "content_type": "The type of content it is, e.g. 'real-world footage', 'video game', 'animation', 'cartoon', 'CGI', 'VTuber', etc.",
    "specific_style": "Specific genre, aesthetic, or platform style (e.g., anime, 3D animation, mobile gameplay, vlog, tutorial, news broadcast, etc.)",
    "production_quality": "Visible production level: e.g., 'professional studio', 'amateur handheld', 'webcam recording', 'TV broadcast', etc.",
    "summary": "One clear, comprehensive sentence summarizing the visual content of the frame. Like the description, get right to the point.",
    "logos": ["logo1 with visual description", "logo2 with visual description", ...]
}

Rules:
- Be specific and literal. Focus on what is explicitly visible.
- Do NOT include interpretations of emotion, mood, or narrative unless it's visually explicit.
- No artistic or cinematic analysis.
- Always include the language of any text in the image if present as an object, e.g. "English text", "Japanese text", "Russian text", etc.
- Maximum 10 objects and 5 actions.
- Return an empty array for 'logos' if none are present.
- Always output strictly valid JSON with proper escaping.
- Output **only the JSON**, no extra text or explanation.
"""

# -------- Utils --------
def extract_top_level_json(s: str):
    """Parse JSON; if extra text around it, extract the first balanced {...} block."""
    # Fast path
    try:
        return json.loads(s)
    except Exception:
        pass
    # Brace-stack extraction
    start = None
    depth = 0
    for i, ch in enumerate(s):
        if ch == '{':
            if depth == 0:
                start = i
            depth += 1
        elif ch == '}':
            if depth > 0:
                depth -= 1
                if depth == 0 and start is not None:
                    chunk = s[start:i+1]
                    try:
                        return json.loads(chunk)
                    except Exception:
                        # continue scanning for the next candidate
                        start = None
    return None

def build_messages(image: Image.Image):
    return [
        {"role": "system", "content": [{"type": "text", "text": SYSTEM_PROMPT}]},
        {"role": "user",   "content": [{"type": "image", "image": image},
                                       {"type": "text",  "text": USER_PROMPT}]}
    ]

def downscale_if_huge(pil: Image.Image, max_side: int = 1792) -> Image.Image:
    if pil is None:
        return pil
    w, h = pil.size
    m = max(w, h)
    if m <= max_side:
        return pil.convert("RGB")
    s = max_side / m
    return pil.convert("RGB").resize((int(w*s), int(h*s)), Image.BICUBIC)

# -------- Load model (A100) --------
processor = tokenizer = model = None
LOAD_ERROR = None

try:
    cfg = AutoConfig.from_pretrained(MODEL_ID, token=HF_TOKEN, trust_remote_code=True)
    if "clip" in cfg.__class__.__name__.lower():
        raise RuntimeError(f"MODEL_ID '{MODEL_ID}' is a CLIP/encoder repo; need a causal VLM.")

    print("[boot] loading processor…", flush=True)
    try:
        processor = AutoProcessor.from_pretrained(
            MODEL_ID, token=HF_TOKEN, trust_remote_code=True, use_fast=True
        )
    except TypeError:
        processor = AutoProcessor.from_pretrained(
            MODEL_ID, token=HF_TOKEN, trust_remote_code=True
        )

    print("[boot] loading model…", flush=True)
    # Force full-precision path on A100 first; add quantized path later if desired
    model = AutoModelForCausalLM.from_pretrained(
        MODEL_ID,
        token=HF_TOKEN,
        device_map="auto",
        torch_dtype=DTYPE,
        trust_remote_code=True,
        # quantization_config=None,  # keep commented if you want to honor repo quant; uncomment to force dequant
    )

    tokenizer = getattr(processor, "tokenizer", None) or AutoTokenizer.from_pretrained(
        MODEL_ID, token=HF_TOKEN, trust_remote_code=True, use_fast=True
    )
    print("[boot] ready.", flush=True)

except Exception as e:
    LOAD_ERROR = f"{e}\n\n{traceback.format_exc()}"

# -------- Inference --------
def generate(image: Image.Image) -> Tuple[str, Dict[str, Any] | None, bool]:
    if image is None:
        return "Please upload an image.", None, False
    if model is None or processor is None:
        return f"❌ Load error:\n{LOAD_ERROR}", None, False

    image = downscale_if_huge(image)

    # Build prompt
    if hasattr(processor, "apply_chat_template"):
        prompt = processor.apply_chat_template(build_messages(image), add_generation_prompt=True, tokenize=False)
    else:
        # fallback join (rare)
        prompt = USER_PROMPT

    # Tokenize with vision
    inputs = processor(text=prompt, images=image, return_tensors="pt").to(model.device)

    # Common gen kwargs
    eos = getattr(model.config, "eos_token_id", None)

    tried = []

    # (1) Greedy, no sampling (most stable, no temperature arg)
    try:
        g = dict(do_sample=False, max_new_tokens=MAX_NEW_TOKENS)
        if eos is not None:
            g["eos_token_id"] = eos
        with torch.inference_mode():
            out = model.generate(**inputs, **g)
        text = (processor.decode(out[0], skip_special_tokens=True)
                if hasattr(processor, "decode")
                else tokenizer.decode(out[0], skip_special_tokens=True))
        parsed = extract_top_level_json(text)
        if isinstance(parsed, dict):
            return json.dumps(parsed, indent=2), parsed, True
        tried.append(("greedy", "parsed-failed"))
    except Exception as e:
        tried.append(("greedy", f"err={e}"))

    # (2) Sampling with temperature=0.1
    try:
        g = dict(do_sample=True, temperature=TEMP, max_new_tokens=MAX_NEW_TOKENS)
        if eos is not None:
            g["eos_token_id"] = eos
        with torch.inference_mode():
            out = model.generate(**inputs, **g)
        text = (processor.decode(out[0], skip_special_tokens=True)
                if hasattr(processor, "decode")
                else tokenizer.decode(out[0], skip_special_tokens=True))
        parsed = extract_top_level_json(text)
        if isinstance(parsed, dict):
            return json.dumps(parsed, indent=2), parsed, True
        tried.append(("sample_t0.1", "parsed-failed"))
    except Exception as e:
        tried.append(("sample_t0.1", f"err={e}"))

    # (3) Shorter greedy
    try:
        g = dict(do_sample=False, max_new_tokens=min(512, MAX_NEW_TOKENS))
        if eos is not None:
            g["eos_token_id"] = eos
        with torch.inference_mode():
            out = model.generate(**inputs, **g)
        text = (processor.decode(out[0], skip_special_tokens=True)
                if hasattr(processor, "decode")
                else tokenizer.decode(out[0], skip_special_tokens=True))
        parsed = extract_top_level_json(text)
        if isinstance(parsed, dict):
            return json.dumps(parsed, indent=2), parsed, True
        tried.append(("greedy_short", "parsed-failed"))
    except Exception as e:
        tried.append(("greedy_short", f"err={e}"))

    # Debug info if all fail
    return "Generation failed.\nTried: " + "\n".join([f"{t[0]} -> {t[1]}" for t in tried]), None, False

# -------- UI --------
with gr.Blocks(theme=gr.themes.Soft(), analytics_enabled=False, title="Keyframe Annotator (Gemma-3 VLM · A100)") as demo:
    gr.Markdown("# Keyframe Annotator (Gemma-3-12B FT · A100)\nUpload an image to get **strict JSON** annotations.")
    if LOAD_ERROR:
        with gr.Accordion("Startup Error Details", open=False):
            gr.Markdown(f"```\n{LOAD_ERROR}\n```")

    with gr.Row():
        with gr.Column(scale=1):
            image = gr.Image(type="pil", label="Upload Image", image_mode="RGB")
            btn = gr.Button("Annotate", variant="primary")
        with gr.Column(scale=1):
            out_text = gr.Code(label="Output (JSON or error)")
            out_json = gr.JSON(label="Parsed JSON")
            ok_flag = gr.Checkbox(label="Valid JSON", value=False, interactive=False)

    btn.click(generate, inputs=[image], outputs=[out_text, out_json, ok_flag])

demo.queue(max_size=32).launch()