Add files using upload-large-folder tool
Browse files- START_HERE.md +79 -0
- from +0 -0
- import +0 -0
- results/comparison/uipress_256/uipress_256/clip_scores.json +58 -0
- results/comparison/uipress_256/uipress_256/html_predictions/0.html +49 -0
- results/comparison/uipress_256/uipress_256/html_predictions/10.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/100.html +25 -0
- results/comparison/uipress_256/uipress_256/html_predictions/101.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/103.html +25 -0
- results/comparison/uipress_256/uipress_256/html_predictions/106.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/108.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/11.html +594 -0
- results/comparison/uipress_256/uipress_256/html_predictions/115.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/116.html +11 -0
- results/comparison/uipress_256/uipress_256/html_predictions/119.html +8 -0
- results/comparison/uipress_256/uipress_256/html_predictions/121.html +10 -0
- results/comparison/uipress_256/uipress_256/html_predictions/123.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/125.html +35 -0
- results/comparison/uipress_256/uipress_256/html_predictions/126.html +9 -0
- results/comparison/uipress_256/uipress_256/html_predictions/131.html +558 -0
- results/comparison/uipress_256/uipress_256/html_predictions/137.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/138.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/139.html +5 -0
- results/comparison/uipress_256/uipress_256/html_predictions/14.html +625 -0
- results/comparison/uipress_256/uipress_256/html_predictions/141.html +5 -0
- results/comparison/uipress_256/uipress_256/per_sample.json +352 -0
- results/comparison/uipress_256/uipress_256/summary.json +8 -0
- results/comparison/visionzip_256_quick2/visionzip_256/html_predictions/0.html +464 -0
- results/comparison/visionzip_256_quick2/visionzip_256/per_sample.json +9 -0
- results/comparison/visionzip_256_quick2/visionzip_256/summary.json +8 -0
- results/element_analysis.json +200 -0
- scripts/eval_all.py +298 -153
- scripts/step_case_study.py +230 -0
- scripts/step_clip_batch.py +170 -0
- scripts/step_element_analysis.py +256 -0
- scripts/step_ssim_bootstrap.py +228 -0
- scripts/train_compressor.py +76 -46
- sync_up.py +83 -15
START_HERE.md
CHANGED
|
@@ -4,6 +4,49 @@
|
|
| 4 |
> **统一模型**: Qwen3-VL-8B-Instruct
|
| 5 |
> **预计总时间**: 2-3 天
|
| 6 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
---
|
| 8 |
|
| 9 |
## 0. 项目结构
|
|
@@ -223,6 +266,42 @@ results/comparison/
|
|
| 223 |
└── qwen3_res_230400/
|
| 224 |
```
|
| 225 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 226 |
### 预期对比表
|
| 227 |
|
| 228 |
| 方法 | Tokens | CLIP (预期) | 延迟 | 显存 |
|
|
|
|
| 4 |
> **统一模型**: Qwen3-VL-8B-Instruct
|
| 5 |
> **预计总时间**: 2-3 天
|
| 6 |
|
| 7 |
+
## 当前任务状态（已按实际运行标记）
|
| 8 |
+
|
| 9 |
+
- [x] `1. 环境配置`：依赖与基础环境已可用。
|
| 10 |
+
- [x] `2. 数据准备`：训练与评估数据已就位并可被脚本读取。
|
| 11 |
+
- [x] `3.1 Smoke Test`：单卡冒烟已跑通。
|
| 12 |
+
- [x] `3.2 正式训练`：已完成（最新日志：`Epoch 4: avg_loss=0.2288`，并保存 `checkpoints/optical/epoch4.pt`）。
|
| 13 |
+
- [ ] `3.3 选择最佳 checkpoint`：待最终确认（当前已有 `epoch0/epoch1/epoch4/latest/best`，建议按最低 loss 更新 `best.pt`）。
|
| 14 |
+
- [x] `4.1 一键并行评估（50样本）`：已完成。
|
| 15 |
+
- [x] `4.2 逐个评估（已完成项）`：`baseline`、`visionzip-256`、`visionzip-128`、`efficientui(prune=0.6)`、`uipress-256`。
|
| 16 |
+
- [x] `4.2 逐个评估（补齐项）`：`resolution(230400/1003520)`、`visionzip-64`、`efficientui(prune=0.8)` 已完成。
|
| 17 |
+
- [x] `4.3 计算 CLIP 分数`：已完成（`results/benchmark/all_clip_scores.json`）。
|
| 18 |
+
- [x] `SSIM/bootstrap`：已完成（`results/benchmark/ssim_scores.json`、`results/benchmark/bootstrap_ci.json`）。
|
| 19 |
+
- [x] `element_analysis`：已完成（`results/element_analysis.json`）。
|
| 20 |
+
- [ ] `case study`：尚未执行。
|
| 21 |
+
|
| 22 |
+
### 执行约定（当前）
|
| 23 |
+
|
| 24 |
+
- 所有新任务统一使用 `nohup` 后台启动，并写入 `logs/*.nohup.log`。
|
| 25 |
+
|
| 26 |
+
### 现在可继续做（按优先级）
|
| 27 |
+
|
| 28 |
+
1) **最终 checkpoint 选择（推荐先做）**
|
| 29 |
+
```bash
|
| 30 |
+
for f in checkpoints/optical/epoch*.pt; do
|
| 31 |
+
echo -n "$f: "
|
| 32 |
+
python -c "import torch; c=torch.load('$f', map_location='cpu'); print(f'loss={c[\"loss\"]:.4f}')"
|
| 33 |
+
done
|
| 34 |
+
# 将最佳 epoch 覆盖为 best.pt
|
| 35 |
+
# cp checkpoints/optical/epochX.pt checkpoints/optical/best.pt
|
| 36 |
+
```
|
| 37 |
+
|
| 38 |
+
2) **补跑 case study（nohup）**
|
| 39 |
+
```bash
|
| 40 |
+
nohup bash -lc 'PYTHONPATH=. python scripts/step_case_study.py' \
|
| 41 |
+
> logs/case_study.nohup.log 2>&1 < /dev/null &
|
| 42 |
+
```
|
| 43 |
+
|
| 44 |
+
3) **可选：重跑失败样本渲染后再算 SSIM（例如 129/130）**
|
| 45 |
+
```bash
|
| 46 |
+
nohup bash -lc 'PYTHONPATH=. python scripts/step_ssim_bootstrap.py --benchmark_dir results/benchmark --ref_dir data/ref_screenshots' \
|
| 47 |
+
> logs/ssim_rerun.nohup.log 2>&1 < /dev/null &
|
| 48 |
+
```
|
| 49 |
+
|
| 50 |
---
|
| 51 |
|
| 52 |
## 0. 项目结构
|
|
|
|
| 266 |
└── qwen3_res_230400/
|
| 267 |
```
|
| 268 |
|
| 269 |
+
### 当前实测结果（50样本，已完成）
|
| 270 |
+
|
| 271 |
+
| 方法 | n_success | 平均视觉 tokens | 平均延迟 | 平均峰值显存 |
|
| 272 |
+
|:---|---:|---:|---:|---:|
|
| 273 |
+
| Qwen3-VL full | 50/50 | 7299.2 | 90.83s | 16.96GB |
|
| 274 |
+
| VisionZip-256 | 50/50 | 256.0 | 110.44s | 17.07GB |
|
| 275 |
+
| VisionZip-128 | 50/50 | 128.0 | 108.41s | 17.07GB |
|
| 276 |
+
| VisionZip-64 | 50/50 | 64.0 | 92.04s | 17.04GB |
|
| 277 |
+
| EfficientUI-60% | 50/50 | 729.9 | 99.52s | 17.02GB |
|
| 278 |
+
| EfficientUI-80% | 50/50 | 364.0 | 102.37s | 17.05GB |
|
| 279 |
+
| Qwen3-Res-230400 | 50/50 | 844.5 | 94.00s | 16.74GB |
|
| 280 |
+
| Qwen3-Res-1003520 | 50/50 | 3747.5 | 76.96s | 16.79GB |
|
| 281 |
+
| **UIPress-256** | **50/50** | **256.0** | **52.52s** | **17.31GB** |
|
| 282 |
+
|
| 283 |
+
### 当前 CLIP / SSIM（50样本）
|
| 284 |
+
|
| 285 |
+
| 方法 | CLIP | SSIM |
|
| 286 |
+
|:---|---:|---:|
|
| 287 |
+
| qwen3_res_230400 | 0.7768 | 0.6592 |
|
| 288 |
+
| qwen3_res_1003520 | 0.7750 | 0.6612 |
|
| 289 |
+
| qwen3_full | 0.7563 | 0.6647* |
|
| 290 |
+
| efficientui_prune60 | 0.7523 | 0.6487 |
|
| 291 |
+
| efficientui_prune80 | 0.7380 | 0.6232 |
|
| 292 |
+
| visionzip_256 | 0.7333 | 0.6489 |
|
| 293 |
+
| visionzip_128 | 0.7245 | 0.6461 |
|
| 294 |
+
| uipress_256 | 0.7232 | 0.6323 |
|
| 295 |
+
| visionzip_64 | 0.7197 | 0.6452 |
|
| 296 |
+
|
| 297 |
+
\* `qwen3_full` 与 `qwen3_res_1003520` 在 SSIM 渲染阶段各有少量超时样本（统计时已按可用样本数计算）。
|
| 298 |
+
|
| 299 |
+
### 当前还能马上做什么（全部可立即启动）
|
| 300 |
+
|
| 301 |
+
1. **恢复训练（优先）**：当前训练在 `E2 S4272/10000` 中断，可从 `checkpoints/optical/latest.pt` 继续。
|
| 302 |
+
2. **补齐剩余评估档位**：`resolution`、`visionzip-64`、`efficientui(prune=0.8)`。
|
| 303 |
+
3. **后处理与统计**：运行 `step_clip_batch.py`、`step_ssim_bootstrap.py`，然后做 `step_element_analysis.py` / `step_case_study.py`。
|
| 304 |
+
|
| 305 |
### 预期对比表
|
| 306 |
|
| 307 |
| 方法 | Tokens | CLIP (预期) | 延迟 | 显存 |
|
from
ADDED
|
File without changes
|
import
ADDED
|
File without changes
|
results/comparison/uipress_256/uipress_256/clip_scores.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"n": 50,
|
| 3 |
+
"avg_clip": 0.7232,
|
| 4 |
+
"min_clip": 0.4485,
|
| 5 |
+
"max_clip": 0.8768,
|
| 6 |
+
"per_sample": {
|
| 7 |
+
"0": 0.6477,
|
| 8 |
+
"1": 0.6416,
|
| 9 |
+
"10": 0.8694,
|
| 10 |
+
"100": 0.6019,
|
| 11 |
+
"101": 0.8444,
|
| 12 |
+
"102": 0.6862,
|
| 13 |
+
"103": 0.6421,
|
| 14 |
+
"104": 0.733,
|
| 15 |
+
"105": 0.7572,
|
| 16 |
+
"106": 0.653,
|
| 17 |
+
"107": 0.7705,
|
| 18 |
+
"108": 0.6961,
|
| 19 |
+
"109": 0.7902,
|
| 20 |
+
"11": 0.5583,
|
| 21 |
+
"110": 0.7974,
|
| 22 |
+
"111": 0.755,
|
| 23 |
+
"112": 0.8091,
|
| 24 |
+
"113": 0.6079,
|
| 25 |
+
"114": 0.6899,
|
| 26 |
+
"115": 0.8383,
|
| 27 |
+
"116": 0.7303,
|
| 28 |
+
"117": 0.6555,
|
| 29 |
+
"118": 0.6899,
|
| 30 |
+
"119": 0.715,
|
| 31 |
+
"12": 0.6144,
|
| 32 |
+
"120": 0.722,
|
| 33 |
+
"121": 0.8399,
|
| 34 |
+
"122": 0.4485,
|
| 35 |
+
"123": 0.7539,
|
| 36 |
+
"125": 0.6337,
|
| 37 |
+
"126": 0.7624,
|
| 38 |
+
"127": 0.8295,
|
| 39 |
+
"128": 0.7319,
|
| 40 |
+
"129": 0.834,
|
| 41 |
+
"13": 0.8209,
|
| 42 |
+
"130": 0.7926,
|
| 43 |
+
"131": 0.754,
|
| 44 |
+
"132": 0.8207,
|
| 45 |
+
"133": 0.8479,
|
| 46 |
+
"134": 0.6171,
|
| 47 |
+
"135": 0.8655,
|
| 48 |
+
"136": 0.5819,
|
| 49 |
+
"137": 0.7602,
|
| 50 |
+
"138": 0.4775,
|
| 51 |
+
"139": 0.8768,
|
| 52 |
+
"14": 0.7486,
|
| 53 |
+
"140": 0.821,
|
| 54 |
+
"141": 0.7551,
|
| 55 |
+
"142": 0.7435,
|
| 56 |
+
"143": 0.5266
|
| 57 |
+
}
|
| 58 |
+
}
|
results/comparison/uipress_256/uipress_256/html_predictions/0.html
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
<!DOCTYPE html>
|
| 2 |
+
<html>
|
| 3 |
+
<body>
|
| 4 |
+
<div class="container">
|
| 5 |
+
<div class="header">
|
| 6 |
+
<div class="logo">
|
| 7 |
+
<h1>WELCOME TO OUR WEBSITE</h1>
|
| 8 |
+
</div>
|
| 9 |
+
<nav class="navbar">
|
| 10 |
+
<ul>
|
| 11 |
+
<li><a href="#">Home</a></li>
|
| 12 |
+
<li><a href="#">About</a></li>
|
| 13 |
+
<li><a href="#">Services</a></li>
|
| 14 |
+
<li><a href="#">Contact</a></li>
|
| 15 |
+
</ul>
|
| 16 |
+
</div>
|
| 17 |
+
</div>
|
| 18 |
+
<div class="hero">
|
| 19 |
+
<h2>Our Mission</h2>
|
| 20 |
+
<p>At our company, we are dedicated to providing exceptional services and products to our valued customers.</p>
|
| 21 |
+
</div>
|
| 22 |
+
<div class="content">
|
| 23 |
+
<div class="section">
|
| 24 |
+
<h3>Our Services</h3>
|
| 25 |
+
<ul>
|
| 26 |
+
<li>Service 1</li>
|
| 27 |
+
<li>Service 2</li>
|
| 28 |
+
<li>Service 2</li>
|
| 29 |
+
</ul>
|
| 30 |
+
</div>
|
| 31 |
+
<div class="section">
|
| 32 |
+
<h3>Our Team</h3>
|
| 33 |
+
<div class="team-member">
|
| 34 |
+
<img src="team-member.jpg" alt="Team Member">
|
| 35 |
+
<h4>John Doe</h4>
|
| 36 |
+
<p>Team Leader</p>
|
| 37 |
+
</div>
|
| 38 |
+
<div class="team-member">
|
| 39 |
+
<img src="team-member.jpg" alt="Team Member">
|
| 40 |
+
<h4>Jane Smith</h4>
|
| 41 |
+
<p>Team Member</p>
|
| 42 |
+
</div>
|
| 43 |
+
</div>
|
| 44 |
+
</div>
|
| 45 |
+
<footer>
|
| 46 |
+
<p>© 2024 Our Company. All rights reserved.</p>
|
| 47 |
+
</footer>
|
| 48 |
+
</body>
|
| 49 |
+
</html>
|
results/comparison/uipress_256/uipress_256/html_predictions/10.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/100.html
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
<!DOCTYPE html>
|
| 2 |
+
<html>
|
| 3 |
+
<body>
|
| 4 |
+
<div class="container">
|
| 5 |
+
<div class="header">
|
| 6 |
+
<div class="logo">
|
| 7 |
+
<h1>WELCOME TO OUR WEBSITE</h1>
|
| 8 |
+
</div>
|
| 9 |
+
<div class="nav">
|
| 10 |
+
<a href="#">Home</a>
|
| 11 |
+
<a href="#">About</a>
|
| 12 |
+
<a href="#">Services</a>
|
| 13 |
+
<a href="#">Contact</a>
|
| 14 |
+
</div>
|
| 15 |
+
</div>
|
| 16 |
+
<div class="hero">
|
| 17 |
+
<h2>Our Services</h2>
|
| 18 |
+
<p>We offer a wide range of services to meet your needs.</p>
|
| 19 |
+
</div>
|
| 20 |
+
<div class="footer">
|
| 21 |
+
<p>© 2024 My Website. All rights reserved.</p>
|
| 22 |
+
</div>
|
| 23 |
+
</div>
|
| 24 |
+
</body>
|
| 25 |
+
</html>
|
results/comparison/uipress_256/uipress_256/html_predictions/101.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/103.html
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
<!DOCTYPE html>
|
| 2 |
+
<html>
|
| 3 |
+
<body>
|
| 4 |
+
<div class="container">
|
| 5 |
+
<div class="header">
|
| 6 |
+
<div class="logo">
|
| 7 |
+
<h1>WELCOME TO OUR WEBSITE</h1>
|
| 8 |
+
</div>
|
| 9 |
+
<div class="nav">
|
| 10 |
+
<a href="#">Home</a>
|
| 11 |
+
<a href="#">About</a>
|
| 12 |
+
<a href="#">Services</a>
|
| 13 |
+
<a href="#">Contact</a>
|
| 14 |
+
</div>
|
| 15 |
+
</div>
|
| 16 |
+
<div class="hero">
|
| 17 |
+
<h2>Our Services</h2>
|
| 18 |
+
<p>We offer a wide range of services to meet your needs.</p>
|
| 19 |
+
</div>
|
| 20 |
+
<div class="footer">
|
| 21 |
+
<p>© 2024 Our Website. All rights reserved.</p>
|
| 22 |
+
</div>
|
| 23 |
+
</div>
|
| 24 |
+
</body>
|
| 25 |
+
</html>
|
results/comparison/uipress_256/uipress_256/html_predictions/106.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/108.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/11.html
ADDED
|
@@ -0,0 +1,594 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<title>WELCOME TO THE OFFICIAL WEBSITE OF THE HOSPITAL OF THE CITY OF LONDON</title>
|
| 6 |
+
<style>
|
| 7 |
+
body {
|
| 8 |
+
margin: 0;
|
| 9 |
+
padding: 0;
|
| 10 |
+
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
|
| 11 |
+
background-color: #f5f5f;
|
| 12 |
+
overflow-x: hidden;
|
| 13 |
+
overflow-y: auto;
|
| 14 |
+
}
|
| 15 |
+
.container {
|
| 16 |
+
width: 100%;
|
| 17 |
+
height: 100%;
|
| 18 |
+
position: relative;
|
| 19 |
+
}
|
| 20 |
+
.header {
|
| 21 |
+
width: 0;
|
| 22 |
+
height: 100%;
|
| 23 |
+
position: absolute;
|
| 24 |
+
top: 0;
|
| 25 |
+
left: 0;
|
| 26 |
+
z-index: 100;
|
| 27 |
+
background-color: #000000;
|
| 28 |
+
color: #ffffff;
|
| 29 |
+
font-size: 16px;
|
| 30 |
+
padding: 0;
|
| 31 |
+
margin: 0;
|
| 32 |
+
overflow: hidden;
|
| 33 |
+
}
|
| 34 |
+
.header .logo {
|
| 35 |
+
width: 100%;
|
| 36 |
+
height: 100%;
|
| 37 |
+
position: absolute;
|
| 38 |
+
top: 0;
|
| 39 |
+
left: 0;
|
| 40 |
+
z-index: 100;
|
| 41 |
+
background-color: #000000;
|
| 42 |
+
color: #ffffff;
|
| 43 |
+
font-size: 16px;
|
| 44 |
+
padding: 0;
|
| 45 |
+
margin: 0;
|
| 46 |
+
overflow: hidden;
|
| 47 |
+
}
|
| 48 |
+
.header .logo .logo {
|
| 49 |
+
width: 100%;
|
| 50 |
+
height: 100%;
|
| 51 |
+
position: absolute;
|
| 52 |
+
top: 0;
|
| 53 |
+
left: 0;
|
| 54 |
+
z-index: 100;
|
| 55 |
+
background-color: #000000;
|
| 56 |
+
color: #ffffff;
|
| 57 |
+
font-size: 16px;
|
| 58 |
+
padding: 0;
|
| 59 |
+
margin: 0;
|
| 60 |
+
overflow: hidden;
|
| 61 |
+
}
|
| 62 |
+
.header .logo .logo img {
|
| 63 |
+
width: 100%;
|
| 64 |
+
height: 100%;
|
| 65 |
+
position: absolute;
|
| 66 |
+
top: 0;
|
| 67 |
+
left: 0;
|
| 68 |
+
z-index: 100;
|
| 69 |
+
background-color: #000000;
|
| 70 |
+
color: #ffffff;
|
| 71 |
+
font-size: 16px;
|
| 72 |
+
padding: 0;
|
| 73 |
+
margin: 0;
|
| 74 |
+
overflow: hidden;
|
| 75 |
+
}
|
| 76 |
+
.header .logo .logo img {
|
| 77 |
+
width: 100%;
|
| 78 |
+
height: 100%;
|
| 79 |
+
position: absolute;
|
| 80 |
+
top: 0;
|
| 81 |
+
left: 0;
|
| 82 |
+
z-index: 100;
|
| 83 |
+
background-color: #000000;
|
| 84 |
+
color: #ffffff;
|
| 85 |
+
font-size: 16px;
|
| 86 |
+
padding: 0;
|
| 87 |
+
margin: 0;
|
| 88 |
+
overflow: hidden;
|
| 89 |
+
}
|
| 90 |
+
.header .logo .logo img {
|
| 91 |
+
width: 100%;
|
| 92 |
+
height: 100%;
|
| 93 |
+
position: absolute;
|
| 94 |
+
top: 0;
|
| 95 |
+
left: 0;
|
| 96 |
+
z-index: 100;
|
| 97 |
+
background-color: #0000000;
|
| 98 |
+
color: #ffffff;
|
| 99 |
+
font-size: 16px;
|
| 100 |
+
padding: 0;
|
| 101 |
+
margin: 0;
|
| 102 |
+
overflow: hidden;
|
| 103 |
+
}
|
| 104 |
+
.header .logo .logo img {
|
| 105 |
+
width: 100%;
|
| 106 |
+
height: 100%;
|
| 107 |
+
position: absolute;
|
| 108 |
+
top: 0;
|
| 109 |
+
left: 0;
|
| 110 |
+
z-index: 100;
|
| 111 |
+
background-color: #000000;
|
| 112 |
+
color: #ffffff;
|
| 113 |
+
font-size: 16px;
|
| 114 |
+
padding: 0;
|
| 115 |
+
margin: 0;
|
| 116 |
+
overflow: hidden;
|
| 117 |
+
}
|
| 118 |
+
.header .logo .logo img {
|
| 119 |
+
width: 100%;
|
| 120 |
+
height: 100%;
|
| 121 |
+
position: absolute;
|
| 122 |
+
top: 0;
|
| 123 |
+
left: 0;
|
| 124 |
+
z-index: 100;
|
| 125 |
+
background-color: #000000;
|
| 126 |
+
color: #ffffff;
|
| 127 |
+
font-size: 16px;
|
| 128 |
+
padding: 0;
|
| 129 |
+
margin: 0;
|
| 130 |
+
overflow: hidden;
|
| 131 |
+
}
|
| 132 |
+
.header .logo .logo img {
|
| 133 |
+
width: 100%;
|
| 134 |
+
height: 100%;
|
| 135 |
+
position: absolute;
|
| 136 |
+
top: 0;
|
| 137 |
+
left: 0;
|
| 138 |
+
z-index: 100;
|
| 139 |
+
background-color: #000000;
|
| 140 |
+
color: #ffffff;
|
| 141 |
+
font-size: 16px;
|
| 142 |
+
padding: 0;
|
| 143 |
+
margin: 0;
|
| 144 |
+
overflow: hidden;
|
| 145 |
+
}
|
| 146 |
+
.header .logo .logo img {
|
| 147 |
+
width: 100%;
|
| 148 |
+
height: 100%;
|
| 149 |
+
position: absolute;
|
| 150 |
+
top: 0;
|
| 151 |
+
left: 0;
|
| 152 |
+
z-index: 100;
|
| 153 |
+
background-color: #000000;
|
| 154 |
+
color: #ffffff;
|
| 155 |
+
font-size: 16px;
|
| 156 |
+
padding: 0;
|
| 157 |
+
margin: 0;
|
| 158 |
+
overflow: hidden;
|
| 159 |
+
}
|
| 160 |
+
.header .logo .logo img {
|
| 161 |
+
width: 100%;
|
| 162 |
+
height: 100%;
|
| 163 |
+
position: absolute;
|
| 164 |
+
top: 0;
|
| 165 |
+
left: 0;
|
| 166 |
+
z-index: 100;
|
| 167 |
+
background-color: #000000;
|
| 168 |
+
color: #ffffff;
|
| 169 |
+
font-size: 16px;
|
| 170 |
+
padding: 0;
|
| 171 |
+
margin: 0;
|
| 172 |
+
overflow: hidden;
|
| 173 |
+
}
|
| 174 |
+
.header .logo .logo img {
|
| 175 |
+
width: 100%;
|
| 176 |
+
height: 100%;
|
| 177 |
+
position: absolute;
|
| 178 |
+
top: 0;
|
| 179 |
+
left: 0;
|
| 180 |
+
z-index: 100;
|
| 181 |
+
background-color: #000000;
|
| 182 |
+
color: #ffffff;
|
| 183 |
+
font-size: 16px;
|
| 184 |
+
padding: 0;
|
| 185 |
+
margin: 0;
|
| 186 |
+
overflow: hidden;
|
| 187 |
+
}
|
| 188 |
+
.header .logo .logo img {
|
| 189 |
+
width: 100%;
|
| 190 |
+
height: 100%;
|
| 191 |
+
position: absolute;
|
| 192 |
+
top: 0;
|
| 193 |
+
left: 0;
|
| 194 |
+
z-index: 100;
|
| 195 |
+
background-color: #000000;
|
| 196 |
+
color: #ffffff;
|
| 197 |
+
font-size: 16px;
|
| 198 |
+
padding: 0;
|
| 199 |
+
margin: 0;
|
| 200 |
+
overflow: hidden;
|
| 201 |
+
}
|
| 202 |
+
.header .logo .logo img {
|
| 203 |
+
width: 100%;
|
| 204 |
+
height: 100%;
|
| 205 |
+
position: absolute;
|
| 206 |
+
top: 0;
|
| 207 |
+
left: 0;
|
| 208 |
+
z-index: 100;
|
| 209 |
+
background-color: #000000;
|
| 210 |
+
color: #ffffff;
|
| 211 |
+
font-size: 16px;
|
| 212 |
+
padding: 0;
|
| 213 |
+
margin: 0;
|
| 214 |
+
overflow: hidden;
|
| 215 |
+
}
|
| 216 |
+
.header .logo .logo img {
|
| 217 |
+
width: 100%;
|
| 218 |
+
height: 100%;
|
| 219 |
+
position: absolute;
|
| 220 |
+
top: 0;
|
| 221 |
+
left: 0;
|
| 222 |
+
z-index: 100;
|
| 223 |
+
background-color: #000000;
|
| 224 |
+
color: #ffffff;
|
| 225 |
+
font-size: 16px;
|
| 226 |
+
padding: 0;
|
| 227 |
+
margin: 0;
|
| 228 |
+
overflow: hidden;
|
| 229 |
+
}
|
| 230 |
+
.header .logo .logo img {
|
| 231 |
+
width: 100%;
|
| 232 |
+
height: 100%;
|
| 233 |
+
position: absolute;
|
| 234 |
+
top: 0;
|
| 235 |
+
left: 0;
|
| 236 |
+
z-index: 100;
|
| 237 |
+
background-color: #000000;
|
| 238 |
+
color: #ffffff;
|
| 239 |
+
font-size: 16px;
|
| 240 |
+
padding: 0;
|
| 241 |
+
margin: 0;
|
| 242 |
+
overflow: hidden;
|
| 243 |
+
}
|
| 244 |
+
.header .logo .logo img {
|
| 245 |
+
width: 100%;
|
| 246 |
+
height: 100%;
|
| 247 |
+
position: absolute;
|
| 248 |
+
top: 0;
|
| 249 |
+
left: 0;
|
| 250 |
+
z-index: 100;
|
| 251 |
+
background-color: #000000;
|
| 252 |
+
color: #ffffff;
|
| 253 |
+
font-size: 16px;
|
| 254 |
+
padding: 0;
|
| 255 |
+
margin: 0;
|
| 256 |
+
overflow: hidden;
|
| 257 |
+
}
|
| 258 |
+
.header .logo .logo img {
|
| 259 |
+
width: 100%;
|
| 260 |
+
height: 100%;
|
| 261 |
+
position: absolute;
|
| 262 |
+
top: 0;
|
| 263 |
+
left: 0;
|
| 264 |
+
z-index: 100;
|
| 265 |
+
background-color: #000000;
|
| 266 |
+
color: #ffffff;
|
| 267 |
+
font-size: 16px;
|
| 268 |
+
padding: 0;
|
| 269 |
+
margin: 0;
|
| 270 |
+
overflow: hidden;
|
| 271 |
+
}
|
| 272 |
+
.header .logo .logo img {
|
| 273 |
+
width: 100%;
|
| 274 |
+
height: 100%;
|
| 275 |
+
position: absolute;
|
| 276 |
+
top: 0;
|
| 277 |
+
left: 0;
|
| 278 |
+
z-index: 100;
|
| 279 |
+
background-color: #000000;
|
| 280 |
+
color: #ffffff;
|
| 281 |
+
font-size: 16px;
|
| 282 |
+
padding: 0;
|
| 283 |
+
margin: 0;
|
| 284 |
+
overflow: hidden;
|
| 285 |
+
}
|
| 286 |
+
.header .logo .logo img {
|
| 287 |
+
width: 100%;
|
| 288 |
+
height: 100%;
|
| 289 |
+
position: absolute;
|
| 290 |
+
top: 0;
|
| 291 |
+
left: 0;
|
| 292 |
+
z-index: 100;
|
| 293 |
+
background-color: #000000;
|
| 294 |
+
color: #ffffff;
|
| 295 |
+
font-size: 16px;
|
| 296 |
+
padding: 0;
|
| 297 |
+
margin: 0;
|
| 298 |
+
overflow: hidden;
|
| 299 |
+
}
|
| 300 |
+
.header .logo .logo img {
|
| 301 |
+
width: 100%;
|
| 302 |
+
height: 100%;
|
| 303 |
+
position: absolute;
|
| 304 |
+
top: 0;
|
| 305 |
+
left: 0;
|
| 306 |
+
z-index: 100;
|
| 307 |
+
background-color: #000000;
|
| 308 |
+
color: #ffffff;
|
| 309 |
+
font-size: 16px;
|
| 310 |
+
padding: 0;
|
| 311 |
+
margin: 0;
|
| 312 |
+
overflow: hidden;
|
| 313 |
+
}
|
| 314 |
+
.header .logo .logo img {
|
| 315 |
+
width: 100%;
|
| 316 |
+
height: 100%;
|
| 317 |
+
position: absolute;
|
| 318 |
+
top: 0;
|
| 319 |
+
left: 0;
|
| 320 |
+
z-index: 100;
|
| 321 |
+
background-color: #000000;
|
| 322 |
+
color: #ffffff;
|
| 323 |
+
font-size: 16px;
|
| 324 |
+
padding: 0;
|
| 325 |
+
margin: 0;
|
| 326 |
+
overflow: hidden;
|
| 327 |
+
}
|
| 328 |
+
.header .logo .logo img {
|
| 329 |
+
width: 100%;
|
| 330 |
+
height: 100%;
|
| 331 |
+
position: absolute;
|
| 332 |
+
top: 0;
|
| 333 |
+
left: 0;
|
| 334 |
+
z-index: 100;
|
| 335 |
+
background-color: #000000;
|
| 336 |
+
color: #ffffff;
|
| 337 |
+
font-size: 16px;
|
| 338 |
+
padding: 0;
|
| 339 |
+
margin: 0;
|
| 340 |
+
overflow: hidden;
|
| 341 |
+
}
|
| 342 |
+
.header .logo .logo img {
|
| 343 |
+
width: 100%;
|
| 344 |
+
height: 100%;
|
| 345 |
+
position: absolute;
|
| 346 |
+
top: 0;
|
| 347 |
+
left: 0;
|
| 348 |
+
z-index: 100;
|
| 349 |
+
background-color: #000000;
|
| 350 |
+
color: #ffffff;
|
| 351 |
+
font-size: 16px;
|
| 352 |
+
padding: 0;
|
| 353 |
+
margin: 0;
|
| 354 |
+
overflow: hidden;
|
| 355 |
+
}
|
| 356 |
+
.header .logo .logo img {
|
| 357 |
+
width: 100%;
|
| 358 |
+
height: 100%;
|
| 359 |
+
position: absolute;
|
| 360 |
+
top: 0;
|
| 361 |
+
left: 0;
|
| 362 |
+
z-index: 100;
|
| 363 |
+
background-color: #000000;
|
| 364 |
+
color: #ffffff;
|
| 365 |
+
font-size: 16px;
|
| 366 |
+
padding: 0;
|
| 367 |
+
margin: 0;
|
| 368 |
+
overflow: hidden;
|
| 369 |
+
}
|
| 370 |
+
.header .logo .logo img {
|
| 371 |
+
width: 100%;
|
| 372 |
+
height: 100%;
|
| 373 |
+
position: absolute;
|
| 374 |
+
top: 0;
|
| 375 |
+
left: 0;
|
| 376 |
+
z-index: 100;
|
| 377 |
+
background-color: #000000;
|
| 378 |
+
color: #ffffff;
|
| 379 |
+
font-size: 16px;
|
| 380 |
+
padding: 0;
|
| 381 |
+
margin: 0;
|
| 382 |
+
overflow: hidden;
|
| 383 |
+
}
|
| 384 |
+
.header .logo .logo img {
|
| 385 |
+
width: 100%;
|
| 386 |
+
height: 100%;
|
| 387 |
+
position: absolute;
|
| 388 |
+
top: 0;
|
| 389 |
+
left: 0;
|
| 390 |
+
z-index: 100;
|
| 391 |
+
background-color: #000000;
|
| 392 |
+
color: #ffffff;
|
| 393 |
+
font-size: 16px;
|
| 394 |
+
padding: 0;
|
| 395 |
+
margin: 0;
|
| 396 |
+
overflow: hidden;
|
| 397 |
+
}
|
| 398 |
+
.header .logo .logo img {
|
| 399 |
+
width: 100%;
|
| 400 |
+
height: 100%;
|
| 401 |
+
position: absolute;
|
| 402 |
+
top: 0;
|
| 403 |
+
left: 0;
|
| 404 |
+
z-index: 100;
|
| 405 |
+
background-color: #000000;
|
| 406 |
+
color: #ffffff;
|
| 407 |
+
font-size: 16px;
|
| 408 |
+
padding: 0;
|
| 409 |
+
margin: 0;
|
| 410 |
+
overflow: hidden;
|
| 411 |
+
}
|
| 412 |
+
.header .logo .logo img {
|
| 413 |
+
width: 100%;
|
| 414 |
+
height: 100%;
|
| 415 |
+
position: absolute;
|
| 416 |
+
top: 0;
|
| 417 |
+
left: 0;
|
| 418 |
+
z-index: 100;
|
| 419 |
+
background-color: #000000;
|
| 420 |
+
color: #ffffff;
|
| 421 |
+
font-size: 16px;
|
| 422 |
+
padding: 0;
|
| 423 |
+
margin: 0;
|
| 424 |
+
overflow: hidden;
|
| 425 |
+
}
|
| 426 |
+
.header .logo .logo img {
|
| 427 |
+
width: 100%;
|
| 428 |
+
height: 100%;
|
| 429 |
+
position: absolute;
|
| 430 |
+
top: 0;
|
| 431 |
+
left: 0;
|
| 432 |
+
z-index: 100;
|
| 433 |
+
background-color: #000000;
|
| 434 |
+
color: #ffffff;
|
| 435 |
+
font-size: 16px;
|
| 436 |
+
padding: 0;
|
| 437 |
+
margin: 0;
|
| 438 |
+
overflow: hidden;
|
| 439 |
+
}
|
| 440 |
+
.header .logo .logo img {
|
| 441 |
+
width: 100%;
|
| 442 |
+
height: 100%;
|
| 443 |
+
position: absolute;
|
| 444 |
+
top: 0;
|
| 445 |
+
left: 0;
|
| 446 |
+
z-index: 100;
|
| 447 |
+
background-color: #000000;
|
| 448 |
+
color: #ffffff;
|
| 449 |
+
font-size: 16px;
|
| 450 |
+
padding: 0;
|
| 451 |
+
margin: 0;
|
| 452 |
+
overflow: hidden;
|
| 453 |
+
}
|
| 454 |
+
.header .logo .logo img {
|
| 455 |
+
width: 100%;
|
| 456 |
+
height: 100%;
|
| 457 |
+
position: absolute;
|
| 458 |
+
top: 0;
|
| 459 |
+
left: 0;
|
| 460 |
+
z-index: 100;
|
| 461 |
+
background-color: #000000;
|
| 462 |
+
color: #ffffff;
|
| 463 |
+
font-size: 16px;
|
| 464 |
+
padding: 0;
|
| 465 |
+
margin: 0;
|
| 466 |
+
overflow: hidden;
|
| 467 |
+
}
|
| 468 |
+
.header .logo .logo img {
|
| 469 |
+
width: 100%;
|
| 470 |
+
height: 100%;
|
| 471 |
+
position: absolute;
|
| 472 |
+
top: 0;
|
| 473 |
+
left: 0;
|
| 474 |
+
z-index: 100;
|
| 475 |
+
background-color: #000000;
|
| 476 |
+
color: #ffffff;
|
| 477 |
+
font-size: 16px;
|
| 478 |
+
padding: 0;
|
| 479 |
+
margin: 0;
|
| 480 |
+
overflow: hidden;
|
| 481 |
+
}
|
| 482 |
+
.header .logo .logo img {
|
| 483 |
+
width: 100%;
|
| 484 |
+
height: 100%;
|
| 485 |
+
position: absolute;
|
| 486 |
+
top: 0;
|
| 487 |
+
left: 0;
|
| 488 |
+
z-index: 100;
|
| 489 |
+
background-color: #000000;
|
| 490 |
+
color: #ffffff;
|
| 491 |
+
font-size: 16px;
|
| 492 |
+
padding: 0;
|
| 493 |
+
margin: 0;
|
| 494 |
+
overflow: hidden;
|
| 495 |
+
}
|
| 496 |
+
.header .logo .logo img {
|
| 497 |
+
width: 100%;
|
| 498 |
+
height: 100%;
|
| 499 |
+
position: absolute;
|
| 500 |
+
top: 0;
|
| 501 |
+
left: 0;
|
| 502 |
+
z-index: 100;
|
| 503 |
+
background-color: #000000;
|
| 504 |
+
color: #ffffff;
|
| 505 |
+
font-size: 16px;
|
| 506 |
+
padding: 0;
|
| 507 |
+
margin: 0;
|
| 508 |
+
overflow: hidden;
|
| 509 |
+
}
|
| 510 |
+
.header .logo .logo img {
|
| 511 |
+
width: 100%;
|
| 512 |
+
height: 100%;
|
| 513 |
+
position: absolute;
|
| 514 |
+
top: 0;
|
| 515 |
+
left: 0;
|
| 516 |
+
z-index: 100;
|
| 517 |
+
background-color: #000000;
|
| 518 |
+
color: #ffffff;
|
| 519 |
+
font-size: 16px;
|
| 520 |
+
padding: 0;
|
| 521 |
+
margin: 0;
|
| 522 |
+
overflow: hidden;
|
| 523 |
+
}
|
| 524 |
+
.header .logo .logo img {
|
| 525 |
+
width: 100%;
|
| 526 |
+
height: 100%;
|
| 527 |
+
position: absolute;
|
| 528 |
+
top: 0;
|
| 529 |
+
left: 0;
|
| 530 |
+
z-index: 100;
|
| 531 |
+
background-color: #000000;
|
| 532 |
+
color: #ffffff;
|
| 533 |
+
font-size: 16px;
|
| 534 |
+
padding: 0;
|
| 535 |
+
margin: 0;
|
| 536 |
+
overflow: hidden;
|
| 537 |
+
}
|
| 538 |
+
.header .logo .logo img {
|
| 539 |
+
width: 100%;
|
| 540 |
+
height: 100%;
|
| 541 |
+
position: absolute;
|
| 542 |
+
top: 0;
|
| 543 |
+
left: 0;
|
| 544 |
+
z-index: 100;
|
| 545 |
+
background-color: #000000;
|
| 546 |
+
color: #ffffff;
|
| 547 |
+
font-size: 16px;
|
| 548 |
+
padding: 0;
|
| 549 |
+
margin: 0;
|
| 550 |
+
overflow: hidden;
|
| 551 |
+
}
|
| 552 |
+
.header .logo .logo img {
|
| 553 |
+
width: 100%;
|
| 554 |
+
height: 100%;
|
| 555 |
+
position: absolute;
|
| 556 |
+
top: 0;
|
| 557 |
+
left: 0;
|
| 558 |
+
z-index: 100;
|
| 559 |
+
background-color: #000000;
|
| 560 |
+
color: #ffffff;
|
| 561 |
+
font-size: 16px;
|
| 562 |
+
padding: 0;
|
| 563 |
+
margin: 0;
|
| 564 |
+
overflow: hidden;
|
| 565 |
+
}
|
| 566 |
+
.header .logo .logo img {
|
| 567 |
+
width: 100%;
|
| 568 |
+
height: 100%;
|
| 569 |
+
position: absolute;
|
| 570 |
+
top: 0;
|
| 571 |
+
left: 0;
|
| 572 |
+
z-index: 100;
|
| 573 |
+
background-color: #000000;
|
| 574 |
+
color: #ffffff;
|
| 575 |
+
font-size: 16px;
|
| 576 |
+
padding: 0;
|
| 577 |
+
margin: 0;
|
| 578 |
+
overflow: hidden;
|
| 579 |
+
}
|
| 580 |
+
.header .logo .logo img {
|
| 581 |
+
width: 100%;
|
| 582 |
+
height: 100%;
|
| 583 |
+
position: absolute;
|
| 584 |
+
top: 0;
|
| 585 |
+
left: 0;
|
| 586 |
+
z-index: 100;
|
| 587 |
+
background-color: #000000;
|
| 588 |
+
color: #ffffff;
|
| 589 |
+
font-size: 16px;
|
| 590 |
+
padding: 0;
|
| 591 |
+
margin: 0;
|
| 592 |
+
overflow: hidden;
|
| 593 |
+
}
|
| 594 |
+
.
|
results/comparison/uipress_256/uipress_256/html_predictions/115.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/116.html
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<title>WELCOME TO THE OFFICIAL WEBSITE</title>
|
| 6 |
+
<style>
|
| 7 |
+
body {
|
| 8 |
+
margin: 0;
|
| 9 |
+
padding: 0;
|
| 10 |
+
font-family: 'Segoe UI', Tahoma, sans-serif;
|
| 11 |
+
background: #f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f5f
|
results/comparison/uipress_256/uipress_256/html_predictions/119.html
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<body>
|
| 5 |
+
<div class="container">
|
| 6 |
+
<div class="header">
|
| 7 |
+
<div class="logo">
|
| 8 |
+
<img src="https://www.12345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345
|
results/comparison/uipress_256/uipress_256/html_predictions/121.html
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<title>Our Story</title>
|
| 6 |
+
<style>
|
| 7 |
+
body {
|
| 8 |
+
font-family: Arial, sans-serif;
|
| 9 |
+
margin: 0;
|
| 10 |
+
padding:
|
results/comparison/uipress_256/uipress_256/html_predictions/123.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/125.html
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
<!DOCTYPE html>
|
| 2 |
+
<html>
|
| 3 |
+
<body>
|
| 4 |
+
<div style="container">
|
| 5 |
+
<div style="header">
|
| 6 |
+
<h1>WELCOME TO THE OFFICIAL WEBSITE OF THE HOSPITAL</h1>
|
| 7 |
+
</div>
|
| 8 |
+
<div style="content">
|
| 9 |
+
<p style="text-align: center;">
|
| 10 |
+
<img src="https://via.placeholder.com/800x150" alt="Hospital Logo" />
|
| 11 |
+
<h2>Our Mission</h2>
|
| 12 |
+
<p style="text-align: justify;">
|
| 13 |
+
To provide high-quality, affordable, and accessible healthcare services to the community, with a focus on patient-centered care and continuous improvement.
|
| 14 |
+
</p>
|
| 15 |
+
<h2>Our Vision</h2>
|
| 16 |
+
<p style="text-align: justify;">
|
| 17 |
+
To be the leading healthcare provider in our region, recognized for excellence in patient care, innovation, and community impact.
|
| 18 |
+
</p>
|
| 19 |
+
<h2>Our Values</h2>
|
| 20 |
+
<ul style="list-style-type: square; padding-left: 20px;">
|
| 21 |
+
<li>Compassion</li>
|
| 22 |
+
<li>Integrity</li>
|
| 23 |
+
<li>Excellence</li>
|
| 24 |
+
<li>Respect</li>
|
| 25 |
+
<li>Empathy</li>
|
| 26 |
+
</ul>
|
| 27 |
+
</div>
|
| 28 |
+
<div style="footer">
|
| 29 |
+
<p style="text-align: center;">
|
| 30 |
+
© 2024 | All Rights Reserved
|
| 31 |
+
</p>
|
| 32 |
+
</div>
|
| 33 |
+
</div>
|
| 34 |
+
</body>
|
| 35 |
+
</html>
|
results/comparison/uipress_256/uipress_256/html_predictions/126.html
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<title>WELCOME TO THE OFFICIAL WEBSITE</title>
|
| 6 |
+
<style>
|
| 7 |
+
body {
|
| 8 |
+
margin: 0;
|
| 9 |
+
padding:
|
results/comparison/uipress_256/uipress_256/html_predictions/131.html
ADDED
|
@@ -0,0 +1,558 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<title>WELCOME TO THE OFFICIAL WEBSITE OF THE FASHION DESIGNER</title>
|
| 6 |
+
<style>
|
| 7 |
+
body {
|
| 8 |
+
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
|
| 9 |
+
margin: 0;
|
| 10 |
+
padding: 0;
|
| 11 |
+
background-color: #f5f5f;
|
| 12 |
+
overflow-x: hidden;
|
| 13 |
+
overflow-y: scroll;
|
| 14 |
+
}
|
| 15 |
+
.header {
|
| 16 |
+
position: fixed;
|
| 17 |
+
top: 0;
|
| 18 |
+
left: 0;
|
| 19 |
+
width: 100%;
|
| 20 |
+
height: 100vh;
|
| 21 |
+
background-color: #000000;
|
| 22 |
+
z-index: 999;
|
| 23 |
+
display: flex;
|
| 24 |
+
flex-direction: column;
|
| 25 |
+
justify-content: space-between;
|
| 26 |
+
align-items: center;
|
| 27 |
+
padding: 10px 0;
|
| 28 |
+
box-sizing: border-box;
|
| 29 |
+
}
|
| 30 |
+
.header .logo {
|
| 31 |
+
width: 100%;
|
| 32 |
+
height: 100px;
|
| 33 |
+
display: flex;
|
| 34 |
+
justify-content: center;
|
| 35 |
+
align-items: center;
|
| 36 |
+
background-color: #000000;
|
| 37 |
+
color: #ffffff;
|
| 38 |
+
font-size: 24px;
|
| 39 |
+
font-weight: bold;
|
| 40 |
+
text-transform: uppercase;
|
| 41 |
+
letter-spacing: 2px;
|
| 42 |
+
text-align: center;
|
| 43 |
+
padding: 10px 0;
|
| 44 |
+
}
|
| 45 |
+
.header .nav {
|
| 46 |
+
width: 100%;
|
| 47 |
+
height: 60px;
|
| 48 |
+
display: flex;
|
| 49 |
+
justify-content: space-between;
|
| 50 |
+
align-items: center;
|
| 51 |
+
padding: 0 10px;
|
| 52 |
+
box-sizing: border-box;
|
| 53 |
+
}
|
| 54 |
+
.header .nav a {
|
| 55 |
+
color: #ffffff;
|
| 56 |
+
text-decoration: none;
|
| 57 |
+
font-size: 16px;
|
| 58 |
+
font-weight: bold;
|
| 59 |
+
padding: 0 10px;
|
| 60 |
+
transition: all 0.3s;
|
| 61 |
+
}
|
| 62 |
+
.header .nav a:hover {
|
| 63 |
+
color: #ffcc00;
|
| 64 |
+
transform: scale(1.05);
|
| 65 |
+
}
|
| 66 |
+
.header .nav .logo {
|
| 67 |
+
width: 100%;
|
| 68 |
+
height: 100%;
|
| 69 |
+
display: flex;
|
| 70 |
+
justify-content: center;
|
| 71 |
+
align-items: center;
|
| 72 |
+
background-color: #000000;
|
| 73 |
+
color: #ffffff;
|
| 74 |
+
font-size: 24px;
|
| 75 |
+
font-weight: bold;
|
| 76 |
+
text-transform: uppercase;
|
| 77 |
+
letter-spacing: 2px;
|
| 78 |
+
text-align: center;
|
| 79 |
+
padding: 10px 0;
|
| 80 |
+
}
|
| 81 |
+
.header .nav .logo a {
|
| 82 |
+
color: #ffffff;
|
| 83 |
+
text-decoration: none;
|
| 84 |
+
font-size: 16px;
|
| 85 |
+
font-weight: bold;
|
| 86 |
+
padding: 0 10px;
|
| 87 |
+
transition: all 0.3s;
|
| 88 |
+
}
|
| 89 |
+
.header .nav .logo a:hover {
|
| 90 |
+
color: #ffcc00;
|
| 91 |
+
transform: scale(1.05);
|
| 92 |
+
}
|
| 93 |
+
.header .nav .logo a {
|
| 94 |
+
color: #ffffff;
|
| 95 |
+
text-decoration: none;
|
| 96 |
+
font-size: 16px;
|
| 97 |
+
font-weight: bold;
|
| 98 |
+
padding: 0 10px;
|
| 99 |
+
transition: all 0.3s;
|
| 100 |
+
}
|
| 101 |
+
.header .nav .logo a:hover {
|
| 102 |
+
color: #ffcc00;
|
| 103 |
+
transform: scale(1.05);
|
| 104 |
+
}
|
| 105 |
+
.header .nav .logo a {
|
| 106 |
+
color: #ffffff;
|
| 107 |
+
text-decoration: none;
|
| 108 |
+
font-size: 16px;
|
| 109 |
+
font-weight: bold;
|
| 110 |
+
padding: 0 10px;
|
| 111 |
+
transition: all 0.3s;
|
| 112 |
+
}
|
| 113 |
+
.header .nav .logo a:hover {
|
| 114 |
+
color: #ffcc00;
|
| 115 |
+
transform: scale(1.05);
|
| 116 |
+
}
|
| 117 |
+
.header .nav .logo a {
|
| 118 |
+
color: #ffffff;
|
| 119 |
+
text-decoration: none;
|
| 120 |
+
font-size: 16px;
|
| 121 |
+
font-weight: bold;
|
| 122 |
+
padding: 0 10px;
|
| 123 |
+
transition: all 0.3s;
|
| 124 |
+
}
|
| 125 |
+
.header .nav .logo a:hover {
|
| 126 |
+
color: #ffcc00;
|
| 127 |
+
transform: scale(1.05);
|
| 128 |
+
}
|
| 129 |
+
.header .nav .logo a {
|
| 130 |
+
color: #ffffff;
|
| 131 |
+
text-decoration: none;
|
| 132 |
+
font-size: 16px;
|
| 133 |
+
font-weight: bold;
|
| 134 |
+
padding: 0 10px;
|
| 135 |
+
transition: all 0.3s;
|
| 136 |
+
}
|
| 137 |
+
.header .nav .logo a:hover {
|
| 138 |
+
color: #ffcc00;
|
| 139 |
+
transform: scale(1.05);
|
| 140 |
+
}
|
| 141 |
+
.header .nav .logo a {
|
| 142 |
+
color: #ffffff;
|
| 143 |
+
text-decoration: none;
|
| 144 |
+
font-size: 16px;
|
| 145 |
+
font-weight: bold;
|
| 146 |
+
padding: 0 10px;
|
| 147 |
+
transition: all 0.3s;
|
| 148 |
+
}
|
| 149 |
+
.header .nav .logo a:hover {
|
| 150 |
+
color: #ffcc00;
|
| 151 |
+
transform: scale(1.05);
|
| 152 |
+
}
|
| 153 |
+
.header .nav .logo a {
|
| 154 |
+
color: #ffffff;
|
| 155 |
+
text-decoration: none;
|
| 156 |
+
font-size: 16px;
|
| 157 |
+
font-weight: bold;
|
| 158 |
+
padding: 0 10px;
|
| 159 |
+
transition: all 0.3s;
|
| 160 |
+
}
|
| 161 |
+
.header .nav .logo a:hover {
|
| 162 |
+
color: #ffcc00;
|
| 163 |
+
transform: scale(1.05);
|
| 164 |
+
}
|
| 165 |
+
.header .nav .logo a {
|
| 166 |
+
color: #ffffff;
|
| 167 |
+
text-decoration: none;
|
| 168 |
+
font-size: 16px;
|
| 169 |
+
font-weight: bold;
|
| 170 |
+
padding: 0 10px;
|
| 171 |
+
transition: all 0.3s;
|
| 172 |
+
}
|
| 173 |
+
.header .nav .logo a:hover {
|
| 174 |
+
color: #ffcc00;
|
| 175 |
+
transform: scale(1.05);
|
| 176 |
+
}
|
| 177 |
+
.header .nav .logo a {
|
| 178 |
+
color: #ffffff;
|
| 179 |
+
text-decoration: none;
|
| 180 |
+
font-size: 16px;
|
| 181 |
+
font-weight: bold;
|
| 182 |
+
padding: 0 10px;
|
| 183 |
+
transition: all 0.3s;
|
| 184 |
+
}
|
| 185 |
+
.header .nav .logo a:hover {
|
| 186 |
+
color: #ffcc00;
|
| 187 |
+
transform: scale(1.05);
|
| 188 |
+
}
|
| 189 |
+
.header .nav .logo a {
|
| 190 |
+
color: #ffffff;
|
| 191 |
+
text-decoration: none;
|
| 192 |
+
font-size: 16px;
|
| 193 |
+
font-weight: bold;
|
| 194 |
+
padding: 0 10px;
|
| 195 |
+
transition: all 0.3s;
|
| 196 |
+
}
|
| 197 |
+
.header .nav .logo a:hover {
|
| 198 |
+
color: #ffcc00;
|
| 199 |
+
transform: scale(1.05);
|
| 200 |
+
}
|
| 201 |
+
.header .nav .logo a {
|
| 202 |
+
color: #ffffff;
|
| 203 |
+
text-decoration: none;
|
| 204 |
+
font-size: 16px;
|
| 205 |
+
font-weight: bold;
|
| 206 |
+
padding: 0 10px;
|
| 207 |
+
transition: all 0.3s;
|
| 208 |
+
}
|
| 209 |
+
.header .nav .logo a:hover {
|
| 210 |
+
color: #ffcc00;
|
| 211 |
+
transform: scale(1.05);
|
| 212 |
+
}
|
| 213 |
+
.header .nav .logo a {
|
| 214 |
+
color: #ffffff;
|
| 215 |
+
text-decoration: none;
|
| 216 |
+
font-size: 16px;
|
| 217 |
+
font-weight: bold;
|
| 218 |
+
padding: 0 10px;
|
| 219 |
+
transition: all 0.3s;
|
| 220 |
+
}
|
| 221 |
+
.header .nav .logo a:hover {
|
| 222 |
+
color: #ffcc00;
|
| 223 |
+
transform: scale(1.05);
|
| 224 |
+
}
|
| 225 |
+
.header .nav .logo a {
|
| 226 |
+
color: #ffffff;
|
| 227 |
+
text-decoration: none;
|
| 228 |
+
font-size: 16px;
|
| 229 |
+
font-weight: bold;
|
| 230 |
+
padding: 0 10px;
|
| 231 |
+
transition: all 0.3s;
|
| 232 |
+
}
|
| 233 |
+
.header .nav .logo a:hover {
|
| 234 |
+
color: #ffcc00;
|
| 235 |
+
transform: scale(1.05);
|
| 236 |
+
}
|
| 237 |
+
.header .nav .logo a {
|
| 238 |
+
color: #ffffff;
|
| 239 |
+
text-decoration: none;
|
| 240 |
+
font-size: 16px;
|
| 241 |
+
font-weight: bold;
|
| 242 |
+
padding: 0 10px;
|
| 243 |
+
transition: all 0.3s;
|
| 244 |
+
}
|
| 245 |
+
.header .nav .logo a:hover {
|
| 246 |
+
color: #ffcc00;
|
| 247 |
+
transform: scale(1.05);
|
| 248 |
+
}
|
| 249 |
+
.header .nav .logo a {
|
| 250 |
+
color: #ffffff;
|
| 251 |
+
text-decoration: none;
|
| 252 |
+
font-size: 16px;
|
| 253 |
+
font-weight: bold;
|
| 254 |
+
padding: 0 10px;
|
| 255 |
+
transition: all 0.3s;
|
| 256 |
+
}
|
| 257 |
+
.header .nav .logo a:hover {
|
| 258 |
+
color: #ffcc00;
|
| 259 |
+
transform: scale(1.05);
|
| 260 |
+
}
|
| 261 |
+
.header .nav .logo a {
|
| 262 |
+
color: #ffffff;
|
| 263 |
+
text-decoration: none;
|
| 264 |
+
font-size: 16px;
|
| 265 |
+
font-weight: bold;
|
| 266 |
+
padding: 0 10px;
|
| 267 |
+
transition: all 0.3s;
|
| 268 |
+
}
|
| 269 |
+
.header .nav .logo a:hover {
|
| 270 |
+
color: #ffcc00;
|
| 271 |
+
transform: scale(1.05);
|
| 272 |
+
}
|
| 273 |
+
.header .nav .logo a {
|
| 274 |
+
color: #ffffff;
|
| 275 |
+
text-decoration: none;
|
| 276 |
+
font-size: 16px;
|
| 277 |
+
font-weight: bold;
|
| 278 |
+
padding: 0 10px;
|
| 279 |
+
transition: all 0.3s;
|
| 280 |
+
}
|
| 281 |
+
.header .nav .logo a:hover {
|
| 282 |
+
color: #ffcc00;
|
| 283 |
+
transform: scale(1.05);
|
| 284 |
+
}
|
| 285 |
+
.header .nav .logo a {
|
| 286 |
+
color: #ffffff;
|
| 287 |
+
text-decoration: none;
|
| 288 |
+
font-size: 16px;
|
| 289 |
+
font-weight: bold;
|
| 290 |
+
padding: 0 10px;
|
| 291 |
+
transition: all 0.3s;
|
| 292 |
+
}
|
| 293 |
+
.header .nav .logo a:hover {
|
| 294 |
+
color: #ffcc00;
|
| 295 |
+
transform: scale(1.05);
|
| 296 |
+
}
|
| 297 |
+
.header .nav .logo a {
|
| 298 |
+
color: #ffffff;
|
| 299 |
+
text-decoration: none;
|
| 300 |
+
font-size: 16px;
|
| 301 |
+
font-weight: bold;
|
| 302 |
+
padding: 0 10px;
|
| 303 |
+
transition: all 0.3s;
|
| 304 |
+
}
|
| 305 |
+
.header .nav .logo a:hover {
|
| 306 |
+
color: #ffcc00;
|
| 307 |
+
transform: scale(1.05);
|
| 308 |
+
}
|
| 309 |
+
.header .nav .logo a {
|
| 310 |
+
color: #ffffff;
|
| 311 |
+
text-decoration: none;
|
| 312 |
+
font-size: 16px;
|
| 313 |
+
font-weight: bold;
|
| 314 |
+
padding: 0 10px;
|
| 315 |
+
transition: all 0.3s;
|
| 316 |
+
}
|
| 317 |
+
.header .nav .logo a:hover {
|
| 318 |
+
color: #ffcc00;
|
| 319 |
+
transform: scale(1.05);
|
| 320 |
+
}
|
| 321 |
+
.header .nav .logo a {
|
| 322 |
+
color: #ffffff;
|
| 323 |
+
text-decoration: none;
|
| 324 |
+
font-size: 16px;
|
| 325 |
+
font-weight: bold;
|
| 326 |
+
padding: 0 10px;
|
| 327 |
+
transition: all 0.3s;
|
| 328 |
+
}
|
| 329 |
+
.header .nav .logo a:hover {
|
| 330 |
+
color: #ffcc00;
|
| 331 |
+
transform: scale(1.05);
|
| 332 |
+
}
|
| 333 |
+
.header .nav .logo a {
|
| 334 |
+
color: #ffffff;
|
| 335 |
+
text-decoration: none;
|
| 336 |
+
font-size: 16px;
|
| 337 |
+
font-weight: bold;
|
| 338 |
+
padding: 0 10px;
|
| 339 |
+
transition: all 0.3s;
|
| 340 |
+
}
|
| 341 |
+
.header .nav .logo a:hover {
|
| 342 |
+
color: #ffcc00;
|
| 343 |
+
transform: scale(1.05);
|
| 344 |
+
}
|
| 345 |
+
.header .nav .logo a {
|
| 346 |
+
color: #ffffff;
|
| 347 |
+
text-decoration: none;
|
| 348 |
+
font-size: 16px;
|
| 349 |
+
font-weight: bold;
|
| 350 |
+
padding: 0 10px;
|
| 351 |
+
transition: all 0.3s;
|
| 352 |
+
}
|
| 353 |
+
.header .nav .logo a:hover {
|
| 354 |
+
color: #ffcc00;
|
| 355 |
+
transform: scale(1.05);
|
| 356 |
+
}
|
| 357 |
+
.header .nav .logo a {
|
| 358 |
+
color: #ffffff;
|
| 359 |
+
text-decoration: none;
|
| 360 |
+
font-size: 16px;
|
| 361 |
+
font-weight: bold;
|
| 362 |
+
padding: 0 10px;
|
| 363 |
+
transition: all 0.3s;
|
| 364 |
+
}
|
| 365 |
+
.header .nav .logo a:hover {
|
| 366 |
+
color: #ffcc00;
|
| 367 |
+
transform: scale(1.05);
|
| 368 |
+
}
|
| 369 |
+
.header .nav .logo a {
|
| 370 |
+
color: #ffffff;
|
| 371 |
+
text-decoration: none;
|
| 372 |
+
font-size: 16px;
|
| 373 |
+
font-weight: bold;
|
| 374 |
+
padding: 0 10px;
|
| 375 |
+
transition: all 0.3s;
|
| 376 |
+
}
|
| 377 |
+
.header .nav .logo a:hover {
|
| 378 |
+
color: #ffcc00;
|
| 379 |
+
transform: scale(1.05);
|
| 380 |
+
}
|
| 381 |
+
.header .nav .logo a {
|
| 382 |
+
color: #ffffff;
|
| 383 |
+
text-decoration: none;
|
| 384 |
+
font-size: 16px;
|
| 385 |
+
font-weight: bold;
|
| 386 |
+
padding: 0 10px;
|
| 387 |
+
transition: all 0.3s;
|
| 388 |
+
}
|
| 389 |
+
.header .nav .logo a:hover {
|
| 390 |
+
color: #ffcc00;
|
| 391 |
+
transform: scale(1.05);
|
| 392 |
+
}
|
| 393 |
+
.header .nav .logo a {
|
| 394 |
+
color: #ffffff;
|
| 395 |
+
text-decoration: none;
|
| 396 |
+
font-size: 16px;
|
| 397 |
+
font-weight: bold;
|
| 398 |
+
padding: 0 10px;
|
| 399 |
+
transition: all 0.3s;
|
| 400 |
+
}
|
| 401 |
+
.header .nav .logo a:hover {
|
| 402 |
+
color: #ffcc00;
|
| 403 |
+
transform: scale(1.05);
|
| 404 |
+
}
|
| 405 |
+
.header .nav .logo a {
|
| 406 |
+
color: #ffffff;
|
| 407 |
+
text-decoration: none;
|
| 408 |
+
font-size: 16px;
|
| 409 |
+
font-weight: bold;
|
| 410 |
+
padding: 0 10px;
|
| 411 |
+
transition: all 0.3s;
|
| 412 |
+
}
|
| 413 |
+
.header .nav .logo a:hover {
|
| 414 |
+
color: #ffcc00;
|
| 415 |
+
transform: scale(1.05);
|
| 416 |
+
}
|
| 417 |
+
.header .nav .logo a {
|
| 418 |
+
color: #ffffff;
|
| 419 |
+
text-decoration: none;
|
| 420 |
+
font-size: 16px;
|
| 421 |
+
font-weight: bold;
|
| 422 |
+
padding: 0 10px;
|
| 423 |
+
transition: all 0.3s;
|
| 424 |
+
}
|
| 425 |
+
.header .nav .logo a:hover {
|
| 426 |
+
color: #ffcc00;
|
| 427 |
+
transform: scale(1.05);
|
| 428 |
+
}
|
| 429 |
+
.header .nav .logo a {
|
| 430 |
+
color: #ffffff;
|
| 431 |
+
text-decoration: none;
|
| 432 |
+
font-size: 16px;
|
| 433 |
+
font-weight: bold;
|
| 434 |
+
padding: 0 10px;
|
| 435 |
+
transition: all 0.3s;
|
| 436 |
+
}
|
| 437 |
+
.header .nav .logo a:hover {
|
| 438 |
+
color: #ffcc00;
|
| 439 |
+
transform: scale(1.05);
|
| 440 |
+
}
|
| 441 |
+
.header .nav .logo a {
|
| 442 |
+
color: #ffffff;
|
| 443 |
+
text-decoration: none;
|
| 444 |
+
font-size: 16px;
|
| 445 |
+
font-weight: bold;
|
| 446 |
+
padding: 0 10px;
|
| 447 |
+
transition: all 0.3s;
|
| 448 |
+
}
|
| 449 |
+
.header .nav .logo a:hover {
|
| 450 |
+
color: #ffcc00;
|
| 451 |
+
transform: scale(1.05);
|
| 452 |
+
}
|
| 453 |
+
.header .nav .logo a {
|
| 454 |
+
color: #ffffff;
|
| 455 |
+
text-decoration: none;
|
| 456 |
+
font-size: 16px;
|
| 457 |
+
font-weight: bold;
|
| 458 |
+
padding: 0 10px;
|
| 459 |
+
transition: all 0.3s;
|
| 460 |
+
}
|
| 461 |
+
.header .nav .logo a:hover {
|
| 462 |
+
color: #ffcc00;
|
| 463 |
+
transform: scale(1.05);
|
| 464 |
+
}
|
| 465 |
+
.header .nav .logo a {
|
| 466 |
+
color: #ffffff;
|
| 467 |
+
text-decoration: none;
|
| 468 |
+
font-size: 16px;
|
| 469 |
+
font-weight: bold;
|
| 470 |
+
padding: 0 10px;
|
| 471 |
+
transition: all 0.3s;
|
| 472 |
+
}
|
| 473 |
+
.header .nav .logo a:hover {
|
| 474 |
+
color: #ffcc00;
|
| 475 |
+
transform: scale(1.05);
|
| 476 |
+
}
|
| 477 |
+
.header .nav .logo a {
|
| 478 |
+
color: #ffffff;
|
| 479 |
+
text-decoration: none;
|
| 480 |
+
font-size: 16px;
|
| 481 |
+
font-weight: bold;
|
| 482 |
+
padding: 0 10px;
|
| 483 |
+
transition: all 0.3s;
|
| 484 |
+
}
|
| 485 |
+
.header .nav .logo a:hover {
|
| 486 |
+
color: #ffcc00;
|
| 487 |
+
transform: scale(1.05);
|
| 488 |
+
}
|
| 489 |
+
.header .nav .logo a {
|
| 490 |
+
color: #ffffff;
|
| 491 |
+
text-decoration: none;
|
| 492 |
+
font-size: 16px;
|
| 493 |
+
font-weight: bold;
|
| 494 |
+
padding: 0 10px;
|
| 495 |
+
transition: all 0.3s;
|
| 496 |
+
}
|
| 497 |
+
.header .nav .logo a:hover {
|
| 498 |
+
color: #ffcc00;
|
| 499 |
+
transform: scale(1.05);
|
| 500 |
+
}
|
| 501 |
+
.header .nav .logo a {
|
| 502 |
+
color: #ffffff;
|
| 503 |
+
text-decoration: none;
|
| 504 |
+
font-size: 16px;
|
| 505 |
+
font-weight: bold;
|
| 506 |
+
padding: 0 10px;
|
| 507 |
+
transition: all 0.3s;
|
| 508 |
+
}
|
| 509 |
+
.header .nav .logo a:hover {
|
| 510 |
+
color: #ffcc00;
|
| 511 |
+
transform: scale(1.05);
|
| 512 |
+
}
|
| 513 |
+
.header .nav .logo a {
|
| 514 |
+
color: #ffffff;
|
| 515 |
+
text-decoration: none;
|
| 516 |
+
font-size: 16px;
|
| 517 |
+
font-weight: bold;
|
| 518 |
+
padding: 0 10px;
|
| 519 |
+
transition: all 0.3s;
|
| 520 |
+
}
|
| 521 |
+
.header .nav .logo a:hover {
|
| 522 |
+
color: #ffcc00;
|
| 523 |
+
transform: scale(1.05);
|
| 524 |
+
}
|
| 525 |
+
.header .nav .logo a {
|
| 526 |
+
color: #ffffff;
|
| 527 |
+
text-decoration: none;
|
| 528 |
+
font-size: 16px;
|
| 529 |
+
font-weight: bold;
|
| 530 |
+
padding: 0 10px;
|
| 531 |
+
transition: all 0.3s;
|
| 532 |
+
}
|
| 533 |
+
.header .nav .logo a:hover {
|
| 534 |
+
color: #ffcc00;
|
| 535 |
+
transform: scale(1.05);
|
| 536 |
+
}
|
| 537 |
+
.header .nav .logo a {
|
| 538 |
+
color: #ffffff;
|
| 539 |
+
text-decoration: none;
|
| 540 |
+
font-size: 16px;
|
| 541 |
+
font-weight: bold;
|
| 542 |
+
padding: 0 10px;
|
| 543 |
+
transition: all 0.3s;
|
| 544 |
+
}
|
| 545 |
+
.header .nav .logo a:hover {
|
| 546 |
+
color: #ffcc00;
|
| 547 |
+
transform: scale(1.05);
|
| 548 |
+
}
|
| 549 |
+
.header .nav .logo a {
|
| 550 |
+
color: #ffffff;
|
| 551 |
+
text-decoration: none;
|
| 552 |
+
font-size: 16px;
|
| 553 |
+
font-weight: bold;
|
| 554 |
+
padding: 0 10px;
|
| 555 |
+
transition: all 0.3s;
|
| 556 |
+
}
|
| 557 |
+
.header .nav .logo a:hover {
|
| 558 |
+
color: #
|
results/comparison/uipress_256/uipress_256/html_predictions/137.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/138.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/139.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/html_predictions/14.html
ADDED
|
@@ -0,0 +1,625 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<title>WELCOME TO THE OFFICIAL WEBSITE OF THE HOSPITAL OF THE CITY OF BAGUIOS</title>
|
| 6 |
+
<style>
|
| 7 |
+
body {
|
| 8 |
+
margin: 0;
|
| 9 |
+
padding: 0;
|
| 10 |
+
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
|
| 11 |
+
background-color: #f5f5f;
|
| 12 |
+
overflow-x: hidden;
|
| 13 |
+
overflow-y: auto;
|
| 14 |
+
}
|
| 15 |
+
.container {
|
| 16 |
+
width: 100%;
|
| 17 |
+
height: 100%;
|
| 18 |
+
position: relative;
|
| 19 |
+
padding: 0;
|
| 20 |
+
margin: 0;
|
| 21 |
+
}
|
| 22 |
+
.header {
|
| 23 |
+
width: 100%;
|
| 24 |
+
height: 100%;
|
| 25 |
+
position: relative;
|
| 26 |
+
background-color: #000000;
|
| 27 |
+
padding: 0;
|
| 28 |
+
margin: 0;
|
| 29 |
+
}
|
| 30 |
+
.header .logo {
|
| 31 |
+
width: 100%;
|
| 32 |
+
height: 100%;
|
| 33 |
+
position: absolute;
|
| 34 |
+
top: 0;
|
| 35 |
+
left: 0;
|
| 36 |
+
padding: 0;
|
| 37 |
+
margin: 0;
|
| 38 |
+
z-index: 100;
|
| 39 |
+
}
|
| 40 |
+
.header .logo img {
|
| 41 |
+
width: 100%;
|
| 42 |
+
height: 100%;
|
| 43 |
+
position: absolute;
|
| 44 |
+
top: 0;
|
| 45 |
+
left: 0;
|
| 46 |
+
z-index: 100;
|
| 47 |
+
}
|
| 48 |
+
.header .logo .logo {
|
| 49 |
+
width: 100%;
|
| 50 |
+
height: 100%;
|
| 51 |
+
position: absolute;
|
| 52 |
+
top: 0;
|
| 53 |
+
left: 0;
|
| 54 |
+
z-index: 100;
|
| 55 |
+
}
|
| 56 |
+
.header .logo .logo img {
|
| 57 |
+
width: 100%;
|
| 58 |
+
height: 100%;
|
| 59 |
+
position: absolute;
|
| 60 |
+
top: 0;
|
| 61 |
+
left: 0;
|
| 62 |
+
z-index: 100;
|
| 63 |
+
}
|
| 64 |
+
.header .logo .logo {
|
| 65 |
+
width: 100%;
|
| 66 |
+
height: 100%;
|
| 67 |
+
position: absolute;
|
| 68 |
+
top: 0;
|
| 69 |
+
left: 0;
|
| 70 |
+
z-index: 100;
|
| 71 |
+
}
|
| 72 |
+
.header .logo .logo img {
|
| 73 |
+
width: 100%;
|
| 74 |
+
height: 100%;
|
| 75 |
+
position: absolute;
|
| 76 |
+
top: 0;
|
| 77 |
+
left: 0;
|
| 78 |
+
z-index: 100;
|
| 79 |
+
}
|
| 80 |
+
.header .logo .logo {
|
| 81 |
+
width: 100%;
|
| 82 |
+
height: 100%;
|
| 83 |
+
position: absolute;
|
| 84 |
+
top: 0;
|
| 85 |
+
left: 0;
|
| 86 |
+
z-index: 100;
|
| 87 |
+
}
|
| 88 |
+
.header .logo .logo img {
|
| 89 |
+
width: 100%;
|
| 90 |
+
height: 100%;
|
| 91 |
+
position: absolute;
|
| 92 |
+
top: 0;
|
| 93 |
+
left: 0;
|
| 94 |
+
z-index: 100;
|
| 95 |
+
}
|
| 96 |
+
.header .logo .logo {
|
| 97 |
+
width: 100%;
|
| 98 |
+
height: 100%;
|
| 99 |
+
position: absolute;
|
| 100 |
+
top: 0;
|
| 101 |
+
left: 0;
|
| 102 |
+
z-index: 100;
|
| 103 |
+
}
|
| 104 |
+
.header .logo .logo img {
|
| 105 |
+
width: 100%;
|
| 106 |
+
height: 100%;
|
| 107 |
+
position: absolute;
|
| 108 |
+
top: 0;
|
| 109 |
+
left: 0;
|
| 110 |
+
z-index: 100;
|
| 111 |
+
}
|
| 112 |
+
.header .logo .logo {
|
| 113 |
+
width: 100%;
|
| 114 |
+
height: 100%;
|
| 115 |
+
position: absolute;
|
| 116 |
+
top: 0;
|
| 117 |
+
left: 0;
|
| 118 |
+
z-index: 100;
|
| 119 |
+
}
|
| 120 |
+
.header .logo .logo img {
|
| 121 |
+
width: 100%;
|
| 122 |
+
height: 100%;
|
| 123 |
+
position: absolute;
|
| 124 |
+
top: 0;
|
| 125 |
+
left: 0;
|
| 126 |
+
z-index: 100;
|
| 127 |
+
}
|
| 128 |
+
.header .logo .logo {
|
| 129 |
+
width: 100%;
|
| 130 |
+
height: 100%;
|
| 131 |
+
position: absolute;
|
| 132 |
+
top: 0;
|
| 133 |
+
left: 0;
|
| 134 |
+
z-index: 100;
|
| 135 |
+
}
|
| 136 |
+
.header .logo .logo img {
|
| 137 |
+
width: 100%;
|
| 138 |
+
height: 100%;
|
| 139 |
+
position: absolute;
|
| 140 |
+
top: 0;
|
| 141 |
+
left: 0;
|
| 142 |
+
z-index: 100;
|
| 143 |
+
}
|
| 144 |
+
.header .logo .logo {
|
| 145 |
+
width: 100%;
|
| 146 |
+
height: 100%;
|
| 147 |
+
position: absolute;
|
| 148 |
+
top: 0;
|
| 149 |
+
left: 0;
|
| 150 |
+
z-index: 100;
|
| 151 |
+
}
|
| 152 |
+
.header .logo .logo img {
|
| 153 |
+
width: 100%;
|
| 154 |
+
height: 100%;
|
| 155 |
+
position: absolute;
|
| 156 |
+
top: 0;
|
| 157 |
+
left: 0;
|
| 158 |
+
z-index: 100;
|
| 159 |
+
}
|
| 160 |
+
.header .logo .logo {
|
| 161 |
+
width: 100%;
|
| 162 |
+
height: 100%;
|
| 163 |
+
position: absolute;
|
| 164 |
+
top: 0;
|
| 165 |
+
left: 0;
|
| 166 |
+
z-index: 100;
|
| 167 |
+
}
|
| 168 |
+
.header .logo .logo img {
|
| 169 |
+
width: 100%;
|
| 170 |
+
height: 100%;
|
| 171 |
+
position: absolute;
|
| 172 |
+
top: 0;
|
| 173 |
+
left: 0;
|
| 174 |
+
z-index: 100;
|
| 175 |
+
}
|
| 176 |
+
.header .logo .logo {
|
| 177 |
+
width: 100%;
|
| 178 |
+
height: 100%;
|
| 179 |
+
position: absolute;
|
| 180 |
+
top: 0;
|
| 181 |
+
left: 0;
|
| 182 |
+
z-index: 100;
|
| 183 |
+
}
|
| 184 |
+
.header .logo .logo img {
|
| 185 |
+
width: 100%;
|
| 186 |
+
height: 100%;
|
| 187 |
+
position: absolute;
|
| 188 |
+
top: 0;
|
| 189 |
+
left: 0;
|
| 190 |
+
z-index: 100;
|
| 191 |
+
}
|
| 192 |
+
.header .logo .logo {
|
| 193 |
+
width: 100%;
|
| 194 |
+
height: 100%;
|
| 195 |
+
position: absolute;
|
| 196 |
+
top: 0;
|
| 197 |
+
left: 0;
|
| 198 |
+
z-index: 100;
|
| 199 |
+
}
|
| 200 |
+
.header .logo .logo img {
|
| 201 |
+
width: 100%;
|
| 202 |
+
height: 100%;
|
| 203 |
+
position: absolute;
|
| 204 |
+
top: 0;
|
| 205 |
+
left: 0;
|
| 206 |
+
z-index: 100;
|
| 207 |
+
}
|
| 208 |
+
.header .logo .logo {
|
| 209 |
+
width: 100%;
|
| 210 |
+
height: 100%;
|
| 211 |
+
position: absolute;
|
| 212 |
+
top: 0;
|
| 213 |
+
left: 0;
|
| 214 |
+
z-index: 100;
|
| 215 |
+
}
|
| 216 |
+
.header .logo .logo img {
|
| 217 |
+
width: 100%;
|
| 218 |
+
height: 100%;
|
| 219 |
+
position: absolute;
|
| 220 |
+
top: 0;
|
| 221 |
+
left: 0;
|
| 222 |
+
z-index: 100;
|
| 223 |
+
}
|
| 224 |
+
.header .logo .logo {
|
| 225 |
+
width: 100%;
|
| 226 |
+
height: 100%;
|
| 227 |
+
position: absolute;
|
| 228 |
+
top: 0;
|
| 229 |
+
left: 0;
|
| 230 |
+
z-index: 100;
|
| 231 |
+
}
|
| 232 |
+
.header .logo .logo img {
|
| 233 |
+
width: 100%;
|
| 234 |
+
height: 100%;
|
| 235 |
+
position: absolute;
|
| 236 |
+
top: 0;
|
| 237 |
+
left: 0;
|
| 238 |
+
z-index: 100;
|
| 239 |
+
}
|
| 240 |
+
.header .logo .logo {
|
| 241 |
+
width: 100%;
|
| 242 |
+
height: 100%;
|
| 243 |
+
position: absolute;
|
| 244 |
+
top: 0;
|
| 245 |
+
left: 0;
|
| 246 |
+
z-index: 100;
|
| 247 |
+
}
|
| 248 |
+
.header .logo .logo img {
|
| 249 |
+
width: 100%;
|
| 250 |
+
height: 100%;
|
| 251 |
+
position: absolute;
|
| 252 |
+
top: 0;
|
| 253 |
+
left: 0;
|
| 254 |
+
z-index: 100;
|
| 255 |
+
}
|
| 256 |
+
.header .logo .logo {
|
| 257 |
+
width: 100%;
|
| 258 |
+
height: 100%;
|
| 259 |
+
position: absolute;
|
| 260 |
+
top: 0;
|
| 261 |
+
left: 0;
|
| 262 |
+
z-index: 100;
|
| 263 |
+
}
|
| 264 |
+
.header .logo .logo img {
|
| 265 |
+
width: 100%;
|
| 266 |
+
height: 100%;
|
| 267 |
+
position: absolute;
|
| 268 |
+
top: 0;
|
| 269 |
+
left: 0;
|
| 270 |
+
z-index: 100;
|
| 271 |
+
}
|
| 272 |
+
.header .logo .logo {
|
| 273 |
+
width: 100%;
|
| 274 |
+
height: 100%;
|
| 275 |
+
position: absolute;
|
| 276 |
+
top: 0;
|
| 277 |
+
left: 0;
|
| 278 |
+
z-index: 100;
|
| 279 |
+
}
|
| 280 |
+
.header .logo .logo img {
|
| 281 |
+
width: 100%;
|
| 282 |
+
height: 100%;
|
| 283 |
+
position: absolute;
|
| 284 |
+
top: 0;
|
| 285 |
+
left: 0;
|
| 286 |
+
z-index: 100;
|
| 287 |
+
}
|
| 288 |
+
.header .logo .logo {
|
| 289 |
+
width: 100%;
|
| 290 |
+
height: 100%;
|
| 291 |
+
position: absolute;
|
| 292 |
+
top: 0;
|
| 293 |
+
left: 0;
|
| 294 |
+
z-index: 100;
|
| 295 |
+
}
|
| 296 |
+
.header .logo .logo img {
|
| 297 |
+
width: 100%;
|
| 298 |
+
height: 100%;
|
| 299 |
+
position: absolute;
|
| 300 |
+
top: 0;
|
| 301 |
+
left: 0;
|
| 302 |
+
z-index: 100;
|
| 303 |
+
}
|
| 304 |
+
.header .logo .logo {
|
| 305 |
+
width: 100%;
|
| 306 |
+
height: 100%;
|
| 307 |
+
position: absolute;
|
| 308 |
+
top: 0;
|
| 309 |
+
left: 0;
|
| 310 |
+
z-index: 100;
|
| 311 |
+
}
|
| 312 |
+
.header .logo .logo img {
|
| 313 |
+
width: 100%;
|
| 314 |
+
height: 100%;
|
| 315 |
+
position: absolute;
|
| 316 |
+
top: 0;
|
| 317 |
+
left: 0;
|
| 318 |
+
z-index: 100;
|
| 319 |
+
}
|
| 320 |
+
.header .logo .logo {
|
| 321 |
+
width: 100%;
|
| 322 |
+
height: 100%;
|
| 323 |
+
position: absolute;
|
| 324 |
+
top: 0;
|
| 325 |
+
left: 0;
|
| 326 |
+
z-index: 100;
|
| 327 |
+
}
|
| 328 |
+
.header .logo .logo img {
|
| 329 |
+
width: 100%;
|
| 330 |
+
height: 100%;
|
| 331 |
+
position: absolute;
|
| 332 |
+
top: 0;
|
| 333 |
+
left: 0;
|
| 334 |
+
z-index: 100;
|
| 335 |
+
}
|
| 336 |
+
.header .logo .logo {
|
| 337 |
+
width: 100%;
|
| 338 |
+
height: 100%;
|
| 339 |
+
position: absolute;
|
| 340 |
+
top: 0;
|
| 341 |
+
left: 0;
|
| 342 |
+
z-index: 100;
|
| 343 |
+
}
|
| 344 |
+
.header .logo .logo img {
|
| 345 |
+
width: 100%;
|
| 346 |
+
height: 100%;
|
| 347 |
+
position: absolute;
|
| 348 |
+
top: 0;
|
| 349 |
+
left: 0;
|
| 350 |
+
z-index: 100;
|
| 351 |
+
}
|
| 352 |
+
.header .logo .logo {
|
| 353 |
+
width: 100%;
|
| 354 |
+
height: 100%;
|
| 355 |
+
position: absolute;
|
| 356 |
+
top: 0;
|
| 357 |
+
left: 0;
|
| 358 |
+
z-index: 100;
|
| 359 |
+
}
|
| 360 |
+
.header .logo .logo img {
|
| 361 |
+
width: 100%;
|
| 362 |
+
height: 100%;
|
| 363 |
+
position: absolute;
|
| 364 |
+
top: 0;
|
| 365 |
+
left: 0;
|
| 366 |
+
z-index: 100;
|
| 367 |
+
}
|
| 368 |
+
.header .logo .logo {
|
| 369 |
+
width: 100%;
|
| 370 |
+
height: 100%;
|
| 371 |
+
position: absolute;
|
| 372 |
+
top: 0;
|
| 373 |
+
left: 0;
|
| 374 |
+
z-index: 100;
|
| 375 |
+
}
|
| 376 |
+
.header .logo .logo img {
|
| 377 |
+
width: 100%;
|
| 378 |
+
height: 100%;
|
| 379 |
+
position: absolute;
|
| 380 |
+
top: 0;
|
| 381 |
+
left: 0;
|
| 382 |
+
z-index: 100;
|
| 383 |
+
}
|
| 384 |
+
.header .logo .logo {
|
| 385 |
+
width: 100%;
|
| 386 |
+
height: 100%;
|
| 387 |
+
position: absolute;
|
| 388 |
+
top: 0;
|
| 389 |
+
left: 0;
|
| 390 |
+
z-index: 100;
|
| 391 |
+
}
|
| 392 |
+
.header .logo .logo img {
|
| 393 |
+
width: 100%;
|
| 394 |
+
height: 100%;
|
| 395 |
+
position: absolute;
|
| 396 |
+
top: 0;
|
| 397 |
+
left: 0;
|
| 398 |
+
z-index: 100;
|
| 399 |
+
}
|
| 400 |
+
.header .logo .logo {
|
| 401 |
+
width: 100%;
|
| 402 |
+
height: 100%;
|
| 403 |
+
position: absolute;
|
| 404 |
+
top: 0;
|
| 405 |
+
left: 0;
|
| 406 |
+
z-index: 100;
|
| 407 |
+
}
|
| 408 |
+
.header .logo .logo img {
|
| 409 |
+
width: 100%;
|
| 410 |
+
height: 100%;
|
| 411 |
+
position: absolute;
|
| 412 |
+
top: 0;
|
| 413 |
+
left: 0;
|
| 414 |
+
z-index: 100;
|
| 415 |
+
}
|
| 416 |
+
.header .logo .logo {
|
| 417 |
+
width: 100%;
|
| 418 |
+
height: 100%;
|
| 419 |
+
position: absolute;
|
| 420 |
+
top: 0;
|
| 421 |
+
left: 0;
|
| 422 |
+
z-index: 100;
|
| 423 |
+
}
|
| 424 |
+
.header .logo .logo img {
|
| 425 |
+
width: 100%;
|
| 426 |
+
height: 100%;
|
| 427 |
+
position: absolute;
|
| 428 |
+
top: 0;
|
| 429 |
+
left: 0;
|
| 430 |
+
z-index: 100;
|
| 431 |
+
}
|
| 432 |
+
.header .logo .logo {
|
| 433 |
+
width: 100%;
|
| 434 |
+
height: 100%;
|
| 435 |
+
position: absolute;
|
| 436 |
+
top: 0;
|
| 437 |
+
left: 0;
|
| 438 |
+
z-index: 100;
|
| 439 |
+
}
|
| 440 |
+
.header .logo .logo img {
|
| 441 |
+
width: 100%;
|
| 442 |
+
height: 100%;
|
| 443 |
+
position: absolute;
|
| 444 |
+
top: 0;
|
| 445 |
+
left: 0;
|
| 446 |
+
z-index: 100;
|
| 447 |
+
}
|
| 448 |
+
.header .logo .logo {
|
| 449 |
+
width: 100%;
|
| 450 |
+
height: 100%;
|
| 451 |
+
position: absolute;
|
| 452 |
+
top: 0;
|
| 453 |
+
left: 0;
|
| 454 |
+
z-index: 100;
|
| 455 |
+
}
|
| 456 |
+
.header .logo .logo img {
|
| 457 |
+
width: 100%;
|
| 458 |
+
height: 100%;
|
| 459 |
+
position: absolute;
|
| 460 |
+
top: 0;
|
| 461 |
+
left: 0;
|
| 462 |
+
z-index: 100;
|
| 463 |
+
}
|
| 464 |
+
.header .logo .logo {
|
| 465 |
+
width: 100%;
|
| 466 |
+
height: 100%;
|
| 467 |
+
position: absolute;
|
| 468 |
+
top: 0;
|
| 469 |
+
left: 0;
|
| 470 |
+
z-index: 100;
|
| 471 |
+
}
|
| 472 |
+
.header .logo .logo img {
|
| 473 |
+
width: 100%;
|
| 474 |
+
height: 100%;
|
| 475 |
+
position: absolute;
|
| 476 |
+
top: 0;
|
| 477 |
+
left: 0;
|
| 478 |
+
z-index: 100;
|
| 479 |
+
}
|
| 480 |
+
.header .logo .logo {
|
| 481 |
+
width: 100%;
|
| 482 |
+
height: 100%;
|
| 483 |
+
position: absolute;
|
| 484 |
+
top: 0;
|
| 485 |
+
left: 0;
|
| 486 |
+
z-index: 100;
|
| 487 |
+
}
|
| 488 |
+
.header .logo .logo img {
|
| 489 |
+
width: 100%;
|
| 490 |
+
height: 100%;
|
| 491 |
+
position: absolute;
|
| 492 |
+
top: 0;
|
| 493 |
+
left: 0;
|
| 494 |
+
z-index: 100;
|
| 495 |
+
}
|
| 496 |
+
.header .logo .logo {
|
| 497 |
+
width: 100%;
|
| 498 |
+
height: 100%;
|
| 499 |
+
position: absolute;
|
| 500 |
+
top: 0;
|
| 501 |
+
left: 0;
|
| 502 |
+
z-index: 100;
|
| 503 |
+
}
|
| 504 |
+
.header .logo .logo img {
|
| 505 |
+
width: 100%;
|
| 506 |
+
height: 100%;
|
| 507 |
+
position: absolute;
|
| 508 |
+
top: 0;
|
| 509 |
+
left: 0;
|
| 510 |
+
z-index: 100;
|
| 511 |
+
}
|
| 512 |
+
.header .logo .logo {
|
| 513 |
+
width: 100%;
|
| 514 |
+
height: 100%;
|
| 515 |
+
position: absolute;
|
| 516 |
+
top: 0;
|
| 517 |
+
left: 0;
|
| 518 |
+
z-index: 100;
|
| 519 |
+
}
|
| 520 |
+
.header .logo .logo img {
|
| 521 |
+
width: 100%;
|
| 522 |
+
height: 100%;
|
| 523 |
+
position: absolute;
|
| 524 |
+
top: 0;
|
| 525 |
+
left: 0;
|
| 526 |
+
z-index: 100;
|
| 527 |
+
}
|
| 528 |
+
.header .logo .logo {
|
| 529 |
+
width: 100%;
|
| 530 |
+
height: 100%;
|
| 531 |
+
position: absolute;
|
| 532 |
+
top: 0;
|
| 533 |
+
left: 0;
|
| 534 |
+
z-index: 100;
|
| 535 |
+
}
|
| 536 |
+
.header .logo .logo img {
|
| 537 |
+
width: 100%;
|
| 538 |
+
height: 100%;
|
| 539 |
+
position: absolute;
|
| 540 |
+
top: 0;
|
| 541 |
+
left: 0;
|
| 542 |
+
z-index: 100;
|
| 543 |
+
}
|
| 544 |
+
.header .logo .logo {
|
| 545 |
+
width: 100%;
|
| 546 |
+
height: 100%;
|
| 547 |
+
position: absolute;
|
| 548 |
+
top: 0;
|
| 549 |
+
left: 0;
|
| 550 |
+
z-index: 100;
|
| 551 |
+
}
|
| 552 |
+
.header .logo .logo img {
|
| 553 |
+
width: 100%;
|
| 554 |
+
height: 100%;
|
| 555 |
+
position: absolute;
|
| 556 |
+
top: 0;
|
| 557 |
+
left: 0;
|
| 558 |
+
z-index: 100;
|
| 559 |
+
}
|
| 560 |
+
.header .logo .logo {
|
| 561 |
+
width: 100%;
|
| 562 |
+
height: 100%;
|
| 563 |
+
position: absolute;
|
| 564 |
+
top: 0;
|
| 565 |
+
left: 0;
|
| 566 |
+
z-index: 100;
|
| 567 |
+
}
|
| 568 |
+
.header .logo .logo img {
|
| 569 |
+
width: 100%;
|
| 570 |
+
height: 100%;
|
| 571 |
+
position: absolute;
|
| 572 |
+
top: 0;
|
| 573 |
+
left: 0;
|
| 574 |
+
z-index: 100;
|
| 575 |
+
}
|
| 576 |
+
.header .logo .logo {
|
| 577 |
+
width: 100%;
|
| 578 |
+
height: 100%;
|
| 579 |
+
position: absolute;
|
| 580 |
+
top: 0;
|
| 581 |
+
left: 0;
|
| 582 |
+
z-index: 100;
|
| 583 |
+
}
|
| 584 |
+
.header .logo .logo img {
|
| 585 |
+
width: 100%;
|
| 586 |
+
height: 100%;
|
| 587 |
+
position: absolute;
|
| 588 |
+
top: 0;
|
| 589 |
+
left: 0;
|
| 590 |
+
z-index: 100;
|
| 591 |
+
}
|
| 592 |
+
.header .logo .logo {
|
| 593 |
+
width: 100%;
|
| 594 |
+
height: 100%;
|
| 595 |
+
position: absolute;
|
| 596 |
+
top: 0;
|
| 597 |
+
left: 0;
|
| 598 |
+
z-index: 100;
|
| 599 |
+
}
|
| 600 |
+
.header .logo .logo img {
|
| 601 |
+
width: 100%;
|
| 602 |
+
height: 100%;
|
| 603 |
+
position: absolute;
|
| 604 |
+
top: 0;
|
| 605 |
+
left: 0;
|
| 606 |
+
z-index: 100;
|
| 607 |
+
}
|
| 608 |
+
.header .logo .logo {
|
| 609 |
+
width: 100%;
|
| 610 |
+
height: 100%;
|
| 611 |
+
position: absolute;
|
| 612 |
+
top: 0;
|
| 613 |
+
left: 0;
|
| 614 |
+
z-index: 100;
|
| 615 |
+
}
|
| 616 |
+
.header .logo .logo img {
|
| 617 |
+
width: 100%;
|
| 618 |
+
height: 100%;
|
| 619 |
+
position: absolute;
|
| 620 |
+
top: 0;
|
| 621 |
+
left: 0;
|
| 622 |
+
z-index: 100;
|
| 623 |
+
}
|
| 624 |
+
.header .logo .logo {
|
| 625 |
+
width:
|
results/comparison/uipress_256/uipress_256/html_predictions/141.html
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
results/comparison/uipress_256/uipress_256/per_sample.json
ADDED
|
@@ -0,0 +1,352 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"id": "0",
|
| 4 |
+
"n_visual_tokens": 256,
|
| 5 |
+
"latency_s": 16.51,
|
| 6 |
+
"peak_mem_gb": 17.13,
|
| 7 |
+
"output_len": 1490
|
| 8 |
+
},
|
| 9 |
+
{
|
| 10 |
+
"id": "1",
|
| 11 |
+
"n_visual_tokens": 256,
|
| 12 |
+
"latency_s": 1.15,
|
| 13 |
+
"peak_mem_gb": 16.94,
|
| 14 |
+
"output_len": 61
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"id": "10",
|
| 18 |
+
"n_visual_tokens": 256,
|
| 19 |
+
"latency_s": 1.93,
|
| 20 |
+
"peak_mem_gb": 17.62,
|
| 21 |
+
"output_len": 61
|
| 22 |
+
},
|
| 23 |
+
{
|
| 24 |
+
"id": "100",
|
| 25 |
+
"n_visual_tokens": 256,
|
| 26 |
+
"latency_s": 7.31,
|
| 27 |
+
"peak_mem_gb": 17.11,
|
| 28 |
+
"output_len": 683
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"id": "101",
|
| 32 |
+
"n_visual_tokens": 256,
|
| 33 |
+
"latency_s": 1.59,
|
| 34 |
+
"peak_mem_gb": 17.37,
|
| 35 |
+
"output_len": 61
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"id": "102",
|
| 39 |
+
"n_visual_tokens": 256,
|
| 40 |
+
"latency_s": 13.52,
|
| 41 |
+
"peak_mem_gb": 17.15,
|
| 42 |
+
"output_len": 1410
|
| 43 |
+
},
|
| 44 |
+
{
|
| 45 |
+
"id": "103",
|
| 46 |
+
"n_visual_tokens": 256,
|
| 47 |
+
"latency_s": 7.08,
|
| 48 |
+
"peak_mem_gb": 17.11,
|
| 49 |
+
"output_len": 684
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"id": "104",
|
| 53 |
+
"n_visual_tokens": 256,
|
| 54 |
+
"latency_s": 158.76,
|
| 55 |
+
"peak_mem_gb": 17.54,
|
| 56 |
+
"output_len": 10665
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"id": "105",
|
| 60 |
+
"n_visual_tokens": 256,
|
| 61 |
+
"latency_s": 11.99,
|
| 62 |
+
"peak_mem_gb": 17.13,
|
| 63 |
+
"output_len": 1129
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"id": "106",
|
| 67 |
+
"n_visual_tokens": 256,
|
| 68 |
+
"latency_s": 1.61,
|
| 69 |
+
"peak_mem_gb": 17.38,
|
| 70 |
+
"output_len": 61
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"id": "107",
|
| 74 |
+
"n_visual_tokens": 256,
|
| 75 |
+
"latency_s": 2.28,
|
| 76 |
+
"peak_mem_gb": 17.36,
|
| 77 |
+
"output_len": 135
|
| 78 |
+
},
|
| 79 |
+
{
|
| 80 |
+
"id": "108",
|
| 81 |
+
"n_visual_tokens": 256,
|
| 82 |
+
"latency_s": 1.64,
|
| 83 |
+
"peak_mem_gb": 17.29,
|
| 84 |
+
"output_len": 61
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"id": "109",
|
| 88 |
+
"n_visual_tokens": 256,
|
| 89 |
+
"latency_s": 19.8,
|
| 90 |
+
"peak_mem_gb": 17.01,
|
| 91 |
+
"output_len": 2014
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"id": "11",
|
| 95 |
+
"n_visual_tokens": 256,
|
| 96 |
+
"latency_s": 156.35,
|
| 97 |
+
"peak_mem_gb": 17.39,
|
| 98 |
+
"output_len": 15523
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"id": "110",
|
| 102 |
+
"n_visual_tokens": 256,
|
| 103 |
+
"latency_s": 160.41,
|
| 104 |
+
"peak_mem_gb": 17.56,
|
| 105 |
+
"output_len": 14937
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"id": "111",
|
| 109 |
+
"n_visual_tokens": 256,
|
| 110 |
+
"latency_s": 160.89,
|
| 111 |
+
"peak_mem_gb": 17.49,
|
| 112 |
+
"output_len": 11093
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"id": "112",
|
| 116 |
+
"n_visual_tokens": 256,
|
| 117 |
+
"latency_s": 12.47,
|
| 118 |
+
"peak_mem_gb": 17.01,
|
| 119 |
+
"output_len": 1092
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"id": "113",
|
| 123 |
+
"n_visual_tokens": 256,
|
| 124 |
+
"latency_s": 0.96,
|
| 125 |
+
"peak_mem_gb": 16.94,
|
| 126 |
+
"output_len": 61
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"id": "114",
|
| 130 |
+
"n_visual_tokens": 256,
|
| 131 |
+
"latency_s": 7.41,
|
| 132 |
+
"peak_mem_gb": 17.89,
|
| 133 |
+
"output_len": 524
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"id": "115",
|
| 137 |
+
"n_visual_tokens": 256,
|
| 138 |
+
"latency_s": 0.98,
|
| 139 |
+
"peak_mem_gb": 16.94,
|
| 140 |
+
"output_len": 61
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"id": "116",
|
| 144 |
+
"n_visual_tokens": 256,
|
| 145 |
+
"latency_s": 189.32,
|
| 146 |
+
"peak_mem_gb": 17.67,
|
| 147 |
+
"output_len": 4278
|
| 148 |
+
},
|
| 149 |
+
{
|
| 150 |
+
"id": "117",
|
| 151 |
+
"n_visual_tokens": 256,
|
| 152 |
+
"latency_s": 27.95,
|
| 153 |
+
"peak_mem_gb": 16.98,
|
| 154 |
+
"output_len": 1173
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"id": "118",
|
| 158 |
+
"n_visual_tokens": 256,
|
| 159 |
+
"latency_s": 4.75,
|
| 160 |
+
"peak_mem_gb": 17.62,
|
| 161 |
+
"output_len": 114
|
| 162 |
+
},
|
| 163 |
+
{
|
| 164 |
+
"id": "119",
|
| 165 |
+
"n_visual_tokens": 256,
|
| 166 |
+
"latency_s": 338.31,
|
| 167 |
+
"peak_mem_gb": 17.44,
|
| 168 |
+
"output_len": 4219
|
| 169 |
+
},
|
| 170 |
+
{
|
| 171 |
+
"id": "12",
|
| 172 |
+
"n_visual_tokens": 256,
|
| 173 |
+
"latency_s": 2.06,
|
| 174 |
+
"peak_mem_gb": 16.94,
|
| 175 |
+
"output_len": 61
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"id": "120",
|
| 179 |
+
"n_visual_tokens": 256,
|
| 180 |
+
"latency_s": 26.24,
|
| 181 |
+
"peak_mem_gb": 16.98,
|
| 182 |
+
"output_len": 1032
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"id": "121",
|
| 186 |
+
"n_visual_tokens": 256,
|
| 187 |
+
"latency_s": 6.52,
|
| 188 |
+
"peak_mem_gb": 17.74,
|
| 189 |
+
"output_len": 180
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"id": "122",
|
| 193 |
+
"n_visual_tokens": 256,
|
| 194 |
+
"latency_s": 3.51,
|
| 195 |
+
"peak_mem_gb": 17.46,
|
| 196 |
+
"output_len": 61
|
| 197 |
+
},
|
| 198 |
+
{
|
| 199 |
+
"id": "123",
|
| 200 |
+
"n_visual_tokens": 256,
|
| 201 |
+
"latency_s": 2.06,
|
| 202 |
+
"peak_mem_gb": 16.94,
|
| 203 |
+
"output_len": 61
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"id": "125",
|
| 207 |
+
"n_visual_tokens": 256,
|
| 208 |
+
"latency_s": 27.72,
|
| 209 |
+
"peak_mem_gb": 17.21,
|
| 210 |
+
"output_len": 1207
|
| 211 |
+
},
|
| 212 |
+
{
|
| 213 |
+
"id": "126",
|
| 214 |
+
"n_visual_tokens": 256,
|
| 215 |
+
"latency_s": 6.21,
|
| 216 |
+
"peak_mem_gb": 17.69,
|
| 217 |
+
"output_len": 158
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"id": "127",
|
| 221 |
+
"n_visual_tokens": 256,
|
| 222 |
+
"latency_s": 2.07,
|
| 223 |
+
"peak_mem_gb": 16.94,
|
| 224 |
+
"output_len": 61
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"id": "128",
|
| 228 |
+
"n_visual_tokens": 256,
|
| 229 |
+
"latency_s": 27.48,
|
| 230 |
+
"peak_mem_gb": 17.05,
|
| 231 |
+
"output_len": 1363
|
| 232 |
+
},
|
| 233 |
+
{
|
| 234 |
+
"id": "129",
|
| 235 |
+
"n_visual_tokens": 256,
|
| 236 |
+
"latency_s": 234.42,
|
| 237 |
+
"peak_mem_gb": 17.71,
|
| 238 |
+
"output_len": 14331
|
| 239 |
+
},
|
| 240 |
+
{
|
| 241 |
+
"id": "13",
|
| 242 |
+
"n_visual_tokens": 256,
|
| 243 |
+
"latency_s": 158.92,
|
| 244 |
+
"peak_mem_gb": 17.58,
|
| 245 |
+
"output_len": 13648
|
| 246 |
+
},
|
| 247 |
+
{
|
| 248 |
+
"id": "130",
|
| 249 |
+
"n_visual_tokens": 256,
|
| 250 |
+
"latency_s": 157.18,
|
| 251 |
+
"peak_mem_gb": 17.56,
|
| 252 |
+
"output_len": 14865
|
| 253 |
+
},
|
| 254 |
+
{
|
| 255 |
+
"id": "131",
|
| 256 |
+
"n_visual_tokens": 256,
|
| 257 |
+
"latency_s": 154.34,
|
| 258 |
+
"peak_mem_gb": 17.38,
|
| 259 |
+
"output_len": 15728
|
| 260 |
+
},
|
| 261 |
+
{
|
| 262 |
+
"id": "132",
|
| 263 |
+
"n_visual_tokens": 256,
|
| 264 |
+
"latency_s": 160.23,
|
| 265 |
+
"peak_mem_gb": 17.55,
|
| 266 |
+
"output_len": 4718
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"id": "133",
|
| 270 |
+
"n_visual_tokens": 256,
|
| 271 |
+
"latency_s": 1.37,
|
| 272 |
+
"peak_mem_gb": 17.33,
|
| 273 |
+
"output_len": 61
|
| 274 |
+
},
|
| 275 |
+
{
|
| 276 |
+
"id": "134",
|
| 277 |
+
"n_visual_tokens": 256,
|
| 278 |
+
"latency_s": 4.61,
|
| 279 |
+
"peak_mem_gb": 17.19,
|
| 280 |
+
"output_len": 396
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"id": "135",
|
| 284 |
+
"n_visual_tokens": 256,
|
| 285 |
+
"latency_s": 160.6,
|
| 286 |
+
"peak_mem_gb": 17.56,
|
| 287 |
+
"output_len": 15537
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"id": "136",
|
| 291 |
+
"n_visual_tokens": 256,
|
| 292 |
+
"latency_s": 2.32,
|
| 293 |
+
"peak_mem_gb": 17.75,
|
| 294 |
+
"output_len": 76
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"id": "137",
|
| 298 |
+
"n_visual_tokens": 256,
|
| 299 |
+
"latency_s": 0.96,
|
| 300 |
+
"peak_mem_gb": 16.94,
|
| 301 |
+
"output_len": 61
|
| 302 |
+
},
|
| 303 |
+
{
|
| 304 |
+
"id": "138",
|
| 305 |
+
"n_visual_tokens": 256,
|
| 306 |
+
"latency_s": 0.97,
|
| 307 |
+
"peak_mem_gb": 16.94,
|
| 308 |
+
"output_len": 61
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"id": "139",
|
| 312 |
+
"n_visual_tokens": 256,
|
| 313 |
+
"latency_s": 1.86,
|
| 314 |
+
"peak_mem_gb": 17.51,
|
| 315 |
+
"output_len": 61
|
| 316 |
+
},
|
| 317 |
+
{
|
| 318 |
+
"id": "14",
|
| 319 |
+
"n_visual_tokens": 256,
|
| 320 |
+
"latency_s": 158.72,
|
| 321 |
+
"peak_mem_gb": 17.41,
|
| 322 |
+
"output_len": 15035
|
| 323 |
+
},
|
| 324 |
+
{
|
| 325 |
+
"id": "140",
|
| 326 |
+
"n_visual_tokens": 256,
|
| 327 |
+
"latency_s": 1.55,
|
| 328 |
+
"peak_mem_gb": 17.37,
|
| 329 |
+
"output_len": 61
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"id": "141",
|
| 333 |
+
"n_visual_tokens": 256,
|
| 334 |
+
"latency_s": 2.13,
|
| 335 |
+
"peak_mem_gb": 17.58,
|
| 336 |
+
"output_len": 61
|
| 337 |
+
},
|
| 338 |
+
{
|
| 339 |
+
"id": "142",
|
| 340 |
+
"n_visual_tokens": 256,
|
| 341 |
+
"latency_s": 1.28,
|
| 342 |
+
"peak_mem_gb": 16.95,
|
| 343 |
+
"output_len": 61
|
| 344 |
+
},
|
| 345 |
+
{
|
| 346 |
+
"id": "143",
|
| 347 |
+
"n_visual_tokens": 256,
|
| 348 |
+
"latency_s": 15.58,
|
| 349 |
+
"peak_mem_gb": 17.05,
|
| 350 |
+
"output_len": 1278
|
| 351 |
+
}
|
| 352 |
+
]
|
results/comparison/uipress_256/uipress_256/summary.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"method": "uipress_256",
|
| 3 |
+
"n_samples": 50,
|
| 4 |
+
"n_success": 50,
|
| 5 |
+
"avg_visual_tokens": 256.0,
|
| 6 |
+
"avg_latency_s": 52.52,
|
| 7 |
+
"avg_peak_mem_gb": 17.31
|
| 8 |
+
}
|
results/comparison/visionzip_256_quick2/visionzip_256/html_predictions/0.html
ADDED
|
@@ -0,0 +1,464 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```html
|
| 2 |
+
<!DOCTYPE html>
|
| 3 |
+
<html>
|
| 4 |
+
<head>
|
| 5 |
+
<meta charset="UTF-
|
| 6 |
+
<title>Banana Oatmeal Muffins</title>
|
| 7 |
+
<style>
|
| 8 |
+
body {
|
| 9 |
+
font-family: Arial, sans-serif;
|
| 10 |
+
background-color: #f9f9f;
|
| 11 |
+
padding: 20px;
|
| 12 |
+
line-height: 1.5;
|
| 13 |
+
color: #333333;
|
| 14 |
+
}
|
| 15 |
+
h1, h2, h3 {
|
| 16 |
+
color: #222222;
|
| 17 |
+
margin-top: 10px;
|
| 18 |
+
margin-bottom: 10px;
|
| 19 |
+
}
|
| 20 |
+
h1 {
|
| 21 |
+
font-size: 24px;
|
| 22 |
+
}
|
| 23 |
+
h2 {
|
| 24 |
+
font-size: 20px;
|
| 25 |
+
}
|
| 26 |
+
h3 {
|
| 27 |
+
font-size: 16px;
|
| 28 |
+
}
|
| 29 |
+
.header {
|
| 30 |
+
display: flex;
|
| 31 |
+
justify-content: space-between;
|
| 32 |
+
align-items: center;
|
| 33 |
+
margin-bottom: 20px;
|
| 34 |
+
padding: 10px 0;
|
| 35 |
+
background-color: #f0f0f0f;
|
| 36 |
+
border-bottom: 1px solid #ddd;
|
| 37 |
+
}
|
| 38 |
+
.header h1 {
|
| 39 |
+
font-size: 24px;
|
| 40 |
+
font-weight: bold;
|
| 41 |
+
}
|
| 42 |
+
.header .btn {
|
| 43 |
+
background-color: #444444;
|
| 44 |
+
color: white;
|
| 45 |
+
padding: 5px 10px;
|
| 46 |
+
border: none;
|
| 47 |
+
border-radius: 4px;
|
| 48 |
+
cursor: pointer;
|
| 49 |
+
font-size: 14px;
|
| 50 |
+
margin-left: 10px;
|
| 51 |
+
}
|
| 52 |
+
.header .btn:hover {
|
| 53 |
+
background-color: #555555;
|
| 54 |
+
}
|
| 55 |
+
.header .btn:disabled {
|
| 56 |
+
background-color: #888888;
|
| 57 |
+
cursor: not-allowed;
|
| 58 |
+
}
|
| 59 |
+
.header .btn:disabled:hover {
|
| 60 |
+
background-color: #888888;
|
| 61 |
+
}
|
| 62 |
+
.header .btn:disabled:active {
|
| 63 |
+
background-color: #888888;
|
| 64 |
+
}
|
| 65 |
+
.header .btn:disabled:active:hover {
|
| 66 |
+
background-color: #888888;
|
| 67 |
+
}
|
| 68 |
+
.header .btn:disabled:active:active {
|
| 69 |
+
background-color: #888888;
|
| 70 |
+
}
|
| 71 |
+
.header .btn:disabled:active:active:hover {
|
| 72 |
+
background-color: #888888;
|
| 73 |
+
}
|
| 74 |
+
.header .btn:disabled:active:active:active {
|
| 75 |
+
background-color: #888888;
|
| 76 |
+
}
|
| 77 |
+
.header .btn:disabled:active:active:hover {
|
| 78 |
+
background-color: #888888;
|
| 79 |
+
}
|
| 80 |
+
.header .btn:disabled:active:active:active {
|
| 81 |
+
background-color: #888888;
|
| 82 |
+
}
|
| 83 |
+
.header .btn:disabled:active:active:hover {
|
| 84 |
+
background-color: #888888;
|
| 85 |
+
}
|
| 86 |
+
.header .btn:disabled:active:active:active {
|
| 87 |
+
background-color: #888888;
|
| 88 |
+
}
|
| 89 |
+
.header .btn:disabled:active:active:hover {
|
| 90 |
+
background-color: #888888;
|
| 91 |
+
}
|
| 92 |
+
.header .btn:disabled:active:active:active {
|
| 93 |
+
background-color: #888888;
|
| 94 |
+
}
|
| 95 |
+
.header .btn:disabled:active:active:hover {
|
| 96 |
+
background-color: #888888;
|
| 97 |
+
}
|
| 98 |
+
.header .btn:disabled:active:active:active {
|
| 99 |
+
background-color: #888888;
|
| 100 |
+
}
|
| 101 |
+
.header .btn:disabled:active:active:hover {
|
| 102 |
+
background-color: #888888;
|
| 103 |
+
}
|
| 104 |
+
.header .btn:disabled:active:active:active {
|
| 105 |
+
background-color: #888888;
|
| 106 |
+
}
|
| 107 |
+
.header .btn:disabled:active:active:hover {
|
| 108 |
+
background-color: #888888;
|
| 109 |
+
}
|
| 110 |
+
.header .btn:disabled:active:active:active {
|
| 111 |
+
background-color: #888888;
|
| 112 |
+
}
|
| 113 |
+
.header .btn:disabled:active:active:hover {
|
| 114 |
+
background-color: #888888;
|
| 115 |
+
}
|
| 116 |
+
.header .btn:disabled:active:active:active {
|
| 117 |
+
background-color: #888888;
|
| 118 |
+
}
|
| 119 |
+
.header .btn:disabled:active:active:hover {
|
| 120 |
+
background-color: #888888;
|
| 121 |
+
}
|
| 122 |
+
.header .btn:disabled:active:active:active {
|
| 123 |
+
background-color: #888888;
|
| 124 |
+
}
|
| 125 |
+
.header .btn:disabled:active:active:hover {
|
| 126 |
+
background-color: #888888;
|
| 127 |
+
}
|
| 128 |
+
.header .btn:disabled:active:active:active {
|
| 129 |
+
background-color: #888888;
|
| 130 |
+
}
|
| 131 |
+
.header .btn:disabled:active:active:hover {
|
| 132 |
+
background-color: #888888;
|
| 133 |
+
}
|
| 134 |
+
.header .btn:disabled:active:active:active {
|
| 135 |
+
background-color: #888888;
|
| 136 |
+
}
|
| 137 |
+
.header .btn:disabled:active:active:hover {
|
| 138 |
+
background-color: #888888;
|
| 139 |
+
}
|
| 140 |
+
.header .btn:disabled:active:active:active {
|
| 141 |
+
background-color: #888888;
|
| 142 |
+
}
|
| 143 |
+
.header .btn:disabled:active:active:hover {
|
| 144 |
+
background-color: #888888;
|
| 145 |
+
}
|
| 146 |
+
.header .btn:disabled:active:active:active {
|
| 147 |
+
background-color: #888888;
|
| 148 |
+
}
|
| 149 |
+
.header .btn:disabled:active:active:hover {
|
| 150 |
+
background-color: #888888;
|
| 151 |
+
}
|
| 152 |
+
.header .btn:disabled:active:active:active {
|
| 153 |
+
background-color: #888888;
|
| 154 |
+
}
|
| 155 |
+
.header .btn:disabled:active:active:hover {
|
| 156 |
+
background-color: #888888;
|
| 157 |
+
}
|
| 158 |
+
.header .btn:disabled:active:active:active {
|
| 159 |
+
background-color: #888888;
|
| 160 |
+
}
|
| 161 |
+
.header .btn:disabled:active:active:hover {
|
| 162 |
+
background-color: #888888;
|
| 163 |
+
}
|
| 164 |
+
.header .btn:disabled:active:active:active {
|
| 165 |
+
background-color: #888888;
|
| 166 |
+
}
|
| 167 |
+
.header .btn:disabled:active:active:hover {
|
| 168 |
+
background-color: #888888;
|
| 169 |
+
}
|
| 170 |
+
.header .btn:disabled:active:active:active {
|
| 171 |
+
background-color: #888888;
|
| 172 |
+
}
|
| 173 |
+
.header .btn:disabled:active:active:hover {
|
| 174 |
+
background-color: #888888;
|
| 175 |
+
}
|
| 176 |
+
.header .btn:disabled:active:active:active {
|
| 177 |
+
background-color: #888888;
|
| 178 |
+
}
|
| 179 |
+
.header .btn:disabled:active:active:hover {
|
| 180 |
+
background-color: #888888;
|
| 181 |
+
}
|
| 182 |
+
.header .btn:disabled:active:active:active {
|
| 183 |
+
background-color: #888888;
|
| 184 |
+
}
|
| 185 |
+
.header .btn:disabled:active:active:hover {
|
| 186 |
+
background-color: #888888;
|
| 187 |
+
}
|
| 188 |
+
.header .btn:disabled:active:active:active {
|
| 189 |
+
background-color: #888888;
|
| 190 |
+
}
|
| 191 |
+
.header .btn:disabled:active:active:hover {
|
| 192 |
+
background-color: #888888;
|
| 193 |
+
}
|
| 194 |
+
.header .btn:disabled:active:active:active {
|
| 195 |
+
background-color: #888888;
|
| 196 |
+
}
|
| 197 |
+
.header .btn:disabled:active:active:hover {
|
| 198 |
+
background-color: #888888;
|
| 199 |
+
}
|
| 200 |
+
.header .btn:disabled:active:active:active {
|
| 201 |
+
background-color: #888888;
|
| 202 |
+
}
|
| 203 |
+
.header .btn:disabled:active:active:hover {
|
| 204 |
+
background-color: #888888;
|
| 205 |
+
}
|
| 206 |
+
.header .btn:disabled:active:active:active {
|
| 207 |
+
background-color: #888888;
|
| 208 |
+
}
|
| 209 |
+
.header .btn:disabled:active:active:hover {
|
| 210 |
+
background-color: #888888;
|
| 211 |
+
}
|
| 212 |
+
.header .btn:disabled:active:active:active {
|
| 213 |
+
background-color: #888888;
|
| 214 |
+
}
|
| 215 |
+
.header .btn:disabled:active:active:hover {
|
| 216 |
+
background-color: #888888;
|
| 217 |
+
}
|
| 218 |
+
.header .btn:disabled:active:active:active {
|
| 219 |
+
background-color: #888888;
|
| 220 |
+
}
|
| 221 |
+
.header .btn:disabled:active:active:hover {
|
| 222 |
+
background-color: #888888;
|
| 223 |
+
}
|
| 224 |
+
.header .btn:disabled:active:active:active {
|
| 225 |
+
background-color: #888888;
|
| 226 |
+
}
|
| 227 |
+
.header .btn:disabled:active:active:hover {
|
| 228 |
+
background-color: #888888;
|
| 229 |
+
}
|
| 230 |
+
.header .btn:disabled:active:active:active {
|
| 231 |
+
background-color: #888888;
|
| 232 |
+
}
|
| 233 |
+
.header .btn:disabled:active:active:hover {
|
| 234 |
+
background-color: #888888;
|
| 235 |
+
}
|
| 236 |
+
.header .btn:disabled:active:active:active {
|
| 237 |
+
background-color: #888888;
|
| 238 |
+
}
|
| 239 |
+
.header .btn:disabled:active:active:hover {
|
| 240 |
+
background-color: #888888;
|
| 241 |
+
}
|
| 242 |
+
.header .btn:disabled:active:active:active {
|
| 243 |
+
background-color: #888888;
|
| 244 |
+
}
|
| 245 |
+
.header .btn:disabled:active:active:hover {
|
| 246 |
+
background-color: #888888;
|
| 247 |
+
}
|
| 248 |
+
.header .btn:disabled:active:active:active {
|
| 249 |
+
background-color: #888888;
|
| 250 |
+
}
|
| 251 |
+
.header .btn:disabled:active:active:hover {
|
| 252 |
+
background-color: #888888;
|
| 253 |
+
}
|
| 254 |
+
.header .btn:disabled:active:active:active {
|
| 255 |
+
background-color: #888888;
|
| 256 |
+
}
|
| 257 |
+
.header .btn:disabled:active:active:hover {
|
| 258 |
+
background-color: #888888;
|
| 259 |
+
}
|
| 260 |
+
.header .btn:disabled:active:active:active {
|
| 261 |
+
background-color: #888888;
|
| 262 |
+
}
|
| 263 |
+
.header .btn:disabled:active:active:hover {
|
| 264 |
+
background-color: #888888;
|
| 265 |
+
}
|
| 266 |
+
.header .btn:disabled:active:active:active {
|
| 267 |
+
background-color: #888888;
|
| 268 |
+
}
|
| 269 |
+
.header .btn:disabled:active:active:hover {
|
| 270 |
+
background-color: #888888;
|
| 271 |
+
}
|
| 272 |
+
.header .btn:disabled:active:active:active {
|
| 273 |
+
background-color: #888888;
|
| 274 |
+
}
|
| 275 |
+
.header .btn:disabled:active:active:hover {
|
| 276 |
+
background-color: #888888;
|
| 277 |
+
}
|
| 278 |
+
.header .btn:disabled:active:active:active {
|
| 279 |
+
background-color: #888888;
|
| 280 |
+
}
|
| 281 |
+
.header .btn:disabled:active:active:hover {
|
| 282 |
+
background-color: #888888;
|
| 283 |
+
}
|
| 284 |
+
.header .btn:disabled:active:active:active {
|
| 285 |
+
background-color: #888888;
|
| 286 |
+
}
|
| 287 |
+
.header .btn:disabled:active:active:hover {
|
| 288 |
+
background-color: #888888;
|
| 289 |
+
}
|
| 290 |
+
.header .btn:disabled:active:active:active {
|
| 291 |
+
background-color: #888888;
|
| 292 |
+
}
|
| 293 |
+
.header .btn:disabled:active:active:hover {
|
| 294 |
+
background-color: #888888;
|
| 295 |
+
}
|
| 296 |
+
.header .btn:disabled:active:active:active {
|
| 297 |
+
background-color: #888888;
|
| 298 |
+
}
|
| 299 |
+
.header .btn:disabled:active:active:hover {
|
| 300 |
+
background-color: #888888;
|
| 301 |
+
}
|
| 302 |
+
.header .btn:disabled:active:active:active {
|
| 303 |
+
background-color: #888888;
|
| 304 |
+
}
|
| 305 |
+
.header .btn:disabled:active:active:hover {
|
| 306 |
+
background-color: #888888;
|
| 307 |
+
}
|
| 308 |
+
.header .btn:disabled:active:active:active {
|
| 309 |
+
background-color: #888888;
|
| 310 |
+
}
|
| 311 |
+
.header .btn:disabled:active:active:hover {
|
| 312 |
+
background-color: #888888;
|
| 313 |
+
}
|
| 314 |
+
.header .btn:disabled:active:active:active {
|
| 315 |
+
background-color: #888888;
|
| 316 |
+
}
|
| 317 |
+
.header .btn:disabled:active:active:hover {
|
| 318 |
+
background-color: #888888;
|
| 319 |
+
}
|
| 320 |
+
.header .btn:disabled:active:active:active {
|
| 321 |
+
background-color: #888888;
|
| 322 |
+
}
|
| 323 |
+
.header .btn:disabled:active:active:hover {
|
| 324 |
+
background-color: #888888;
|
| 325 |
+
}
|
| 326 |
+
.header .btn:disabled:active:active:active {
|
| 327 |
+
background-color: #888888;
|
| 328 |
+
}
|
| 329 |
+
.header .btn:disabled:active:active:hover {
|
| 330 |
+
background-color: #888888;
|
| 331 |
+
}
|
| 332 |
+
.header .btn:disabled:active:active:active {
|
| 333 |
+
background-color: #888888;
|
| 334 |
+
}
|
| 335 |
+
.header .btn:disabled:active:active:hover {
|
| 336 |
+
background-color: #888888;
|
| 337 |
+
}
|
| 338 |
+
.header .btn:disabled:active:active:active {
|
| 339 |
+
background-color: #888888;
|
| 340 |
+
}
|
| 341 |
+
.header .btn:disabled:active:active:hover {
|
| 342 |
+
background-color: #888888;
|
| 343 |
+
}
|
| 344 |
+
.header .btn:disabled:active:active:active {
|
| 345 |
+
background-color: #888888;
|
| 346 |
+
}
|
| 347 |
+
.header .btn:disabled:active:active:hover {
|
| 348 |
+
background-color: #888888;
|
| 349 |
+
}
|
| 350 |
+
.header .btn:disabled:active:active:active {
|
| 351 |
+
background-color: #888888;
|
| 352 |
+
}
|
| 353 |
+
.header .btn:disabled:active:active:hover {
|
| 354 |
+
background-color: #888888;
|
| 355 |
+
}
|
| 356 |
+
.header .btn:disabled:active:active:active {
|
| 357 |
+
background-color: #888888;
|
| 358 |
+
}
|
| 359 |
+
.header .btn:disabled:active:active:hover {
|
| 360 |
+
background-color: #888888;
|
| 361 |
+
}
|
| 362 |
+
.header .btn:disabled:active:active:active {
|
| 363 |
+
background-color: #888888;
|
| 364 |
+
}
|
| 365 |
+
.header .btn:disabled:active:active:hover {
|
| 366 |
+
background-color: #888888;
|
| 367 |
+
}
|
| 368 |
+
.header .btn:disabled:active:active:active {
|
| 369 |
+
background-color: #888888;
|
| 370 |
+
}
|
| 371 |
+
.header .btn:disabled:active:active:hover {
|
| 372 |
+
background-color: #888888;
|
| 373 |
+
}
|
| 374 |
+
.header .btn:disabled:active:active:active {
|
| 375 |
+
background-color: #888888;
|
| 376 |
+
}
|
| 377 |
+
.header .btn:disabled:active:active:hover {
|
| 378 |
+
background-color: #888888;
|
| 379 |
+
}
|
| 380 |
+
.header .btn:disabled:active:active:active {
|
| 381 |
+
background-color: #888888;
|
| 382 |
+
}
|
| 383 |
+
.header .btn:disabled:active:active:hover {
|
| 384 |
+
background-color: #888888;
|
| 385 |
+
}
|
| 386 |
+
.header .btn:disabled:active:active:active {
|
| 387 |
+
background-color: #888888;
|
| 388 |
+
}
|
| 389 |
+
.header .btn:disabled:active:active:hover {
|
| 390 |
+
background-color: #888888;
|
| 391 |
+
}
|
| 392 |
+
.header .btn:disabled:active:active:active {
|
| 393 |
+
background-color: #888888;
|
| 394 |
+
}
|
| 395 |
+
.header .btn:disabled:active:active:hover {
|
| 396 |
+
background-color: #888888;
|
| 397 |
+
}
|
| 398 |
+
.header .btn:disabled:active:active:active {
|
| 399 |
+
background-color: #888888;
|
| 400 |
+
}
|
| 401 |
+
.header .btn:disabled:active:active:hover {
|
| 402 |
+
background-color: #888888;
|
| 403 |
+
}
|
| 404 |
+
.header .btn:disabled:active:active:active {
|
| 405 |
+
background-color: #888888;
|
| 406 |
+
}
|
| 407 |
+
.header .btn:disabled:active:active:hover {
|
| 408 |
+
background-color: #888888;
|
| 409 |
+
}
|
| 410 |
+
.header .btn:disabled:active:active:active {
|
| 411 |
+
background-color: #888888;
|
| 412 |
+
}
|
| 413 |
+
.header .btn:disabled:active:active:hover {
|
| 414 |
+
background-color: #888888;
|
| 415 |
+
}
|
| 416 |
+
.header .btn:disabled:active:active:active {
|
| 417 |
+
background-color: #888888;
|
| 418 |
+
}
|
| 419 |
+
.header .btn:disabled:active:active:hover {
|
| 420 |
+
background-color: #888888;
|
| 421 |
+
}
|
| 422 |
+
.header .btn:disabled:active:active:active {
|
| 423 |
+
background-color: #888888;
|
| 424 |
+
}
|
| 425 |
+
.header .btn:disabled:active:active:hover {
|
| 426 |
+
background-color: #888888;
|
| 427 |
+
}
|
| 428 |
+
.header .btn:disabled:active:active:active {
|
| 429 |
+
background-color: #888888;
|
| 430 |
+
}
|
| 431 |
+
.header .btn:disabled:active:active:hover {
|
| 432 |
+
background-color: #888888;
|
| 433 |
+
}
|
| 434 |
+
.header .btn:disabled:active:active:active {
|
| 435 |
+
background-color: #888888;
|
| 436 |
+
}
|
| 437 |
+
.header .btn:disabled:active:active:hover {
|
| 438 |
+
background-color: #888888;
|
| 439 |
+
}
|
| 440 |
+
.header .btn:disabled:active:active:active {
|
| 441 |
+
background-color: #888888;
|
| 442 |
+
}
|
| 443 |
+
.header .btn:disabled:active:active:hover {
|
| 444 |
+
background-color: #888888;
|
| 445 |
+
}
|
| 446 |
+
.header .btn:disabled:active:active:active {
|
| 447 |
+
background-color: #888888;
|
| 448 |
+
}
|
| 449 |
+
.header .btn:disabled:active:active:hover {
|
| 450 |
+
background-color: #888888;
|
| 451 |
+
}
|
| 452 |
+
.header .btn:disabled:active:active:active {
|
| 453 |
+
background-color: #888888;
|
| 454 |
+
}
|
| 455 |
+
.header .btn:disabled:active:active:hover {
|
| 456 |
+
background-color: #888888;
|
| 457 |
+
}
|
| 458 |
+
.header .btn:disabled:active:active:active {
|
| 459 |
+
background-color: #888888;
|
| 460 |
+
}
|
| 461 |
+
.header .btn:disabled:active:active:hover {
|
| 462 |
+
background-color: #888888;
|
| 463 |
+
}
|
| 464 |
+
.header .btn:disabled
|
results/comparison/visionzip_256_quick2/visionzip_256/per_sample.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"id": "0",
|
| 4 |
+
"n_visual_tokens": 256,
|
| 5 |
+
"latency_s": 157.09,
|
| 6 |
+
"peak_mem_gb": 17.18,
|
| 7 |
+
"output_len": 15111
|
| 8 |
+
}
|
| 9 |
+
]
|
results/comparison/visionzip_256_quick2/visionzip_256/summary.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"method": "visionzip_256",
|
| 3 |
+
"n_samples": 1,
|
| 4 |
+
"n_success": 1,
|
| 5 |
+
"avg_visual_tokens": 256.0,
|
| 6 |
+
"avg_latency_s": 157.09,
|
| 7 |
+
"avg_peak_mem_gb": 17.18
|
| 8 |
+
}
|
results/element_analysis.json
ADDED
|
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"efficientui_prune60": {
|
| 3 |
+
"n_samples": 50,
|
| 4 |
+
"avg_text_f1": 0.6962,
|
| 5 |
+
"avg_element_f1": 0.7658,
|
| 6 |
+
"avg_dom_depth": 4.8,
|
| 7 |
+
"avg_dom_nodes": 25.8,
|
| 8 |
+
"avg_css_properties": 48.2,
|
| 9 |
+
"avg_css_unique_props": 15.5,
|
| 10 |
+
"per_category_f1": {
|
| 11 |
+
"buttons": 0.8213,
|
| 12 |
+
"inputs": 0.8941,
|
| 13 |
+
"images": 0.845,
|
| 14 |
+
"links": 0.6323,
|
| 15 |
+
"headings": 0.5976,
|
| 16 |
+
"lists": 0.7519,
|
| 17 |
+
"tables": 0.96,
|
| 18 |
+
"forms": 0.98,
|
| 19 |
+
"nav": 0.7691,
|
| 20 |
+
"containers": 0.6164,
|
| 21 |
+
"text_inline": 0.5557
|
| 22 |
+
}
|
| 23 |
+
},
|
| 24 |
+
"efficientui_prune80": {
|
| 25 |
+
"n_samples": 50,
|
| 26 |
+
"avg_text_f1": 0.5292,
|
| 27 |
+
"avg_element_f1": 0.6945,
|
| 28 |
+
"avg_dom_depth": 4.0,
|
| 29 |
+
"avg_dom_nodes": 19.5,
|
| 30 |
+
"avg_css_properties": 34.2,
|
| 31 |
+
"avg_css_unique_props": 10.1,
|
| 32 |
+
"per_category_f1": {
|
| 33 |
+
"buttons": 0.82,
|
| 34 |
+
"inputs": 0.8382,
|
| 35 |
+
"images": 0.82,
|
| 36 |
+
"links": 0.5088,
|
| 37 |
+
"headings": 0.421,
|
| 38 |
+
"lists": 0.6778,
|
| 39 |
+
"tables": 0.96,
|
| 40 |
+
"forms": 0.98,
|
| 41 |
+
"nav": 0.736,
|
| 42 |
+
"containers": 0.4202,
|
| 43 |
+
"text_inline": 0.4574
|
| 44 |
+
}
|
| 45 |
+
},
|
| 46 |
+
"qwen3_full": {
|
| 47 |
+
"n_samples": 50,
|
| 48 |
+
"avg_text_f1": 1.0,
|
| 49 |
+
"avg_element_f1": 1.0,
|
| 50 |
+
"avg_dom_depth": 5.7,
|
| 51 |
+
"avg_dom_nodes": 43.6,
|
| 52 |
+
"avg_css_properties": 55.1,
|
| 53 |
+
"avg_css_unique_props": 19.6,
|
| 54 |
+
"per_category_f1": {
|
| 55 |
+
"buttons": 1.0,
|
| 56 |
+
"inputs": 1.0,
|
| 57 |
+
"images": 1.0,
|
| 58 |
+
"links": 1.0,
|
| 59 |
+
"headings": 1.0,
|
| 60 |
+
"lists": 1.0,
|
| 61 |
+
"tables": 1.0,
|
| 62 |
+
"forms": 1.0,
|
| 63 |
+
"nav": 1.0,
|
| 64 |
+
"containers": 1.0,
|
| 65 |
+
"text_inline": 1.0
|
| 66 |
+
}
|
| 67 |
+
},
|
| 68 |
+
"qwen3_res_1003520": {
|
| 69 |
+
"n_samples": 50,
|
| 70 |
+
"avg_text_f1": 0.8869,
|
| 71 |
+
"avg_element_f1": 0.8835,
|
| 72 |
+
"avg_dom_depth": 5.6,
|
| 73 |
+
"avg_dom_nodes": 40.0,
|
| 74 |
+
"avg_css_properties": 59.7,
|
| 75 |
+
"avg_css_unique_props": 20.3,
|
| 76 |
+
"per_category_f1": {
|
| 77 |
+
"buttons": 0.9048,
|
| 78 |
+
"inputs": 0.9533,
|
| 79 |
+
"images": 0.934,
|
| 80 |
+
"links": 0.8491,
|
| 81 |
+
"headings": 0.7725,
|
| 82 |
+
"lists": 0.8799,
|
| 83 |
+
"tables": 0.98,
|
| 84 |
+
"forms": 0.98,
|
| 85 |
+
"nav": 0.8391,
|
| 86 |
+
"containers": 0.8288,
|
| 87 |
+
"text_inline": 0.7974
|
| 88 |
+
}
|
| 89 |
+
},
|
| 90 |
+
"qwen3_res_230400": {
|
| 91 |
+
"n_samples": 50,
|
| 92 |
+
"avg_text_f1": 0.6244,
|
| 93 |
+
"avg_element_f1": 0.7481,
|
| 94 |
+
"avg_dom_depth": 4.6,
|
| 95 |
+
"avg_dom_nodes": 21.8,
|
| 96 |
+
"avg_css_properties": 32.0,
|
| 97 |
+
"avg_css_unique_props": 12.6,
|
| 98 |
+
"per_category_f1": {
|
| 99 |
+
"buttons": 0.8433,
|
| 100 |
+
"inputs": 0.8585,
|
| 101 |
+
"images": 0.836,
|
| 102 |
+
"links": 0.5918,
|
| 103 |
+
"headings": 0.5942,
|
| 104 |
+
"lists": 0.7042,
|
| 105 |
+
"tables": 0.96,
|
| 106 |
+
"forms": 0.96,
|
| 107 |
+
"nav": 0.732,
|
| 108 |
+
"containers": 0.5961,
|
| 109 |
+
"text_inline": 0.5526
|
| 110 |
+
}
|
| 111 |
+
},
|
| 112 |
+
"uipress_256": {
|
| 113 |
+
"n_samples": 50,
|
| 114 |
+
"avg_text_f1": 0.2091,
|
| 115 |
+
"avg_element_f1": 0.6371,
|
| 116 |
+
"avg_dom_depth": 3.9,
|
| 117 |
+
"avg_dom_nodes": 18.4,
|
| 118 |
+
"avg_css_properties": 0.1,
|
| 119 |
+
"avg_css_unique_props": 0.1,
|
| 120 |
+
"per_category_f1": {
|
| 121 |
+
"buttons": 0.74,
|
| 122 |
+
"inputs": 0.8,
|
| 123 |
+
"images": 0.7,
|
| 124 |
+
"links": 0.3995,
|
| 125 |
+
"headings": 0.4197,
|
| 126 |
+
"lists": 0.617,
|
| 127 |
+
"tables": 0.96,
|
| 128 |
+
"forms": 0.96,
|
| 129 |
+
"nav": 0.72,
|
| 130 |
+
"containers": 0.3521,
|
| 131 |
+
"text_inline": 0.3395
|
| 132 |
+
}
|
| 133 |
+
},
|
| 134 |
+
"visionzip_128": {
|
| 135 |
+
"n_samples": 50,
|
| 136 |
+
"avg_text_f1": 0.2804,
|
| 137 |
+
"avg_element_f1": 0.6619,
|
| 138 |
+
"avg_dom_depth": 3.6,
|
| 139 |
+
"avg_dom_nodes": 23.3,
|
| 140 |
+
"avg_css_properties": 23.0,
|
| 141 |
+
"avg_css_unique_props": 8.3,
|
| 142 |
+
"per_category_f1": {
|
| 143 |
+
"buttons": 0.75,
|
| 144 |
+
"inputs": 0.8,
|
| 145 |
+
"images": 0.78,
|
| 146 |
+
"links": 0.437,
|
| 147 |
+
"headings": 0.3963,
|
| 148 |
+
"lists": 0.68,
|
| 149 |
+
"tables": 0.96,
|
| 150 |
+
"forms": 0.94,
|
| 151 |
+
"nav": 0.7533,
|
| 152 |
+
"containers": 0.4076,
|
| 153 |
+
"text_inline": 0.3763
|
| 154 |
+
}
|
| 155 |
+
},
|
| 156 |
+
"visionzip_256": {
|
| 157 |
+
"n_samples": 50,
|
| 158 |
+
"avg_text_f1": 0.3574,
|
| 159 |
+
"avg_element_f1": 0.6661,
|
| 160 |
+
"avg_dom_depth": 3.3,
|
| 161 |
+
"avg_dom_nodes": 9.4,
|
| 162 |
+
"avg_css_properties": 13.8,
|
| 163 |
+
"avg_css_unique_props": 6.1,
|
| 164 |
+
"per_category_f1": {
|
| 165 |
+
"buttons": 0.8,
|
| 166 |
+
"inputs": 0.78,
|
| 167 |
+
"images": 0.8,
|
| 168 |
+
"links": 0.4269,
|
| 169 |
+
"headings": 0.3264,
|
| 170 |
+
"lists": 0.715,
|
| 171 |
+
"tables": 0.96,
|
| 172 |
+
"forms": 0.96,
|
| 173 |
+
"nav": 0.74,
|
| 174 |
+
"containers": 0.3956,
|
| 175 |
+
"text_inline": 0.4234
|
| 176 |
+
}
|
| 177 |
+
},
|
| 178 |
+
"visionzip_64": {
|
| 179 |
+
"n_samples": 50,
|
| 180 |
+
"avg_text_f1": 0.1966,
|
| 181 |
+
"avg_element_f1": 0.6567,
|
| 182 |
+
"avg_dom_depth": 2.9,
|
| 183 |
+
"avg_dom_nodes": 12.4,
|
| 184 |
+
"avg_css_properties": 11.1,
|
| 185 |
+
"avg_css_unique_props": 5.5,
|
| 186 |
+
"per_category_f1": {
|
| 187 |
+
"buttons": 0.78,
|
| 188 |
+
"inputs": 0.82,
|
| 189 |
+
"images": 0.8,
|
| 190 |
+
"links": 0.3603,
|
| 191 |
+
"headings": 0.3433,
|
| 192 |
+
"lists": 0.68,
|
| 193 |
+
"tables": 0.96,
|
| 194 |
+
"forms": 0.96,
|
| 195 |
+
"nav": 0.74,
|
| 196 |
+
"containers": 0.3583,
|
| 197 |
+
"text_inline": 0.4214
|
| 198 |
+
}
|
| 199 |
+
}
|
| 200 |
+
}
|
scripts/eval_all.py
CHANGED
|
@@ -43,6 +43,13 @@ from tqdm import tqdm
|
|
| 43 |
PROJECT_ROOT = Path(__file__).parent.parent
|
| 44 |
sys.path.insert(0, str(PROJECT_ROOT))
|
| 45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
UI2CODE_PROMPT = (
|
| 47 |
"Convert this webpage screenshot to HTML code. "
|
| 48 |
"Generate a complete, self-contained HTML file with inline CSS. "
|
|
@@ -178,42 +185,107 @@ class VisionZipMethod:
|
|
| 178 |
self._install_hook()
|
| 179 |
|
| 180 |
def _install_hook(self):
|
| 181 |
-
"""
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
def
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
|
| 198 |
-
|
| 199 |
-
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
| 203 |
-
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
|
| 207 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208 |
else:
|
| 209 |
-
|
| 210 |
-
|
| 211 |
-
|
| 212 |
-
|
| 213 |
-
|
| 214 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 215 |
|
| 216 |
-
|
|
|
|
| 217 |
|
| 218 |
def _select_tokens(self, tokens):
|
| 219 |
"""Select dominant + contextual tokens based on L2 norm (proxy for attention)."""
|
|
@@ -264,33 +336,7 @@ class VisionZipMethod:
|
|
| 264 |
torch.cuda.reset_peak_memory_stats()
|
| 265 |
t0 = time.time()
|
| 266 |
|
| 267 |
-
# We need to handle the mismatch between expected and actual
|
| 268 |
-
# visual token count. Override image_grid_thw after visual processing.
|
| 269 |
-
orig_grid = inputs.get("image_grid_thw", None)
|
| 270 |
-
|
| 271 |
with torch.no_grad():
|
| 272 |
-
# Process pixel_values through visual encoder (hook will compress)
|
| 273 |
-
pixel_values = inputs.get("pixel_values")
|
| 274 |
-
if pixel_values is not None:
|
| 275 |
-
visual_out = self.model.visual(
|
| 276 |
-
pixel_values, grid_thw=orig_grid,
|
| 277 |
-
)
|
| 278 |
-
# Now rebuild inputs with compressed token count
|
| 279 |
-
n_compressed = visual_out.shape[0]
|
| 280 |
-
new_grid = self._new_grid_thw
|
| 281 |
-
|
| 282 |
-
# Rebuild input_ids with correct number of image tokens
|
| 283 |
-
input_ids = inputs["input_ids"][0].tolist()
|
| 284 |
-
img_positions = [i for i, t in enumerate(input_ids) if t == IMAGE_TOKEN_ID]
|
| 285 |
-
if img_positions:
|
| 286 |
-
before = input_ids[:img_positions[0]]
|
| 287 |
-
after = input_ids[img_positions[-1] + 1:]
|
| 288 |
-
new_ids = before + [IMAGE_TOKEN_ID] * n_compressed + after
|
| 289 |
-
inputs["input_ids"] = torch.tensor([new_ids], device=self.model.device)
|
| 290 |
-
inputs["image_grid_thw"] = new_grid
|
| 291 |
-
if "attention_mask" in inputs:
|
| 292 |
-
inputs["attention_mask"] = torch.ones_like(inputs["input_ids"])
|
| 293 |
-
|
| 294 |
out = self.model.generate(
|
| 295 |
**inputs, max_new_tokens=4096,
|
| 296 |
temperature=0.1, do_sample=True, top_p=0.9,
|
|
@@ -364,58 +410,110 @@ class EfficientUIMethod:
|
|
| 364 |
return importance
|
| 365 |
|
| 366 |
def _install_hook(self):
|
| 367 |
-
"""
|
| 368 |
-
|
| 369 |
-
|
| 370 |
-
|
| 371 |
-
|
| 372 |
-
|
| 373 |
-
|
| 374 |
-
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
|
| 379 |
-
|
| 380 |
-
|
| 381 |
-
|
| 382 |
-
|
| 383 |
-
|
| 384 |
-
|
| 385 |
-
|
| 386 |
-
|
| 387 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 388 |
)
|
| 389 |
-
|
| 390 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 391 |
)
|
| 392 |
|
| 393 |
-
|
| 394 |
-
|
| 395 |
-
|
| 396 |
-
|
| 397 |
-
|
| 398 |
-
|
| 399 |
-
|
| 400 |
-
|
| 401 |
-
|
| 402 |
-
|
| 403 |
-
# Compute new grid
|
| 404 |
-
n_kept = compressed_list[0].shape[0]
|
| 405 |
-
sqrt_n = int(n_kept ** 0.5)
|
| 406 |
-
for h in range(sqrt_n, 0, -1):
|
| 407 |
-
if n_kept % h == 0:
|
| 408 |
-
w = n_kept // h
|
| 409 |
-
break
|
| 410 |
else:
|
| 411 |
-
|
| 412 |
-
|
| 413 |
-
|
| 414 |
-
|
| 415 |
-
|
| 416 |
-
|
|
|
|
|
|
|
| 417 |
|
| 418 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 419 |
|
| 420 |
def generate(self, image):
|
| 421 |
self._current_image = image
|
|
@@ -432,24 +530,6 @@ class EfficientUIMethod:
|
|
| 432 |
t0 = time.time()
|
| 433 |
|
| 434 |
with torch.no_grad():
|
| 435 |
-
pixel_values = inputs.get("pixel_values")
|
| 436 |
-
orig_grid = inputs.get("image_grid_thw")
|
| 437 |
-
|
| 438 |
-
if pixel_values is not None:
|
| 439 |
-
visual_out = self.model.visual(pixel_values, grid_thw=orig_grid)
|
| 440 |
-
n_compressed = visual_out.shape[0]
|
| 441 |
-
|
| 442 |
-
input_ids = inputs["input_ids"][0].tolist()
|
| 443 |
-
img_pos = [i for i, t in enumerate(input_ids) if t == IMAGE_TOKEN_ID]
|
| 444 |
-
if img_pos:
|
| 445 |
-
before = input_ids[:img_pos[0]]
|
| 446 |
-
after = input_ids[img_pos[-1] + 1:]
|
| 447 |
-
new_ids = before + [IMAGE_TOKEN_ID] * n_compressed + after
|
| 448 |
-
inputs["input_ids"] = torch.tensor([new_ids], device=self.model.device)
|
| 449 |
-
inputs["image_grid_thw"] = self._new_grid
|
| 450 |
-
if "attention_mask" in inputs:
|
| 451 |
-
inputs["attention_mask"] = torch.ones_like(inputs["input_ids"])
|
| 452 |
-
|
| 453 |
out = self.model.generate(
|
| 454 |
**inputs, max_new_tokens=4096,
|
| 455 |
temperature=0.1, do_sample=True, top_p=0.9,
|
|
@@ -488,7 +568,7 @@ class UIPressMethod:
|
|
| 488 |
)
|
| 489 |
|
| 490 |
# Load compressor
|
| 491 |
-
llm_hidden = self.model
|
| 492 |
self.compressor = OpticalCompressor(
|
| 493 |
hidden_dim=llm_hidden, target_tokens=target_tokens,
|
| 494 |
).to(self.model.device, torch.bfloat16).eval()
|
|
@@ -496,7 +576,11 @@ class UIPressMethod:
|
|
| 496 |
ckpt = torch.load(checkpoint, map_location=self.model.device)
|
| 497 |
comp_state = ckpt.get("compressor", ckpt)
|
| 498 |
clean_state = {k.replace("module.", ""): v for k, v in comp_state.items()}
|
| 499 |
-
self.compressor.load_state_dict(clean_state)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 500 |
|
| 501 |
# Load LoRA if present
|
| 502 |
if "lora" in ckpt:
|
|
@@ -509,16 +593,95 @@ class UIPressMethod:
|
|
| 509 |
self._install_hook()
|
| 510 |
|
| 511 |
def _install_hook(self):
|
| 512 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 513 |
|
| 514 |
-
|
| 515 |
-
|
| 516 |
-
|
| 517 |
-
|
| 518 |
-
|
| 519 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 520 |
|
| 521 |
-
|
|
|
|
| 522 |
|
| 523 |
def generate(self, image):
|
| 524 |
messages = [{"role": "user", "content": [
|
|
@@ -534,24 +697,6 @@ class UIPressMethod:
|
|
| 534 |
t0 = time.time()
|
| 535 |
|
| 536 |
with torch.no_grad():
|
| 537 |
-
pixel_values = inputs.get("pixel_values")
|
| 538 |
-
orig_grid = inputs.get("image_grid_thw")
|
| 539 |
-
|
| 540 |
-
if pixel_values is not None:
|
| 541 |
-
visual_out = self.model.visual(pixel_values, grid_thw=orig_grid)
|
| 542 |
-
n_compressed = visual_out.shape[0]
|
| 543 |
-
|
| 544 |
-
input_ids = inputs["input_ids"][0].tolist()
|
| 545 |
-
img_pos = [i for i, t in enumerate(input_ids) if t == IMAGE_TOKEN_ID]
|
| 546 |
-
if img_pos:
|
| 547 |
-
before = input_ids[:img_pos[0]]
|
| 548 |
-
after = input_ids[img_pos[-1] + 1:]
|
| 549 |
-
new_ids = before + [IMAGE_TOKEN_ID] * n_compressed + after
|
| 550 |
-
inputs["input_ids"] = torch.tensor([new_ids], device=self.model.device)
|
| 551 |
-
inputs["image_grid_thw"] = self._new_grid
|
| 552 |
-
if "attention_mask" in inputs:
|
| 553 |
-
inputs["attention_mask"] = torch.ones_like(inputs["input_ids"])
|
| 554 |
-
|
| 555 |
out = self.model.generate(
|
| 556 |
**inputs, max_new_tokens=4096,
|
| 557 |
temperature=0.1, do_sample=True, top_p=0.9,
|
|
|
|
| 43 |
PROJECT_ROOT = Path(__file__).parent.parent
|
| 44 |
sys.path.insert(0, str(PROJECT_ROOT))
|
| 45 |
|
| 46 |
+
from models.qwen3_vl_compat import get_visual_module, set_visual_module
|
| 47 |
+
|
| 48 |
+
|
| 49 |
+
def _llm_hidden(model):
|
| 50 |
+
cfg = model.config
|
| 51 |
+
return cfg.text_config.hidden_size if hasattr(cfg, "text_config") else cfg.hidden_size
|
| 52 |
+
|
| 53 |
UI2CODE_PROMPT = (
|
| 54 |
"Convert this webpage screenshot to HTML code. "
|
| 55 |
"Generate a complete, self-contained HTML file with inline CSS. "
|
|
|
|
| 185 |
self._install_hook()
|
| 186 |
|
| 187 |
def _install_hook(self):
|
| 188 |
+
"""Patch Qwen3VLModel.forward to inject compression and fix masked_scatter dimension."""
|
| 189 |
+
self_ = self
|
| 190 |
+
|
| 191 |
+
def hooked_forward(self, input_ids=None, attention_mask=None, position_ids=None,
|
| 192 |
+
past_key_values=None, inputs_embeds=None,
|
| 193 |
+
pixel_values=None, pixel_values_videos=None,
|
| 194 |
+
image_grid_thw=None, video_grid_thw=None,
|
| 195 |
+
mm_token_type_ids=None, cache_position=None, **kwargs):
|
| 196 |
+
if pixel_values is not None and image_grid_thw is not None:
|
| 197 |
+
vo = self.get_image_features(pixel_values, image_grid_thw, return_dict=True)
|
| 198 |
+
pooler = vo.pooler_output
|
| 199 |
+
flat = torch.cat(pooler, dim=0) if isinstance(pooler, (list, tuple)) else pooler
|
| 200 |
+
|
| 201 |
+
# Step 2: compute LLM grid and compress
|
| 202 |
+
sms = self_._get_spatial_merge_size()
|
| 203 |
+
gl = image_grid_thw.clone()
|
| 204 |
+
gl[:, 1] = gl[:, 1] // sms
|
| 205 |
+
gl[:, 2] = gl[:, 2] // sms
|
| 206 |
+
|
| 207 |
+
parts, offset = [], 0
|
| 208 |
+
for i in range(image_grid_thw.shape[0]):
|
| 209 |
+
t, h, w = gl[i].tolist()
|
| 210 |
+
n = int(t) * int(h) * int(w)
|
| 211 |
+
tok = flat[offset:offset + n]
|
| 212 |
+
offset += n
|
| 213 |
+
parts.append(self_._select_tokens(tok))
|
| 214 |
+
|
| 215 |
+
comp = torch.cat(parts, dim=0)
|
| 216 |
+
|
| 217 |
+
# Step 3: new grid_thw for position IDs
|
| 218 |
+
k = self_.keep_tokens
|
| 219 |
+
sq = int(k ** 0.5)
|
| 220 |
+
for hh in range(sq, 0, -1):
|
| 221 |
+
if k % hh == 0:
|
| 222 |
+
ww = k // hh
|
| 223 |
+
break
|
| 224 |
+
else:
|
| 225 |
+
hh, ww = k, 1
|
| 226 |
+
# compute_3d_position_ids expects pre-merge grid, so scale back.
|
| 227 |
+
new_grid = torch.tensor(
|
| 228 |
+
[[1, hh * sms, ww * sms]] * image_grid_thw.shape[0],
|
| 229 |
+
device=image_grid_thw.device, dtype=image_grid_thw.dtype,
|
| 230 |
+
)
|
| 231 |
+
self_._new_grid_thw = new_grid
|
| 232 |
+
|
| 233 |
+
# Step 4: build inputs_embeds and substitute image positions
|
| 234 |
+
if inputs_embeds is None:
|
| 235 |
+
inputs_embeds = self.get_input_embeddings()(input_ids)
|
| 236 |
+
|
| 237 |
+
B, S, D = inputs_embeds.shape
|
| 238 |
+
flat_embeds = inputs_embeds.reshape(B * S, D).contiguous()
|
| 239 |
+
image_mask_2d = (input_ids == self.config.image_token_id) # [B, S]
|
| 240 |
+
image_mask_flat = image_mask_2d.reshape(B * S) # [B*S]
|
| 241 |
+
flat_indices = image_mask_flat.nonzero(as_tuple=True)[0] # [N_orig]
|
| 242 |
+
n_comp = comp.shape[0]
|
| 243 |
+
flat_indices = flat_indices[:n_comp]
|
| 244 |
+
|
| 245 |
+
flat_embeds[flat_indices] = comp.to(inputs_embeds.dtype)
|
| 246 |
+
inputs_embeds = flat_embeds.view(B, S, D)
|
| 247 |
+
|
| 248 |
+
# Mark only injected compressed positions as image tokens for RoPE.
|
| 249 |
+
rope_mm_token_type_ids = torch.zeros_like(input_ids, dtype=torch.int)
|
| 250 |
+
batch_idx = flat_indices // S
|
| 251 |
+
seq_idx = flat_indices % S
|
| 252 |
+
rope_mm_token_type_ids[batch_idx, seq_idx] = 1
|
| 253 |
+
|
| 254 |
+
# Step 5: compute position_ids with new_grid
|
| 255 |
+
position_ids = self.compute_3d_position_ids(
|
| 256 |
+
input_ids=input_ids, image_grid_thw=new_grid,
|
| 257 |
+
video_grid_thw=video_grid_thw, inputs_embeds=inputs_embeds,
|
| 258 |
+
attention_mask=attention_mask, past_key_values=past_key_values,
|
| 259 |
+
mm_token_type_ids=rope_mm_token_type_ids,
|
| 260 |
+
)
|
| 261 |
+
|
| 262 |
+
# Step 6: call language_model
|
| 263 |
+
outputs = self.language_model(
|
| 264 |
+
input_ids=None, position_ids=position_ids,
|
| 265 |
+
attention_mask=attention_mask, past_key_values=past_key_values,
|
| 266 |
+
inputs_embeds=inputs_embeds, cache_position=cache_position,
|
| 267 |
+
visual_pos_masks=None, deepstack_visual_embeds=None, **kwargs,
|
| 268 |
+
)
|
| 269 |
+
from transformers.models.qwen3_vl.modeling_qwen3_vl import Qwen3VLModelOutputWithPast
|
| 270 |
+
return Qwen3VLModelOutputWithPast(
|
| 271 |
+
**outputs, rope_deltas=getattr(self, 'rope_deltas', None)
|
| 272 |
+
)
|
| 273 |
else:
|
| 274 |
+
return self_._orig_fwd(
|
| 275 |
+
input_ids=input_ids, attention_mask=attention_mask,
|
| 276 |
+
position_ids=position_ids, past_key_values=past_key_values,
|
| 277 |
+
inputs_embeds=inputs_embeds, pixel_values=pixel_values,
|
| 278 |
+
pixel_values_videos=pixel_values_videos,
|
| 279 |
+
image_grid_thw=image_grid_thw, video_grid_thw=video_grid_thw,
|
| 280 |
+
mm_token_type_ids=mm_token_type_ids, cache_position=cache_position, **kwargs
|
| 281 |
+
)
|
| 282 |
+
|
| 283 |
+
import types
|
| 284 |
+
self._orig_fwd = self.model.model.forward
|
| 285 |
+
self.model.model.forward = types.MethodType(hooked_forward, self.model.model)
|
| 286 |
|
| 287 |
+
def _get_spatial_merge_size(self):
|
| 288 |
+
return self.model.model.visual.spatial_merge_size
|
| 289 |
|
| 290 |
def _select_tokens(self, tokens):
|
| 291 |
"""Select dominant + contextual tokens based on L2 norm (proxy for attention)."""
|
|
|
|
| 336 |
torch.cuda.reset_peak_memory_stats()
|
| 337 |
t0 = time.time()
|
| 338 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 339 |
with torch.no_grad():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 340 |
out = self.model.generate(
|
| 341 |
**inputs, max_new_tokens=4096,
|
| 342 |
temperature=0.1, do_sample=True, top_p=0.9,
|
|
|
|
| 410 |
return importance
|
| 411 |
|
| 412 |
def _install_hook(self):
|
| 413 |
+
"""Patch Qwen3VLModel.forward to prune tokens by element importance."""
|
| 414 |
+
self_ = self
|
| 415 |
+
|
| 416 |
+
def hooked_forward(self, input_ids=None, attention_mask=None, position_ids=None,
|
| 417 |
+
past_key_values=None, inputs_embeds=None,
|
| 418 |
+
pixel_values=None, pixel_values_videos=None,
|
| 419 |
+
image_grid_thw=None, video_grid_thw=None,
|
| 420 |
+
mm_token_type_ids=None, cache_position=None, **kwargs):
|
| 421 |
+
if pixel_values is not None and image_grid_thw is not None:
|
| 422 |
+
vo = self.get_image_features(pixel_values, image_grid_thw, return_dict=True)
|
| 423 |
+
pooler = vo.pooler_output
|
| 424 |
+
flat = torch.cat(pooler, dim=0) if isinstance(pooler, (list, tuple)) else pooler
|
| 425 |
+
|
| 426 |
+
sms = self_._get_spatial_merge_size()
|
| 427 |
+
gl = image_grid_thw.clone()
|
| 428 |
+
gl[:, 1] = gl[:, 1] // sms
|
| 429 |
+
gl[:, 2] = gl[:, 2] // sms
|
| 430 |
+
|
| 431 |
+
parts, offset = [], 0
|
| 432 |
+
for i in range(image_grid_thw.shape[0]):
|
| 433 |
+
t, h, w = gl[i].tolist()
|
| 434 |
+
n = int(t) * int(h) * int(w)
|
| 435 |
+
tok = flat[offset:offset + n]
|
| 436 |
+
offset += n
|
| 437 |
+
cur_img = self_._current_image
|
| 438 |
+
if cur_img is None:
|
| 439 |
+
cur_img = Image.new("RGB", (224, 224))
|
| 440 |
+
imp = self_._compute_element_mask(cur_img, int(h), int(w))
|
| 441 |
+
imp_flat = torch.tensor(imp.flatten(), device=tok.device)
|
| 442 |
+
n_keep = max(int(n * (1 - self_.prune_ratio)), 16)
|
| 443 |
+
_, top_idx = imp_flat.topk(n_keep)
|
| 444 |
+
top_idx, _ = top_idx.sort()
|
| 445 |
+
parts.append(tok[top_idx])
|
| 446 |
+
|
| 447 |
+
comp = torch.cat(parts, dim=0)
|
| 448 |
+
self_._n_kept = int(comp.shape[0])
|
| 449 |
+
nk = parts[0].shape[0]
|
| 450 |
+
sq = int(nk ** 0.5)
|
| 451 |
+
for hh in range(sq, 0, -1):
|
| 452 |
+
if nk % hh == 0:
|
| 453 |
+
ww = nk // hh
|
| 454 |
+
break
|
| 455 |
+
else:
|
| 456 |
+
hh, ww = nk, 1
|
| 457 |
+
# compute_3d_position_ids expects pre-merge grid, so scale back.
|
| 458 |
+
new_grid = torch.tensor(
|
| 459 |
+
[[1, hh * sms, ww * sms]] * image_grid_thw.shape[0],
|
| 460 |
+
device=image_grid_thw.device, dtype=image_grid_thw.dtype,
|
| 461 |
)
|
| 462 |
+
self_._new_grid = new_grid
|
| 463 |
+
|
| 464 |
+
if inputs_embeds is None:
|
| 465 |
+
inputs_embeds = self.get_input_embeddings()(input_ids)
|
| 466 |
+
|
| 467 |
+
B, S, D = inputs_embeds.shape
|
| 468 |
+
flat_embeds = inputs_embeds.reshape(B * S, D).contiguous()
|
| 469 |
+
image_mask_2d = (input_ids == self.config.image_token_id) # [B, S]
|
| 470 |
+
image_mask_flat = image_mask_2d.reshape(B * S) # [B*S]
|
| 471 |
+
flat_indices = image_mask_flat.nonzero(as_tuple=True)[0] # [N_orig]
|
| 472 |
+
n_comp = comp.shape[0]
|
| 473 |
+
flat_indices = flat_indices[:n_comp]
|
| 474 |
+
|
| 475 |
+
flat_embeds[flat_indices] = comp.to(inputs_embeds.dtype)
|
| 476 |
+
inputs_embeds = flat_embeds.view(B, S, D)
|
| 477 |
+
|
| 478 |
+
# Mark only injected compressed positions as image tokens for RoPE.
|
| 479 |
+
rope_mm_token_type_ids = torch.zeros_like(input_ids, dtype=torch.int)
|
| 480 |
+
batch_idx = flat_indices // S
|
| 481 |
+
seq_idx = flat_indices % S
|
| 482 |
+
rope_mm_token_type_ids[batch_idx, seq_idx] = 1
|
| 483 |
+
|
| 484 |
+
position_ids = self.compute_3d_position_ids(
|
| 485 |
+
input_ids=input_ids, image_grid_thw=new_grid,
|
| 486 |
+
video_grid_thw=video_grid_thw, inputs_embeds=inputs_embeds,
|
| 487 |
+
attention_mask=attention_mask, past_key_values=past_key_values,
|
| 488 |
+
mm_token_type_ids=rope_mm_token_type_ids,
|
| 489 |
)
|
| 490 |
|
| 491 |
+
outputs = self.language_model(
|
| 492 |
+
input_ids=None, position_ids=position_ids,
|
| 493 |
+
attention_mask=attention_mask, past_key_values=past_key_values,
|
| 494 |
+
inputs_embeds=inputs_embeds, cache_position=cache_position,
|
| 495 |
+
visual_pos_masks=None, deepstack_visual_embeds=None, **kwargs,
|
| 496 |
+
)
|
| 497 |
+
from transformers.models.qwen3_vl.modeling_qwen3_vl import Qwen3VLModelOutputWithPast
|
| 498 |
+
return Qwen3VLModelOutputWithPast(
|
| 499 |
+
**outputs, rope_deltas=getattr(self, 'rope_deltas', None)
|
| 500 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 501 |
else:
|
| 502 |
+
return self_._orig_fwd(
|
| 503 |
+
input_ids=input_ids, attention_mask=attention_mask,
|
| 504 |
+
position_ids=position_ids, past_key_values=past_key_values,
|
| 505 |
+
inputs_embeds=inputs_embeds, pixel_values=pixel_values,
|
| 506 |
+
pixel_values_videos=pixel_values_videos,
|
| 507 |
+
image_grid_thw=image_grid_thw, video_grid_thw=video_grid_thw,
|
| 508 |
+
mm_token_type_ids=mm_token_type_ids, cache_position=cache_position, **kwargs
|
| 509 |
+
)
|
| 510 |
|
| 511 |
+
import types
|
| 512 |
+
self._orig_fwd = self.model.model.forward
|
| 513 |
+
self.model.model.forward = types.MethodType(hooked_forward, self.model.model)
|
| 514 |
+
|
| 515 |
+
def _get_spatial_merge_size(self):
|
| 516 |
+
return self.model.model.visual.spatial_merge_size
|
| 517 |
|
| 518 |
def generate(self, image):
|
| 519 |
self._current_image = image
|
|
|
|
| 530 |
t0 = time.time()
|
| 531 |
|
| 532 |
with torch.no_grad():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 533 |
out = self.model.generate(
|
| 534 |
**inputs, max_new_tokens=4096,
|
| 535 |
temperature=0.1, do_sample=True, top_p=0.9,
|
|
|
|
| 568 |
)
|
| 569 |
|
| 570 |
# Load compressor
|
| 571 |
+
llm_hidden = _llm_hidden(self.model)
|
| 572 |
self.compressor = OpticalCompressor(
|
| 573 |
hidden_dim=llm_hidden, target_tokens=target_tokens,
|
| 574 |
).to(self.model.device, torch.bfloat16).eval()
|
|
|
|
| 576 |
ckpt = torch.load(checkpoint, map_location=self.model.device)
|
| 577 |
comp_state = ckpt.get("compressor", ckpt)
|
| 578 |
clean_state = {k.replace("module.", ""): v for k, v in comp_state.items()}
|
| 579 |
+
missing, unexpected = self.compressor.load_state_dict(clean_state, strict=False)
|
| 580 |
+
if missing:
|
| 581 |
+
print(f" Warning: missing compressor keys: {missing}")
|
| 582 |
+
if unexpected:
|
| 583 |
+
print(f" Warning: unexpected compressor keys: {unexpected}")
|
| 584 |
|
| 585 |
# Load LoRA if present
|
| 586 |
if "lora" in ckpt:
|
|
|
|
| 593 |
self._install_hook()
|
| 594 |
|
| 595 |
def _install_hook(self):
|
| 596 |
+
"""Patch Qwen3VLModel.forward to apply OpticalCompressor compression."""
|
| 597 |
+
self_ = self
|
| 598 |
+
|
| 599 |
+
def hooked_forward(self, input_ids=None, attention_mask=None, position_ids=None,
|
| 600 |
+
past_key_values=None, inputs_embeds=None,
|
| 601 |
+
pixel_values=None, pixel_values_videos=None,
|
| 602 |
+
image_grid_thw=None, video_grid_thw=None,
|
| 603 |
+
mm_token_type_ids=None, cache_position=None, **kwargs):
|
| 604 |
+
if pixel_values is not None and image_grid_thw is not None:
|
| 605 |
+
vo = self.get_image_features(pixel_values, image_grid_thw, return_dict=True)
|
| 606 |
+
pooler = vo.pooler_output
|
| 607 |
+
flat = torch.cat(pooler, dim=0) if isinstance(pooler, (list, tuple)) else pooler
|
| 608 |
+
|
| 609 |
+
sms = self_._get_spatial_merge_size()
|
| 610 |
+
gl = image_grid_thw.clone()
|
| 611 |
+
gl[:, 1] = gl[:, 1] // sms
|
| 612 |
+
gl[:, 2] = gl[:, 2] // sms
|
| 613 |
+
|
| 614 |
+
num_images = image_grid_thw.shape[0]
|
| 615 |
+
parts, new_grids_llm, offset = [], [], 0
|
| 616 |
+
for i in range(num_images):
|
| 617 |
+
t, h, w = gl[i].tolist()
|
| 618 |
+
n = int(t) * int(h) * int(w)
|
| 619 |
+
tok = flat[offset:offset + n]
|
| 620 |
+
offset += n
|
| 621 |
+
comp, new_grid_img = self_.compressor(tok.unsqueeze(0), gl[i:i+1])
|
| 622 |
+
parts.append(comp.squeeze(0))
|
| 623 |
+
new_grids_llm.append(new_grid_img.squeeze(0))
|
| 624 |
+
|
| 625 |
+
comp = torch.cat(parts, dim=0)
|
| 626 |
+
new_grid_llm = torch.stack(new_grids_llm, dim=0)
|
| 627 |
+
new_grid = new_grid_llm.clone()
|
| 628 |
+
new_grid[:, 1] = new_grid[:, 1] * sms
|
| 629 |
+
new_grid[:, 2] = new_grid[:, 2] * sms
|
| 630 |
+
self_._new_grid = new_grid
|
| 631 |
+
|
| 632 |
+
if inputs_embeds is None:
|
| 633 |
+
inputs_embeds = self.get_input_embeddings()(input_ids)
|
| 634 |
+
|
| 635 |
+
B, S, D = inputs_embeds.shape
|
| 636 |
+
flat_embeds = inputs_embeds.reshape(B * S, D).contiguous()
|
| 637 |
+
image_mask_2d = (input_ids == self.config.image_token_id) # [B, S]
|
| 638 |
+
image_mask_flat = image_mask_2d.reshape(B * S) # [B*S]
|
| 639 |
+
flat_indices = image_mask_flat.nonzero(as_tuple=True)[0] # [N_orig]
|
| 640 |
+
n_comp = comp.shape[0]
|
| 641 |
+
flat_indices = flat_indices[:n_comp]
|
| 642 |
+
|
| 643 |
+
flat_embeds[flat_indices] = comp.to(inputs_embeds.dtype)
|
| 644 |
+
inputs_embeds = flat_embeds.view(B, S, D)
|
| 645 |
+
|
| 646 |
+
# Mark only injected compressed positions as image tokens for RoPE.
|
| 647 |
+
rope_mm_token_type_ids = torch.zeros_like(input_ids, dtype=torch.int)
|
| 648 |
+
batch_idx = flat_indices // S
|
| 649 |
+
seq_idx = flat_indices % S
|
| 650 |
+
rope_mm_token_type_ids[batch_idx, seq_idx] = 1
|
| 651 |
+
|
| 652 |
+
position_ids = self.compute_3d_position_ids(
|
| 653 |
+
input_ids=input_ids, image_grid_thw=new_grid,
|
| 654 |
+
video_grid_thw=video_grid_thw, inputs_embeds=inputs_embeds,
|
| 655 |
+
attention_mask=attention_mask, past_key_values=past_key_values,
|
| 656 |
+
mm_token_type_ids=rope_mm_token_type_ids,
|
| 657 |
+
)
|
| 658 |
|
| 659 |
+
outputs = self.language_model(
|
| 660 |
+
input_ids=None, position_ids=position_ids,
|
| 661 |
+
attention_mask=attention_mask, past_key_values=past_key_values,
|
| 662 |
+
inputs_embeds=inputs_embeds, cache_position=cache_position,
|
| 663 |
+
visual_pos_masks=None, deepstack_visual_embeds=None, **kwargs,
|
| 664 |
+
)
|
| 665 |
+
from transformers.models.qwen3_vl.modeling_qwen3_vl import Qwen3VLModelOutputWithPast
|
| 666 |
+
return Qwen3VLModelOutputWithPast(
|
| 667 |
+
**outputs, rope_deltas=getattr(self, 'rope_deltas', None)
|
| 668 |
+
)
|
| 669 |
+
else:
|
| 670 |
+
return self_._orig_fwd(
|
| 671 |
+
input_ids=input_ids, attention_mask=attention_mask,
|
| 672 |
+
position_ids=position_ids, past_key_values=past_key_values,
|
| 673 |
+
inputs_embeds=inputs_embeds, pixel_values=pixel_values,
|
| 674 |
+
pixel_values_videos=pixel_values_videos,
|
| 675 |
+
image_grid_thw=image_grid_thw, video_grid_thw=video_grid_thw,
|
| 676 |
+
mm_token_type_ids=mm_token_type_ids, cache_position=cache_position, **kwargs
|
| 677 |
+
)
|
| 678 |
+
|
| 679 |
+
import types
|
| 680 |
+
self._orig_fwd = self.model.model.forward
|
| 681 |
+
self.model.model.forward = types.MethodType(hooked_forward, self.model.model)
|
| 682 |
|
| 683 |
+
def _get_spatial_merge_size(self):
|
| 684 |
+
return self.model.model.visual.spatial_merge_size
|
| 685 |
|
| 686 |
def generate(self, image):
|
| 687 |
messages = [{"role": "user", "content": [
|
|
|
|
| 697 |
t0 = time.time()
|
| 698 |
|
| 699 |
with torch.no_grad():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 700 |
out = self.model.generate(
|
| 701 |
**inputs, max_new_tokens=4096,
|
| 702 |
temperature=0.1, do_sample=True, top_p=0.9,
|
scripts/step_case_study.py
ADDED
|
@@ -0,0 +1,230 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Case Study: Generate side-by-side visual comparisons.
|
| 3 |
+
Selects representative examples and creates HTML comparison pages.
|
| 4 |
+
|
| 5 |
+
Usage:
|
| 6 |
+
python scripts/step_case_study.py
|
| 7 |
+
"""
|
| 8 |
+
|
| 9 |
+
import json
|
| 10 |
+
import os
|
| 11 |
+
import sys
|
| 12 |
+
from pathlib import Path
|
| 13 |
+
|
| 14 |
+
import numpy as np
|
| 15 |
+
from PIL import Image
|
| 16 |
+
|
| 17 |
+
PROJECT_ROOT = Path(__file__).parent.parent
|
| 18 |
+
sys.path.insert(0, str(PROJECT_ROOT))
|
| 19 |
+
|
| 20 |
+
METHODS_TO_COMPARE = [
|
| 21 |
+
"deepseek_tiny",
|
| 22 |
+
"deepseek_base",
|
| 23 |
+
"deepseek_large",
|
| 24 |
+
"qwen3_256",
|
| 25 |
+
"qwen3_1k",
|
| 26 |
+
"qwen3_full",
|
| 27 |
+
]
|
| 28 |
+
|
| 29 |
+
METHOD_LABELS = {
|
| 30 |
+
"deepseek_tiny": "DeepSeek-OCR tiny (73 tok)",
|
| 31 |
+
"deepseek_base": "DeepSeek-OCR base (273 tok)",
|
| 32 |
+
"deepseek_large": "DeepSeek-OCR large (421 tok)",
|
| 33 |
+
"qwen3_256": "Qwen3-VL 256 (722 tok)",
|
| 34 |
+
"qwen3_1k": "Qwen3-VL 1k (3043 tok)",
|
| 35 |
+
"qwen3_full": "Qwen3-VL full (6746 tok)",
|
| 36 |
+
}
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
def select_representative_samples(benchmark_dir, n=8):
|
| 40 |
+
"""Select diverse, representative samples based on CLIP score variance."""
|
| 41 |
+
clips_by_sample = {}
|
| 42 |
+
|
| 43 |
+
for method in METHODS_TO_COMPARE:
|
| 44 |
+
clip_file = Path(benchmark_dir) / method / "clip_scores.json"
|
| 45 |
+
if not clip_file.exists():
|
| 46 |
+
continue
|
| 47 |
+
with open(clip_file) as f:
|
| 48 |
+
data = json.load(f)
|
| 49 |
+
per_sample = data.get("per_sample", {})
|
| 50 |
+
for sid, val in per_sample.items():
|
| 51 |
+
score = val.get("clip_score", val) if isinstance(val, dict) else float(val)
|
| 52 |
+
clips_by_sample.setdefault(sid, {})[method] = score
|
| 53 |
+
|
| 54 |
+
candidates = []
|
| 55 |
+
for sid, scores in clips_by_sample.items():
|
| 56 |
+
if len(scores) < 4:
|
| 57 |
+
continue
|
| 58 |
+
vals = list(scores.values())
|
| 59 |
+
candidates.append({
|
| 60 |
+
"id": sid,
|
| 61 |
+
"mean_clip": np.mean(vals),
|
| 62 |
+
"std_clip": np.std(vals),
|
| 63 |
+
"max_clip": max(vals),
|
| 64 |
+
"min_clip": min(vals),
|
| 65 |
+
"range": max(vals) - min(vals),
|
| 66 |
+
"scores": scores,
|
| 67 |
+
})
|
| 68 |
+
|
| 69 |
+
candidates.sort(key=lambda c: -c["range"])
|
| 70 |
+
|
| 71 |
+
selected = []
|
| 72 |
+
high_quality = [c for c in candidates if c["mean_clip"] > 0.85]
|
| 73 |
+
if high_quality:
|
| 74 |
+
selected.append(high_quality[0])
|
| 75 |
+
|
| 76 |
+
low_quality = [c for c in candidates if c["mean_clip"] < 0.65 and c not in selected]
|
| 77 |
+
if low_quality:
|
| 78 |
+
selected.append(low_quality[0])
|
| 79 |
+
|
| 80 |
+
high_variance = [c for c in candidates if c not in selected]
|
| 81 |
+
high_variance.sort(key=lambda c: -c["range"])
|
| 82 |
+
for c in high_variance[:3]:
|
| 83 |
+
if c not in selected:
|
| 84 |
+
selected.append(c)
|
| 85 |
+
|
| 86 |
+
mid_range = [c for c in candidates if 0.70 < c["mean_clip"] < 0.80 and c not in selected]
|
| 87 |
+
mid_range.sort(key=lambda c: -c["range"])
|
| 88 |
+
for c in mid_range[:3]:
|
| 89 |
+
if c not in selected:
|
| 90 |
+
selected.append(c)
|
| 91 |
+
|
| 92 |
+
return selected[:n]
|
| 93 |
+
|
| 94 |
+
|
| 95 |
+
def render_html_to_png(html_path, output_path, width=1280, height=1024):
|
| 96 |
+
"""Render HTML to PNG screenshot."""
|
| 97 |
+
try:
|
| 98 |
+
from playwright.sync_api import sync_playwright
|
| 99 |
+
with sync_playwright() as p:
|
| 100 |
+
browser = p.chromium.launch(headless=True, args=['--no-sandbox', '--disable-gpu'])
|
| 101 |
+
page = browser.new_page(viewport={"width": width, "height": height})
|
| 102 |
+
page.goto(f"file://{html_path}", wait_until="networkidle", timeout=15000)
|
| 103 |
+
page.wait_for_timeout(500)
|
| 104 |
+
page.screenshot(path=str(output_path), full_page=False)
|
| 105 |
+
browser.close()
|
| 106 |
+
return True
|
| 107 |
+
except Exception as e:
|
| 108 |
+
print(f" Render failed: {e}")
|
| 109 |
+
return False
|
| 110 |
+
|
| 111 |
+
|
| 112 |
+
def generate_case_study_html(selected, benchmark_dir, ref_dir, output_dir):
|
| 113 |
+
"""Generate an HTML page with side-by-side comparisons."""
|
| 114 |
+
output_dir = Path(output_dir)
|
| 115 |
+
output_dir.mkdir(parents=True, exist_ok=True)
|
| 116 |
+
images_dir = output_dir / "images"
|
| 117 |
+
images_dir.mkdir(exist_ok=True)
|
| 118 |
+
|
| 119 |
+
for sample in selected:
|
| 120 |
+
sid = sample["id"]
|
| 121 |
+
ref_src = Path(ref_dir) / f"{sid}.png"
|
| 122 |
+
if ref_src.exists():
|
| 123 |
+
ref_dst = images_dir / f"ref_{sid}.png"
|
| 124 |
+
if not ref_dst.exists():
|
| 125 |
+
img = Image.open(ref_src)
|
| 126 |
+
img.thumbnail((640, 800))
|
| 127 |
+
img.save(str(ref_dst))
|
| 128 |
+
|
| 129 |
+
for method in METHODS_TO_COMPARE:
|
| 130 |
+
html_path = Path(benchmark_dir) / method / "html_predictions" / f"{sid}.html"
|
| 131 |
+
render_path = images_dir / f"{method}_{sid}.png"
|
| 132 |
+
if html_path.exists() and not render_path.exists():
|
| 133 |
+
print(f" Rendering {method}/{sid}...")
|
| 134 |
+
ok = render_html_to_png(str(html_path.resolve()), str(render_path))
|
| 135 |
+
if ok:
|
| 136 |
+
img = Image.open(render_path)
|
| 137 |
+
img.thumbnail((640, 800))
|
| 138 |
+
img.save(str(render_path))
|
| 139 |
+
|
| 140 |
+
rows_html = []
|
| 141 |
+
for i, sample in enumerate(selected):
|
| 142 |
+
sid = sample["id"]
|
| 143 |
+
scores_str = " | ".join(
|
| 144 |
+
f"{METHOD_LABELS.get(m, m).split('(')[0].strip()}: {sample['scores'].get(m, 'N/A'):.3f}"
|
| 145 |
+
if isinstance(sample['scores'].get(m), float) else f"{m}: N/A"
|
| 146 |
+
for m in METHODS_TO_COMPARE
|
| 147 |
+
)
|
| 148 |
+
|
| 149 |
+
cells = [f'<td><img src="images/ref_{sid}.png" alt="ref"><br><b>Original</b></td>']
|
| 150 |
+
for method in METHODS_TO_COMPARE:
|
| 151 |
+
label = METHOD_LABELS.get(method, method)
|
| 152 |
+
clip = sample["scores"].get(method)
|
| 153 |
+
clip_str = f"CLIP: {clip:.3f}" if clip else "N/A"
|
| 154 |
+
img_file = f"images/{method}_{sid}.png"
|
| 155 |
+
cells.append(f'<td><img src="{img_file}" alt="{method}"><br><b>{label}</b><br>{clip_str}</td>')
|
| 156 |
+
|
| 157 |
+
row = f"""
|
| 158 |
+
<tr class="case-header">
|
| 159 |
+
<td colspan="{len(METHODS_TO_COMPARE) + 1}">
|
| 160 |
+
<b>Case {i+1}</b> (Sample ID: {sid}) — Mean CLIP: {sample['mean_clip']:.3f}, Range: {sample['range']:.3f}
|
| 161 |
+
</td>
|
| 162 |
+
</tr>
|
| 163 |
+
<tr class="case-images">
|
| 164 |
+
{''.join(cells)}
|
| 165 |
+
</tr>
|
| 166 |
+
"""
|
| 167 |
+
rows_html.append(row)
|
| 168 |
+
|
| 169 |
+
html = f"""<!DOCTYPE html>
|
| 170 |
+
<html>
|
| 171 |
+
<head>
|
| 172 |
+
<title>UIPress Case Study</title>
|
| 173 |
+
<style>
|
| 174 |
+
body {{ font-family: 'Segoe UI', Arial, sans-serif; margin: 20px; background: #f5f5f5; }}
|
| 175 |
+
h1 {{ color: #333; }}
|
| 176 |
+
table {{ border-collapse: collapse; width: 100%; background: white; box-shadow: 0 2px 4px rgba(0,0,0,0.1); }}
|
| 177 |
+
.case-header td {{ background: #2c3e50; color: white; padding: 10px 15px; font-size: 14px; }}
|
| 178 |
+
.case-images td {{ padding: 8px; text-align: center; vertical-align: top; border: 1px solid #ddd; font-size: 12px; }}
|
| 179 |
+
.case-images img {{ max-width: 200px; max-height: 300px; border: 1px solid #ccc; display: block; margin: 0 auto 5px; }}
|
| 180 |
+
b {{ display: block; margin-top: 3px; }}
|
| 181 |
+
</style>
|
| 182 |
+
</head>
|
| 183 |
+
<body>
|
| 184 |
+
<h1>UIPress: Visual Token Compression Case Study</h1>
|
| 185 |
+
<p>Side-by-side comparison of {len(selected)} representative examples across {len(METHODS_TO_COMPARE)} methods.</p>
|
| 186 |
+
<table>
|
| 187 |
+
{''.join(rows_html)}
|
| 188 |
+
</table>
|
| 189 |
+
</body>
|
| 190 |
+
</html>"""
|
| 191 |
+
|
| 192 |
+
output_file = output_dir / "case_study.html"
|
| 193 |
+
output_file.write_text(html)
|
| 194 |
+
print(f"Case study saved to {output_file}")
|
| 195 |
+
|
| 196 |
+
summary = {
|
| 197 |
+
"n_cases": len(selected),
|
| 198 |
+
"methods": METHODS_TO_COMPARE,
|
| 199 |
+
"cases": [{
|
| 200 |
+
"id": s["id"],
|
| 201 |
+
"mean_clip": round(s["mean_clip"], 4),
|
| 202 |
+
"clip_range": round(s["range"], 4),
|
| 203 |
+
"scores": {k: round(v, 4) for k, v in s["scores"].items()},
|
| 204 |
+
} for s in selected],
|
| 205 |
+
}
|
| 206 |
+
with open(output_dir / "case_study_summary.json", "w") as f:
|
| 207 |
+
json.dump(summary, f, indent=2)
|
| 208 |
+
|
| 209 |
+
return output_file
|
| 210 |
+
|
| 211 |
+
|
| 212 |
+
def main():
|
| 213 |
+
benchmark_dir = PROJECT_ROOT / "results" / "benchmark"
|
| 214 |
+
ref_dir = PROJECT_ROOT / "data" / "ref_screenshots"
|
| 215 |
+
output_dir = PROJECT_ROOT / "results" / "case_study"
|
| 216 |
+
|
| 217 |
+
print("Selecting representative samples...")
|
| 218 |
+
selected = select_representative_samples(str(benchmark_dir), n=8)
|
| 219 |
+
|
| 220 |
+
print(f"\nSelected {len(selected)} cases:")
|
| 221 |
+
for s in selected:
|
| 222 |
+
print(f" ID={s['id']}: mean_clip={s['mean_clip']:.3f}, range={s['range']:.3f}")
|
| 223 |
+
|
| 224 |
+
print("\nGenerating case study...")
|
| 225 |
+
output_file = generate_case_study_html(selected, str(benchmark_dir), str(ref_dir), str(output_dir))
|
| 226 |
+
print(f"\nDone! Open {output_file} in a browser to view.")
|
| 227 |
+
|
| 228 |
+
|
| 229 |
+
if __name__ == "__main__":
|
| 230 |
+
main()
|
scripts/step_clip_batch.py
ADDED
|
@@ -0,0 +1,170 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Batch CLIP scoring for all benchmark results.
|
| 3 |
+
Computes CLIP similarity between generated HTML screenshots and reference images.
|
| 4 |
+
|
| 5 |
+
Usage:
|
| 6 |
+
conda activate uipress-qwen
|
| 7 |
+
CUDA_VISIBLE_DEVICES=X PYTHONPATH=. python scripts/step_clip_batch.py
|
| 8 |
+
"""
|
| 9 |
+
|
| 10 |
+
import os
|
| 11 |
+
os.environ["HF_ENDPOINT"] = os.environ.get("HF_ENDPOINT", "https://hf-mirror.com")
|
| 12 |
+
os.environ["HF_HOME"] = os.environ.get("HF_HOME", "/root/rivermind-data/huggingface")
|
| 13 |
+
|
| 14 |
+
import json
|
| 15 |
+
import sys
|
| 16 |
+
import tempfile
|
| 17 |
+
from pathlib import Path
|
| 18 |
+
|
| 19 |
+
import torch
|
| 20 |
+
from PIL import Image
|
| 21 |
+
from tqdm import tqdm
|
| 22 |
+
|
| 23 |
+
PROJECT_ROOT = Path(__file__).parent.parent
|
| 24 |
+
|
| 25 |
+
|
| 26 |
+
class CLIPScorer:
|
| 27 |
+
def __init__(self, device="cuda"):
|
| 28 |
+
import open_clip
|
| 29 |
+
self.device = device
|
| 30 |
+
self.model, _, self.preprocess = open_clip.create_model_and_transforms(
|
| 31 |
+
"ViT-B-32", pretrained="openai"
|
| 32 |
+
)
|
| 33 |
+
self.model = self.model.to(device).eval()
|
| 34 |
+
|
| 35 |
+
@torch.no_grad()
|
| 36 |
+
def score(self, img1, img2):
|
| 37 |
+
t1 = self.preprocess(img1).unsqueeze(0).to(self.device)
|
| 38 |
+
t2 = self.preprocess(img2).unsqueeze(0).to(self.device)
|
| 39 |
+
f1 = self.model.encode_image(t1)
|
| 40 |
+
f2 = self.model.encode_image(t2)
|
| 41 |
+
f1 = f1 / f1.norm(dim=-1, keepdim=True)
|
| 42 |
+
f2 = f2 / f2.norm(dim=-1, keepdim=True)
|
| 43 |
+
return float((f1 * f2).sum())
|
| 44 |
+
|
| 45 |
+
|
| 46 |
+
def render_html(html_path, output_path, width=1280, height=1024):
|
| 47 |
+
try:
|
| 48 |
+
from playwright.sync_api import sync_playwright
|
| 49 |
+
abs_path = os.path.abspath(html_path)
|
| 50 |
+
with sync_playwright() as p:
|
| 51 |
+
browser = p.chromium.launch(headless=True)
|
| 52 |
+
page = browser.new_page(viewport={"width": width, "height": height})
|
| 53 |
+
page.goto(f"file://{abs_path}", wait_until="networkidle", timeout=30000)
|
| 54 |
+
page.screenshot(path=output_path, full_page=False)
|
| 55 |
+
browser.close()
|
| 56 |
+
return True
|
| 57 |
+
except Exception as e:
|
| 58 |
+
try:
|
| 59 |
+
from selenium import webdriver
|
| 60 |
+
from selenium.webdriver.chrome.options import Options
|
| 61 |
+
opts = Options()
|
| 62 |
+
opts.add_argument("--headless")
|
| 63 |
+
opts.add_argument("--no-sandbox")
|
| 64 |
+
opts.add_argument(f"--window-size={width},{height}")
|
| 65 |
+
driver = webdriver.Chrome(options=opts)
|
| 66 |
+
driver.get(f"file://{os.path.abspath(html_path)}")
|
| 67 |
+
import time; time.sleep(2)
|
| 68 |
+
driver.save_screenshot(output_path)
|
| 69 |
+
driver.quit()
|
| 70 |
+
return True
|
| 71 |
+
except:
|
| 72 |
+
return False
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
def eval_method(method_dir, ref_dir, scorer, tmp_dir):
|
| 76 |
+
html_dir = Path(method_dir) / "html_predictions"
|
| 77 |
+
if not html_dir.exists():
|
| 78 |
+
return None
|
| 79 |
+
|
| 80 |
+
html_files = sorted(html_dir.glob("*.html"))
|
| 81 |
+
if not html_files:
|
| 82 |
+
return None
|
| 83 |
+
|
| 84 |
+
scores = {}
|
| 85 |
+
for hf in tqdm(html_files, desc=f"CLIP {html_dir.parent.name}"):
|
| 86 |
+
sid = hf.stem
|
| 87 |
+
ref_img_path = Path(ref_dir) / f"{sid}.png"
|
| 88 |
+
if not ref_img_path.exists():
|
| 89 |
+
continue
|
| 90 |
+
|
| 91 |
+
ref_img = Image.open(ref_img_path).convert("RGB")
|
| 92 |
+
pred_img_path = os.path.join(tmp_dir, f"{sid}.png")
|
| 93 |
+
ok = render_html(str(hf), pred_img_path)
|
| 94 |
+
|
| 95 |
+
if ok and os.path.exists(pred_img_path):
|
| 96 |
+
pred_img = Image.open(pred_img_path).convert("RGB")
|
| 97 |
+
clip = scorer.score(ref_img, pred_img)
|
| 98 |
+
else:
|
| 99 |
+
clip = 0.0
|
| 100 |
+
scores[sid] = clip
|
| 101 |
+
|
| 102 |
+
if not scores:
|
| 103 |
+
return None
|
| 104 |
+
|
| 105 |
+
vals = list(scores.values())
|
| 106 |
+
return {
|
| 107 |
+
"n": len(vals),
|
| 108 |
+
"avg_clip": round(sum(vals) / len(vals), 4),
|
| 109 |
+
"min_clip": round(min(vals), 4),
|
| 110 |
+
"max_clip": round(max(vals), 4),
|
| 111 |
+
"per_sample": {k: round(v, 4) for k, v in scores.items()},
|
| 112 |
+
}
|
| 113 |
+
|
| 114 |
+
|
| 115 |
+
def main():
|
| 116 |
+
import argparse
|
| 117 |
+
parser = argparse.ArgumentParser()
|
| 118 |
+
parser.add_argument("--benchmark_dir", default=str(PROJECT_ROOT / "results" / "benchmark"))
|
| 119 |
+
parser.add_argument("--ref_dir", default=str(PROJECT_ROOT / "data" / "ref_screenshots"))
|
| 120 |
+
parser.add_argument("--methods", nargs="*", default=None)
|
| 121 |
+
args = parser.parse_args()
|
| 122 |
+
|
| 123 |
+
bench_dir = Path(args.benchmark_dir)
|
| 124 |
+
ref_dir = Path(args.ref_dir)
|
| 125 |
+
|
| 126 |
+
if not ref_dir.exists():
|
| 127 |
+
print(f"Reference dir not found: {ref_dir}")
|
| 128 |
+
sys.exit(1)
|
| 129 |
+
|
| 130 |
+
scorer = CLIPScorer()
|
| 131 |
+
all_clip = {}
|
| 132 |
+
|
| 133 |
+
methods = args.methods or sorted(
|
| 134 |
+
d.name for d in bench_dir.iterdir()
|
| 135 |
+
if d.is_dir() and (d / "html_predictions").exists()
|
| 136 |
+
)
|
| 137 |
+
|
| 138 |
+
with tempfile.TemporaryDirectory() as tmp:
|
| 139 |
+
for method in methods:
|
| 140 |
+
method_dir = bench_dir / method
|
| 141 |
+
if not method_dir.exists():
|
| 142 |
+
continue
|
| 143 |
+
print(f"\n=== {method} ===")
|
| 144 |
+
result = eval_method(method_dir, ref_dir, scorer, tmp)
|
| 145 |
+
if result:
|
| 146 |
+
all_clip[method] = result
|
| 147 |
+
print(f" CLIP: {result['avg_clip']:.4f} (n={result['n']})")
|
| 148 |
+
|
| 149 |
+
clip_file = method_dir / "clip_scores.json"
|
| 150 |
+
with open(clip_file, "w") as f:
|
| 151 |
+
json.dump(result, f, indent=2)
|
| 152 |
+
|
| 153 |
+
agg_file = bench_dir / "all_clip_scores.json"
|
| 154 |
+
summary = {k: {kk: vv for kk, vv in v.items() if kk != "per_sample"}
|
| 155 |
+
for k, v in all_clip.items()}
|
| 156 |
+
with open(agg_file, "w") as f:
|
| 157 |
+
json.dump(summary, f, indent=2)
|
| 158 |
+
|
| 159 |
+
print(f"\n{'='*60}")
|
| 160 |
+
print(f"{'Method':<20} {'CLIP':>8} {'N':>5}")
|
| 161 |
+
print("-" * 40)
|
| 162 |
+
for k in sorted(summary, key=lambda x: summary[x]["avg_clip"], reverse=True):
|
| 163 |
+
v = summary[k]
|
| 164 |
+
print(f"{k:<20} {v['avg_clip']:>8.4f} {v['n']:>5}")
|
| 165 |
+
print(f"{'='*60}")
|
| 166 |
+
print(f"Saved to: {agg_file}")
|
| 167 |
+
|
| 168 |
+
|
| 169 |
+
if __name__ == "__main__":
|
| 170 |
+
main()
|
scripts/step_element_analysis.py
ADDED
|
@@ -0,0 +1,256 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Per-Element HTML Structure Analysis (Cross-Method Comparison)
|
| 3 |
+
==============================================================
|
| 4 |
+
Analyzes generated HTML across all methods without requiring ground truth.
|
| 5 |
+
Uses the best-performing method (qwen3_1k) as reference baseline.
|
| 6 |
+
|
| 7 |
+
Metrics per method:
|
| 8 |
+
- DOM depth, node count, element type distribution
|
| 9 |
+
- CSS property count & diversity
|
| 10 |
+
- Text content F1 vs reference method
|
| 11 |
+
- Element recall/precision vs reference method
|
| 12 |
+
- Output token efficiency (quality per token)
|
| 13 |
+
|
| 14 |
+
Usage:
|
| 15 |
+
python scripts/step_element_analysis.py
|
| 16 |
+
"""
|
| 17 |
+
|
| 18 |
+
import json
|
| 19 |
+
import re
|
| 20 |
+
import sys
|
| 21 |
+
from collections import Counter, defaultdict
|
| 22 |
+
from pathlib import Path
|
| 23 |
+
|
| 24 |
+
PROJECT_ROOT = Path(__file__).parent.parent
|
| 25 |
+
sys.path.insert(0, str(PROJECT_ROOT))
|
| 26 |
+
|
| 27 |
+
from bs4 import BeautifulSoup
|
| 28 |
+
|
| 29 |
+
|
| 30 |
+
TAG_GROUPS = {
|
| 31 |
+
"buttons": {"button"},
|
| 32 |
+
"inputs": {"input", "textarea", "select"},
|
| 33 |
+
"images": {"img", "svg", "picture"},
|
| 34 |
+
"links": {"a"},
|
| 35 |
+
"headings": {"h1", "h2", "h3", "h4", "h5", "h6"},
|
| 36 |
+
"lists": {"ul", "ol", "li"},
|
| 37 |
+
"tables": {"table", "tr", "td", "th"},
|
| 38 |
+
"forms": {"form"},
|
| 39 |
+
"nav": {"nav", "header", "footer", "aside"},
|
| 40 |
+
"containers": {"div", "section", "article", "main"},
|
| 41 |
+
"text_inline": {"p", "span", "label", "strong", "em", "b", "i"},
|
| 42 |
+
}
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
def extract_elements(html_str):
|
| 46 |
+
try:
|
| 47 |
+
soup = BeautifulSoup(html_str, "html.parser")
|
| 48 |
+
except Exception:
|
| 49 |
+
return {}
|
| 50 |
+
counts = {}
|
| 51 |
+
for category, tags in TAG_GROUPS.items():
|
| 52 |
+
counts[category] = sum(len(soup.find_all(tag)) for tag in tags)
|
| 53 |
+
counts["total_tags"] = len(soup.find_all(True))
|
| 54 |
+
return counts
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
def extract_css_props(html_str):
|
| 58 |
+
props = Counter()
|
| 59 |
+
for match in re.finditer(r'style\s*=\s*"([^"]*)"', html_str, re.IGNORECASE):
|
| 60 |
+
for prop in match.group(1).split(";"):
|
| 61 |
+
if ":" in prop:
|
| 62 |
+
name = prop.split(":")[0].strip().lower()
|
| 63 |
+
if name:
|
| 64 |
+
props[name] += 1
|
| 65 |
+
for match in re.finditer(r'<style[^>]*>(.*?)</style>', html_str, re.DOTALL | re.IGNORECASE):
|
| 66 |
+
for prop in re.findall(r'([\w-]+)\s*:', match.group(1)):
|
| 67 |
+
props[prop.lower()] += 1
|
| 68 |
+
return dict(props)
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
def extract_text(html_str):
|
| 72 |
+
try:
|
| 73 |
+
soup = BeautifulSoup(html_str, "html.parser")
|
| 74 |
+
for tag in soup(["script", "style", "meta", "link"]):
|
| 75 |
+
tag.decompose()
|
| 76 |
+
return soup.get_text(separator=" ", strip=True)
|
| 77 |
+
except Exception:
|
| 78 |
+
return ""
|
| 79 |
+
|
| 80 |
+
|
| 81 |
+
def dom_metrics(html_str):
|
| 82 |
+
try:
|
| 83 |
+
soup = BeautifulSoup(html_str, "html.parser")
|
| 84 |
+
except Exception:
|
| 85 |
+
return {"max_depth": 0, "total_nodes": 0}
|
| 86 |
+
|
| 87 |
+
max_depth = 0
|
| 88 |
+
stack = [(soup, 0)]
|
| 89 |
+
while stack:
|
| 90 |
+
el, d = stack.pop()
|
| 91 |
+
if d > max_depth:
|
| 92 |
+
max_depth = d
|
| 93 |
+
if d > 200:
|
| 94 |
+
continue
|
| 95 |
+
for c in el.children:
|
| 96 |
+
if hasattr(c, 'name') and c.name:
|
| 97 |
+
stack.append((c, d + 1))
|
| 98 |
+
|
| 99 |
+
return {
|
| 100 |
+
"max_depth": max_depth,
|
| 101 |
+
"total_nodes": len(soup.find_all(True)),
|
| 102 |
+
}
|
| 103 |
+
|
| 104 |
+
|
| 105 |
+
def char_f1(pred, ref):
|
| 106 |
+
if not pred and not ref:
|
| 107 |
+
return 1.0
|
| 108 |
+
if not pred or not ref:
|
| 109 |
+
return 0.0
|
| 110 |
+
pc, rc = Counter(pred.lower()), Counter(ref.lower())
|
| 111 |
+
common = sum((pc & rc).values())
|
| 112 |
+
if common == 0:
|
| 113 |
+
return 0.0
|
| 114 |
+
p = common / sum(pc.values())
|
| 115 |
+
r = common / sum(rc.values())
|
| 116 |
+
return 2 * p * r / (p + r)
|
| 117 |
+
|
| 118 |
+
|
| 119 |
+
def element_f1(pred_counts, ref_counts):
|
| 120 |
+
results = {}
|
| 121 |
+
for cat in TAG_GROUPS:
|
| 122 |
+
rn = ref_counts.get(cat, 0)
|
| 123 |
+
pn = pred_counts.get(cat, 0)
|
| 124 |
+
if rn == 0 and pn == 0:
|
| 125 |
+
results[cat] = 1.0
|
| 126 |
+
elif rn == 0 or pn == 0:
|
| 127 |
+
results[cat] = 0.0
|
| 128 |
+
else:
|
| 129 |
+
matched = min(pn, rn)
|
| 130 |
+
recall = matched / rn
|
| 131 |
+
precision = matched / pn
|
| 132 |
+
results[cat] = 2 * recall * precision / (recall + precision)
|
| 133 |
+
return results
|
| 134 |
+
|
| 135 |
+
|
| 136 |
+
def analyze_all(benchmark_dir, ref_method="qwen3_1k"):
|
| 137 |
+
bench = Path(benchmark_dir)
|
| 138 |
+
methods = sorted(d.name for d in bench.iterdir()
|
| 139 |
+
if d.is_dir() and (d / "html_predictions").exists())
|
| 140 |
+
|
| 141 |
+
if ref_method not in methods:
|
| 142 |
+
print(f"Reference method {ref_method} not found, using first: {methods[0]}")
|
| 143 |
+
ref_method = methods[0]
|
| 144 |
+
|
| 145 |
+
ref_dir = bench / ref_method / "html_predictions"
|
| 146 |
+
ref_htmls = {}
|
| 147 |
+
for f in sorted(ref_dir.glob("*.html")):
|
| 148 |
+
ref_htmls[f.stem] = f.read_text(encoding="utf-8", errors="ignore")
|
| 149 |
+
|
| 150 |
+
print(f"Reference: {ref_method} ({len(ref_htmls)} samples)")
|
| 151 |
+
|
| 152 |
+
ref_elements = {sid: extract_elements(h) for sid, h in ref_htmls.items()}
|
| 153 |
+
ref_texts = {sid: extract_text(h) for sid, h in ref_htmls.items()}
|
| 154 |
+
ref_css = {sid: extract_css_props(h) for sid, h in ref_htmls.items()}
|
| 155 |
+
|
| 156 |
+
all_results = {}
|
| 157 |
+
|
| 158 |
+
for method in methods:
|
| 159 |
+
html_dir = bench / method / "html_predictions"
|
| 160 |
+
pred_htmls = {}
|
| 161 |
+
for f in sorted(html_dir.glob("*.html")):
|
| 162 |
+
if f.stem in ref_htmls:
|
| 163 |
+
pred_htmls[f.stem] = f.read_text(encoding="utf-8", errors="ignore")
|
| 164 |
+
|
| 165 |
+
if not pred_htmls:
|
| 166 |
+
continue
|
| 167 |
+
|
| 168 |
+
text_f1s = []
|
| 169 |
+
dom_depths = []
|
| 170 |
+
dom_nodes = []
|
| 171 |
+
css_counts = []
|
| 172 |
+
css_unique = []
|
| 173 |
+
elem_f1s = defaultdict(list)
|
| 174 |
+
total_element_f1s = []
|
| 175 |
+
|
| 176 |
+
for sid, pred_html in pred_htmls.items():
|
| 177 |
+
pred_elem = extract_elements(pred_html)
|
| 178 |
+
ref_elem = ref_elements.get(sid, {})
|
| 179 |
+
pred_text = extract_text(pred_html)
|
| 180 |
+
ref_text = ref_texts.get(sid, "")
|
| 181 |
+
pred_css = extract_css_props(pred_html)
|
| 182 |
+
dm = dom_metrics(pred_html)
|
| 183 |
+
|
| 184 |
+
text_f1s.append(char_f1(pred_text, ref_text))
|
| 185 |
+
dom_depths.append(dm["max_depth"])
|
| 186 |
+
dom_nodes.append(dm["total_nodes"])
|
| 187 |
+
css_counts.append(sum(pred_css.values()))
|
| 188 |
+
css_unique.append(len(pred_css))
|
| 189 |
+
|
| 190 |
+
ef1 = element_f1(pred_elem, ref_elem)
|
| 191 |
+
for cat, val in ef1.items():
|
| 192 |
+
elem_f1s[cat].append(val)
|
| 193 |
+
total_element_f1s.append(sum(ef1.values()) / len(ef1))
|
| 194 |
+
|
| 195 |
+
n = len(pred_htmls)
|
| 196 |
+
per_cat = {}
|
| 197 |
+
for cat in TAG_GROUPS:
|
| 198 |
+
vals = elem_f1s[cat]
|
| 199 |
+
per_cat[cat] = round(sum(vals) / len(vals), 4) if vals else 0
|
| 200 |
+
|
| 201 |
+
result = {
|
| 202 |
+
"n_samples": n,
|
| 203 |
+
"avg_text_f1": round(sum(text_f1s) / n, 4),
|
| 204 |
+
"avg_element_f1": round(sum(total_element_f1s) / n, 4),
|
| 205 |
+
"avg_dom_depth": round(sum(dom_depths) / n, 1),
|
| 206 |
+
"avg_dom_nodes": round(sum(dom_nodes) / n, 1),
|
| 207 |
+
"avg_css_properties": round(sum(css_counts) / n, 1),
|
| 208 |
+
"avg_css_unique_props": round(sum(css_unique) / n, 1),
|
| 209 |
+
"per_category_f1": per_cat,
|
| 210 |
+
}
|
| 211 |
+
all_results[method] = result
|
| 212 |
+
|
| 213 |
+
return all_results
|
| 214 |
+
|
| 215 |
+
|
| 216 |
+
def main():
|
| 217 |
+
import argparse
|
| 218 |
+
parser = argparse.ArgumentParser()
|
| 219 |
+
parser.add_argument("--benchmark_dir", default=str(PROJECT_ROOT / "results" / "benchmark"))
|
| 220 |
+
parser.add_argument("--ref_method", default="qwen3_1k")
|
| 221 |
+
parser.add_argument("--output", default=str(PROJECT_ROOT / "results" / "element_analysis.json"))
|
| 222 |
+
args = parser.parse_args()
|
| 223 |
+
|
| 224 |
+
results = analyze_all(args.benchmark_dir, args.ref_method)
|
| 225 |
+
|
| 226 |
+
Path(args.output).parent.mkdir(parents=True, exist_ok=True)
|
| 227 |
+
with open(args.output, "w") as f:
|
| 228 |
+
json.dump(results, f, indent=2)
|
| 229 |
+
|
| 230 |
+
print(f"\n{'='*90}")
|
| 231 |
+
print(f"{'Method':<25} {'TextF1':>8} {'ElemF1':>8} {'Depth':>6} {'Nodes':>7} {'CSS':>6} {'N':>4}")
|
| 232 |
+
print("-" * 70)
|
| 233 |
+
for k in sorted(results, key=lambda x: results[x]["avg_text_f1"], reverse=True):
|
| 234 |
+
v = results[k]
|
| 235 |
+
print(f"{k:<25} {v['avg_text_f1']:>8.4f} {v['avg_element_f1']:>8.4f} "
|
| 236 |
+
f"{v['avg_dom_depth']:>6.1f} {v['avg_dom_nodes']:>7.0f} "
|
| 237 |
+
f"{v['avg_css_properties']:>6.0f} {v['n_samples']:>4}")
|
| 238 |
+
print(f"{'='*90}")
|
| 239 |
+
|
| 240 |
+
print("\nPer-category Element F1 (vs qwen3_1k):")
|
| 241 |
+
cats = list(TAG_GROUPS.keys())
|
| 242 |
+
header = f"{'Method':<25}" + "".join(f"{c[:6]:>8}" for c in cats)
|
| 243 |
+
print(header)
|
| 244 |
+
print("-" * (25 + 8 * len(cats)))
|
| 245 |
+
for method in sorted(results):
|
| 246 |
+
row = f"{method:<25}"
|
| 247 |
+
for cat in cats:
|
| 248 |
+
val = results[method]["per_category_f1"].get(cat, 0)
|
| 249 |
+
row += f"{val:>8.3f}"
|
| 250 |
+
print(row)
|
| 251 |
+
|
| 252 |
+
print(f"\nSaved to: {args.output}")
|
| 253 |
+
|
| 254 |
+
|
| 255 |
+
if __name__ == "__main__":
|
| 256 |
+
main()
|
scripts/step_ssim_bootstrap.py
ADDED
|
@@ -0,0 +1,228 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
SSIM + Bootstrap CI computation for all benchmark methods.
|
| 3 |
+
Works with existing rendered screenshots and per-sample CLIP scores.
|
| 4 |
+
|
| 5 |
+
Usage:
|
| 6 |
+
python scripts/step_ssim_bootstrap.py --benchmark_dir results/benchmark --ref_dir data/ref_screenshots
|
| 7 |
+
python scripts/step_ssim_bootstrap.py --benchmark_dir results/benchmark_websight --ref_dir data/ref_screenshots_websight
|
| 8 |
+
"""
|
| 9 |
+
|
| 10 |
+
import argparse
|
| 11 |
+
import json
|
| 12 |
+
import os
|
| 13 |
+
import sys
|
| 14 |
+
from pathlib import Path
|
| 15 |
+
|
| 16 |
+
import numpy as np
|
| 17 |
+
from PIL import Image
|
| 18 |
+
|
| 19 |
+
PROJECT_ROOT = Path(__file__).parent.parent
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
def compute_ssim_pil(img1, img2, win_size=7):
|
| 23 |
+
"""Compute SSIM between two PIL images using numpy (no skimage dependency)."""
|
| 24 |
+
target_size = (min(img1.width, img2.width, 512), min(img1.height, img2.height, 512))
|
| 25 |
+
a = np.array(img1.resize(target_size).convert("RGB"), dtype=np.float64)
|
| 26 |
+
b = np.array(img2.resize(target_size).convert("RGB"), dtype=np.float64)
|
| 27 |
+
|
| 28 |
+
C1 = (0.01 * 255) ** 2
|
| 29 |
+
C2 = (0.03 * 255) ** 2
|
| 30 |
+
|
| 31 |
+
ssims = []
|
| 32 |
+
for ch in range(3):
|
| 33 |
+
mu1 = uniform_filter(a[:, :, ch], win_size)
|
| 34 |
+
mu2 = uniform_filter(b[:, :, ch], win_size)
|
| 35 |
+
mu1_sq = mu1 ** 2
|
| 36 |
+
mu2_sq = mu2 ** 2
|
| 37 |
+
mu1_mu2 = mu1 * mu2
|
| 38 |
+
sigma1_sq = uniform_filter(a[:, :, ch] ** 2, win_size) - mu1_sq
|
| 39 |
+
sigma2_sq = uniform_filter(b[:, :, ch] ** 2, win_size) - mu2_sq
|
| 40 |
+
sigma12 = uniform_filter(a[:, :, ch] * b[:, :, ch], win_size) - mu1_mu2
|
| 41 |
+
|
| 42 |
+
ssim_map = ((2 * mu1_mu2 + C1) * (2 * sigma12 + C2)) / \
|
| 43 |
+
((mu1_sq + mu2_sq + C1) * (sigma1_sq + sigma2_sq + C2))
|
| 44 |
+
ssims.append(ssim_map.mean())
|
| 45 |
+
|
| 46 |
+
return float(np.mean(ssims))
|
| 47 |
+
|
| 48 |
+
|
| 49 |
+
def uniform_filter(arr, size):
|
| 50 |
+
"""Simple uniform (box) filter."""
|
| 51 |
+
from scipy.ndimage import uniform_filter as _uf
|
| 52 |
+
return _uf(arr, size=size, mode='reflect')
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
def render_html_to_screenshot(html_path, out_path, width=1280, height=1024, timeout=15000):
|
| 56 |
+
"""Render HTML file to PNG screenshot using Playwright."""
|
| 57 |
+
try:
|
| 58 |
+
from playwright.sync_api import sync_playwright
|
| 59 |
+
with sync_playwright() as p:
|
| 60 |
+
browser = p.chromium.launch(headless=True, args=['--no-sandbox', '--disable-gpu'])
|
| 61 |
+
page = browser.new_page(viewport={"width": width, "height": height})
|
| 62 |
+
page.goto(f"file://{html_path}", wait_until="networkidle", timeout=timeout)
|
| 63 |
+
page.wait_for_timeout(1000)
|
| 64 |
+
page.screenshot(path=str(out_path), full_page=True)
|
| 65 |
+
browser.close()
|
| 66 |
+
return True
|
| 67 |
+
except Exception as e:
|
| 68 |
+
print(f" Render failed for {html_path}: {e}")
|
| 69 |
+
return False
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
def compute_ssim_for_method(method_dir, ref_dir, render_cache_dir):
|
| 73 |
+
"""Compute per-sample SSIM for a method."""
|
| 74 |
+
html_dir = Path(method_dir) / "html_predictions"
|
| 75 |
+
if not html_dir.exists():
|
| 76 |
+
return None
|
| 77 |
+
|
| 78 |
+
ref_dir = Path(ref_dir)
|
| 79 |
+
render_dir = Path(render_cache_dir) / Path(method_dir).name
|
| 80 |
+
render_dir.mkdir(parents=True, exist_ok=True)
|
| 81 |
+
|
| 82 |
+
html_files = sorted(html_dir.glob("*.html"))
|
| 83 |
+
per_sample = {}
|
| 84 |
+
|
| 85 |
+
for html_path in html_files:
|
| 86 |
+
sample_id = html_path.stem
|
| 87 |
+
ref_path = ref_dir / f"{sample_id}.png"
|
| 88 |
+
if not ref_path.exists():
|
| 89 |
+
continue
|
| 90 |
+
|
| 91 |
+
rendered_path = render_dir / f"{sample_id}.png"
|
| 92 |
+
if not rendered_path.exists():
|
| 93 |
+
ok = render_html_to_screenshot(str(html_path.resolve()), str(rendered_path))
|
| 94 |
+
if not ok:
|
| 95 |
+
continue
|
| 96 |
+
|
| 97 |
+
try:
|
| 98 |
+
ref_img = Image.open(ref_path).convert("RGB")
|
| 99 |
+
rendered_img = Image.open(rendered_path).convert("RGB")
|
| 100 |
+
ssim = compute_ssim_pil(ref_img, rendered_img)
|
| 101 |
+
per_sample[sample_id] = ssim
|
| 102 |
+
except Exception as e:
|
| 103 |
+
print(f" SSIM error for {sample_id}: {e}")
|
| 104 |
+
continue
|
| 105 |
+
|
| 106 |
+
if not per_sample:
|
| 107 |
+
return None
|
| 108 |
+
|
| 109 |
+
return {
|
| 110 |
+
"n_samples": len(per_sample),
|
| 111 |
+
"avg_ssim": round(float(np.mean(list(per_sample.values()))), 4),
|
| 112 |
+
"std_ssim": round(float(np.std(list(per_sample.values()))), 4),
|
| 113 |
+
"per_sample": per_sample,
|
| 114 |
+
}
|
| 115 |
+
|
| 116 |
+
|
| 117 |
+
def bootstrap_ci(scores, n_bootstrap=10000, ci=0.95, seed=42):
|
| 118 |
+
"""Compute bootstrap confidence interval."""
|
| 119 |
+
rng = np.random.RandomState(seed)
|
| 120 |
+
scores = np.array(scores)
|
| 121 |
+
n = len(scores)
|
| 122 |
+
boot_means = np.array([
|
| 123 |
+
rng.choice(scores, size=n, replace=True).mean()
|
| 124 |
+
for _ in range(n_bootstrap)
|
| 125 |
+
])
|
| 126 |
+
alpha = (1 - ci) / 2
|
| 127 |
+
lo = float(np.percentile(boot_means, 100 * alpha))
|
| 128 |
+
hi = float(np.percentile(boot_means, 100 * (1 - alpha)))
|
| 129 |
+
return {
|
| 130 |
+
"mean": float(scores.mean()),
|
| 131 |
+
"ci_lower": round(lo, 4),
|
| 132 |
+
"ci_upper": round(hi, 4),
|
| 133 |
+
"ci_width": round(hi - lo, 4),
|
| 134 |
+
"std": round(float(scores.std()), 4),
|
| 135 |
+
"n": n,
|
| 136 |
+
}
|
| 137 |
+
|
| 138 |
+
|
| 139 |
+
def compute_bootstrap_for_all(benchmark_dir):
|
| 140 |
+
"""Compute bootstrap CI for all methods from per-sample CLIP scores."""
|
| 141 |
+
benchmark_dir = Path(benchmark_dir)
|
| 142 |
+
results = {}
|
| 143 |
+
|
| 144 |
+
for method_dir in sorted(benchmark_dir.iterdir()):
|
| 145 |
+
if not method_dir.is_dir():
|
| 146 |
+
continue
|
| 147 |
+
clip_file = method_dir / "clip_scores.json"
|
| 148 |
+
if not clip_file.exists():
|
| 149 |
+
continue
|
| 150 |
+
|
| 151 |
+
with open(clip_file) as f:
|
| 152 |
+
clip_data = json.load(f)
|
| 153 |
+
|
| 154 |
+
per_sample = clip_data.get("per_sample", {})
|
| 155 |
+
if not per_sample:
|
| 156 |
+
continue
|
| 157 |
+
|
| 158 |
+
scores = []
|
| 159 |
+
for k, v in per_sample.items():
|
| 160 |
+
if isinstance(v, dict):
|
| 161 |
+
scores.append(v.get("clip_score", 0))
|
| 162 |
+
else:
|
| 163 |
+
scores.append(float(v))
|
| 164 |
+
|
| 165 |
+
if not scores:
|
| 166 |
+
continue
|
| 167 |
+
|
| 168 |
+
ci_result = bootstrap_ci(scores)
|
| 169 |
+
results[method_dir.name] = ci_result
|
| 170 |
+
print(f" {method_dir.name}: CLIP={ci_result['mean']:.4f} [{ci_result['ci_lower']:.4f}, {ci_result['ci_upper']:.4f}]")
|
| 171 |
+
|
| 172 |
+
return results
|
| 173 |
+
|
| 174 |
+
|
| 175 |
+
def main():
|
| 176 |
+
parser = argparse.ArgumentParser()
|
| 177 |
+
parser.add_argument("--benchmark_dir", type=str, default=str(PROJECT_ROOT / "results" / "benchmark"))
|
| 178 |
+
parser.add_argument("--ref_dir", type=str, default=str(PROJECT_ROOT / "data" / "ref_screenshots"))
|
| 179 |
+
parser.add_argument("--render_cache", type=str, default=str(PROJECT_ROOT / "results" / "rendered_screenshots"))
|
| 180 |
+
parser.add_argument("--skip_ssim", action="store_true")
|
| 181 |
+
parser.add_argument("--skip_bootstrap", action="store_true")
|
| 182 |
+
args = parser.parse_args()
|
| 183 |
+
|
| 184 |
+
benchmark_dir = Path(args.benchmark_dir)
|
| 185 |
+
output = {}
|
| 186 |
+
|
| 187 |
+
if not args.skip_bootstrap:
|
| 188 |
+
print("=" * 60)
|
| 189 |
+
print("Computing Bootstrap CI for CLIP scores...")
|
| 190 |
+
print("=" * 60)
|
| 191 |
+
bootstrap_results = compute_bootstrap_for_all(args.benchmark_dir)
|
| 192 |
+
output["bootstrap_ci"] = bootstrap_results
|
| 193 |
+
|
| 194 |
+
ci_file = benchmark_dir / "bootstrap_ci.json"
|
| 195 |
+
with open(ci_file, "w") as f:
|
| 196 |
+
json.dump(bootstrap_results, f, indent=2)
|
| 197 |
+
print(f"\nSaved to {ci_file}")
|
| 198 |
+
|
| 199 |
+
if not args.skip_ssim:
|
| 200 |
+
print("\n" + "=" * 60)
|
| 201 |
+
print("Computing SSIM scores...")
|
| 202 |
+
print("=" * 60)
|
| 203 |
+
ssim_results = {}
|
| 204 |
+
for method_dir in sorted(benchmark_dir.iterdir()):
|
| 205 |
+
if not method_dir.is_dir():
|
| 206 |
+
continue
|
| 207 |
+
html_dir = method_dir / "html_predictions"
|
| 208 |
+
if not html_dir.exists():
|
| 209 |
+
continue
|
| 210 |
+
print(f"\n Processing {method_dir.name}...")
|
| 211 |
+
result = compute_ssim_for_method(str(method_dir), args.ref_dir, args.render_cache)
|
| 212 |
+
if result:
|
| 213 |
+
ssim_results[method_dir.name] = {
|
| 214 |
+
"n_samples": result["n_samples"],
|
| 215 |
+
"avg_ssim": result["avg_ssim"],
|
| 216 |
+
"std_ssim": result["std_ssim"],
|
| 217 |
+
}
|
| 218 |
+
print(f" SSIM={result['avg_ssim']:.4f} ± {result['std_ssim']:.4f} (n={result['n_samples']})")
|
| 219 |
+
output[f"ssim_{method_dir.name}"] = result
|
| 220 |
+
|
| 221 |
+
ssim_file = benchmark_dir / "ssim_scores.json"
|
| 222 |
+
with open(ssim_file, "w") as f:
|
| 223 |
+
json.dump(ssim_results, f, indent=2)
|
| 224 |
+
print(f"\nSaved to {ssim_file}")
|
| 225 |
+
|
| 226 |
+
|
| 227 |
+
if __name__ == "__main__":
|
| 228 |
+
main()
|
scripts/train_compressor.py
CHANGED
|
@@ -244,6 +244,9 @@ class CompressedQwen3VL(nn.Module):
|
|
| 244 |
target_tokens=target_tokens,
|
| 245 |
).to(torch.bfloat16)
|
| 246 |
|
|
|
|
|
|
|
|
|
|
| 247 |
# Add LoRA to LLM decoder
|
| 248 |
self._add_lora(lora_r, lora_alpha)
|
| 249 |
|
|
@@ -271,6 +274,59 @@ class CompressedQwen3VL(nn.Module):
|
|
| 271 |
)
|
| 272 |
setattr(attn, proj_name, self.lora_modules[lora_key])
|
| 273 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 274 |
def prepare_inputs(self, images, htmls):
|
| 275 |
"""Prepare model inputs for a batch of image-html pairs."""
|
| 276 |
batch_messages = []
|
|
@@ -293,61 +349,35 @@ class CompressedQwen3VL(nn.Module):
|
|
| 293 |
return inputs
|
| 294 |
|
| 295 |
def forward(self, images, htmls, device):
|
| 296 |
-
"""Full training forward pass. Returns loss scalar.
|
| 297 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 298 |
inputs = self.prepare_inputs(images, htmls)
|
| 299 |
pixel_values = inputs["pixel_values"].to(device, torch.bfloat16)
|
| 300 |
image_grid_thw = inputs["image_grid_thw"].to(device)
|
| 301 |
input_ids = inputs["input_ids"].to(device)
|
| 302 |
|
| 303 |
-
#
|
| 304 |
-
|
| 305 |
-
visual_embeds = self.base_model.visual(
|
| 306 |
-
pixel_values, grid_thw=image_grid_thw,
|
| 307 |
-
) # [total_tokens, hidden_dim]
|
| 308 |
-
|
| 309 |
-
# 3. Compress
|
| 310 |
-
compressed, new_grid_thw = self.compressor(
|
| 311 |
-
visual_embeds, image_grid_thw,
|
| 312 |
-
) # [total_compressed, hidden_dim]
|
| 313 |
-
|
| 314 |
-
# 4. Build new input sequence
|
| 315 |
-
new_input_ids, labels = self._rebuild_sequence(
|
| 316 |
-
input_ids, htmls, device,
|
| 317 |
-
)
|
| 318 |
-
|
| 319 |
-
# 5. Build inputs_embeds
|
| 320 |
-
lm = self.base_model.model
|
| 321 |
-
if hasattr(lm, "language_model"):
|
| 322 |
-
embed_layer = lm.language_model.embed_tokens
|
| 323 |
-
elif hasattr(lm, "embed_tokens"):
|
| 324 |
-
embed_layer = lm.embed_tokens
|
| 325 |
-
else:
|
| 326 |
-
embed_layer = lm.get_input_embeddings()
|
| 327 |
-
|
| 328 |
-
with torch.no_grad():
|
| 329 |
-
text_embeds = embed_layer(new_input_ids) # [B, seq_len, D]
|
| 330 |
-
|
| 331 |
-
# Scatter compressed visual tokens into the sequence
|
| 332 |
-
inputs_embeds = self._scatter_visual(
|
| 333 |
-
text_embeds, new_input_ids, compressed, new_grid_thw,
|
| 334 |
-
)
|
| 335 |
-
|
| 336 |
-
# 6. Build position_ids (vectorized, no Python loops over tokens)
|
| 337 |
-
position_ids = self._build_position_ids(
|
| 338 |
-
new_input_ids, new_grid_thw, device,
|
| 339 |
-
)
|
| 340 |
|
| 341 |
-
#
|
| 342 |
-
pad_token_id = self.processor.tokenizer.pad_token_id
|
| 343 |
-
if pad_token_id is None:
|
| 344 |
-
pad_token_id = 0
|
| 345 |
attention_mask = (new_input_ids != pad_token_id).long()
|
| 346 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 347 |
outputs = self.base_model(
|
| 348 |
-
|
| 349 |
attention_mask=attention_mask,
|
| 350 |
-
|
|
|
|
|
|
|
| 351 |
labels=labels,
|
| 352 |
)
|
| 353 |
return outputs.loss
|
|
@@ -543,7 +573,7 @@ def train(args):
|
|
| 543 |
sampler = DistributedSampler(dataset) if is_distributed else None
|
| 544 |
loader = DataLoader(
|
| 545 |
dataset, batch_size=args.batch_size, sampler=sampler,
|
| 546 |
-
shuffle=(sampler is None), num_workers=
|
| 547 |
collate_fn=lambda batch: batch,
|
| 548 |
)
|
| 549 |
|
|
|
|
| 244 |
target_tokens=target_tokens,
|
| 245 |
).to(torch.bfloat16)
|
| 246 |
|
| 247 |
+
# Monkey-patch get_image_features so it auto-compresses visual tokens.
|
| 248 |
+
self._patch_vision(compressor=self.compressor, target_tokens=target_tokens)
|
| 249 |
+
|
| 250 |
# Add LoRA to LLM decoder
|
| 251 |
self._add_lora(lora_r, lora_alpha)
|
| 252 |
|
|
|
|
| 274 |
)
|
| 275 |
setattr(attn, proj_name, self.lora_modules[lora_key])
|
| 276 |
|
| 277 |
+
def _patch_vision(self, compressor, target_tokens):
|
| 278 |
+
"""Monkey-patch get_image_features to auto-compress visual tokens.
|
| 279 |
+
|
| 280 |
+
The compressed tokens PER IMAGE = target_tokens (not T*H*W/spatial_merge_size^2).
|
| 281 |
+
We intercept get_image_features so Qwen3-VL's M-RoPE / placeholder_mask /
|
| 282 |
+
masked_scatter all work unchanged — only the actual embedding values change.
|
| 283 |
+
"""
|
| 284 |
+
import functools
|
| 285 |
+
|
| 286 |
+
orig_get_img = self.base_model.model.get_image_features
|
| 287 |
+
|
| 288 |
+
@functools.wraps(orig_get_img)
|
| 289 |
+
def patched_get_image_features(pixel_values, image_grid_thw=None, **kwargs):
|
| 290 |
+
import torch
|
| 291 |
+
from transformers.models.qwen3_vl.modeling_qwen3_vl import (
|
| 292 |
+
BaseModelOutputWithDeepstackFeatures,
|
| 293 |
+
)
|
| 294 |
+
vision_output = orig_get_img(pixel_values, image_grid_thw, **kwargs)
|
| 295 |
+
pooler = vision_output.pooler_output
|
| 296 |
+
if not isinstance(pooler, (list, tuple)):
|
| 297 |
+
flat = pooler
|
| 298 |
+
else:
|
| 299 |
+
flat = torch.cat(pooler, dim=0)
|
| 300 |
+
|
| 301 |
+
# Compute LLM-side grid after spatial merge
|
| 302 |
+
vis = self.base_model.model.visual
|
| 303 |
+
sms = vis.spatial_merge_size
|
| 304 |
+
grid_llm = image_grid_thw.clone()
|
| 305 |
+
grid_llm[:, 1] = grid_llm[:, 1] // sms
|
| 306 |
+
grid_llm[:, 2] = grid_llm[:, 2] // sms
|
| 307 |
+
num_images = image_grid_thw.shape[0]
|
| 308 |
+
compressed_parts = []
|
| 309 |
+
offset = 0
|
| 310 |
+
for i in range(num_images):
|
| 311 |
+
t, h, w = grid_llm[i].tolist()
|
| 312 |
+
n_orig = t * h * w
|
| 313 |
+
orig_slice = flat[offset: offset + n_orig]
|
| 314 |
+
offset += n_orig
|
| 315 |
+
comp_part, _ = compressor(orig_slice.unsqueeze(0), grid_llm[i:i+1])
|
| 316 |
+
compressed_parts.append(comp_part.squeeze(0))
|
| 317 |
+
|
| 318 |
+
compressed_flat = torch.cat(compressed_parts, dim=0)
|
| 319 |
+
split_sizes = [target_tokens] * num_images
|
| 320 |
+
|
| 321 |
+
# Reconstruct pooler_output as list (expected by Qwen3VLModel.forward)
|
| 322 |
+
vision_output.pooler_output = torch.split(compressed_flat, split_sizes)
|
| 323 |
+
# Disable deepstack processing to avoid dimension mismatch
|
| 324 |
+
vision_output.deepstack_features = []
|
| 325 |
+
|
| 326 |
+
return vision_output
|
| 327 |
+
|
| 328 |
+
self.base_model.model.get_image_features = patched_get_image_features
|
| 329 |
+
|
| 330 |
def prepare_inputs(self, images, htmls):
|
| 331 |
"""Prepare model inputs for a batch of image-html pairs."""
|
| 332 |
batch_messages = []
|
|
|
|
| 349 |
return inputs
|
| 350 |
|
| 351 |
def forward(self, images, htmls, device):
|
| 352 |
+
"""Full training forward pass. Returns loss scalar.
|
| 353 |
+
|
| 354 |
+
The _patch_vision monkey-patch auto-compresses visual tokens inside
|
| 355 |
+
get_image_features, so Qwen3-VL's forward handles M-RoPE, placeholder_mask,
|
| 356 |
+
and masked_scatter automatically. We only supply the HTML teacher targets.
|
| 357 |
+
"""
|
| 358 |
inputs = self.prepare_inputs(images, htmls)
|
| 359 |
pixel_values = inputs["pixel_values"].to(device, torch.bfloat16)
|
| 360 |
image_grid_thw = inputs["image_grid_thw"].to(device)
|
| 361 |
input_ids = inputs["input_ids"].to(device)
|
| 362 |
|
| 363 |
+
# Build HTML target (teacher forcing)
|
| 364 |
+
new_input_ids, labels = self._rebuild_sequence(input_ids, htmls, device)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 365 |
|
| 366 |
+
# Attention mask
|
| 367 |
+
pad_token_id = self.processor.tokenizer.pad_token_id or 0
|
|
|
|
|
|
|
| 368 |
attention_mask = (new_input_ids != pad_token_id).long()
|
| 369 |
|
| 370 |
+
# Forward through Qwen3-VL.
|
| 371 |
+
# input_ids: provides #image tokens for M-RoPE / placeholder_mask.
|
| 372 |
+
# pixel_values + image_grid_thw: trigger patched get_image_features
|
| 373 |
+
# which returns compressed embeddings → auto masked-scattered.
|
| 374 |
+
# output_hidden_states=False: disables deepstack_features mismatch.
|
| 375 |
outputs = self.base_model(
|
| 376 |
+
input_ids=new_input_ids,
|
| 377 |
attention_mask=attention_mask,
|
| 378 |
+
pixel_values=pixel_values,
|
| 379 |
+
image_grid_thw=image_grid_thw,
|
| 380 |
+
output_hidden_states=False,
|
| 381 |
labels=labels,
|
| 382 |
)
|
| 383 |
return outputs.loss
|
|
|
|
| 573 |
sampler = DistributedSampler(dataset) if is_distributed else None
|
| 574 |
loader = DataLoader(
|
| 575 |
dataset, batch_size=args.batch_size, sampler=sampler,
|
| 576 |
+
shuffle=(sampler is None), num_workers=0, pin_memory=True,
|
| 577 |
collate_fn=lambda batch: batch,
|
| 578 |
)
|
| 579 |
|
sync_up.py
CHANGED
|
@@ -1,17 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
| 1 |
from huggingface_hub import HfApi
|
| 2 |
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
)
|
| 16 |
-
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import argparse
|
| 2 |
+
import os
|
| 3 |
+
|
| 4 |
from huggingface_hub import HfApi
|
| 5 |
|
| 6 |
+
|
| 7 |
+
def parse_args() -> argparse.Namespace:
|
| 8 |
+
parser = argparse.ArgumentParser(description="Sync local folder to Hugging Face repository.")
|
| 9 |
+
parser.add_argument("--repo_id", default="DesonDai/UIPress")
|
| 10 |
+
parser.add_argument("--repo_type", default="dataset", choices=["dataset", "model", "space"])
|
| 11 |
+
parser.add_argument("--folder_path", default=".")
|
| 12 |
+
parser.add_argument("--path_in_repo", default=".")
|
| 13 |
+
parser.add_argument("--commit_message", default="Sync local changes to HF")
|
| 14 |
+
parser.add_argument(
|
| 15 |
+
"--mode",
|
| 16 |
+
default="large",
|
| 17 |
+
choices=["large", "regular"],
|
| 18 |
+
help="large: upload_large_folder (recommended for big folders); regular: upload_folder",
|
| 19 |
+
)
|
| 20 |
+
parser.add_argument(
|
| 21 |
+
"--token_env",
|
| 22 |
+
default="HF_TOKEN",
|
| 23 |
+
help="Environment variable that stores Hugging Face token.",
|
| 24 |
+
)
|
| 25 |
+
parser.add_argument(
|
| 26 |
+
"--ignore",
|
| 27 |
+
nargs="*",
|
| 28 |
+
default=[
|
| 29 |
+
".git/*",
|
| 30 |
+
".cursor/*",
|
| 31 |
+
"__pycache__/*",
|
| 32 |
+
"*.pyc",
|
| 33 |
+
"OLD/*",
|
| 34 |
+
"*.tar.gz",
|
| 35 |
+
],
|
| 36 |
+
help="Glob patterns to ignore while uploading.",
|
| 37 |
+
)
|
| 38 |
+
parser.add_argument("--num_workers", type=int, default=4)
|
| 39 |
+
return parser.parse_args()
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
def main() -> None:
|
| 43 |
+
args = parse_args()
|
| 44 |
+
repo_id = args.repo_id.strip().strip("/")
|
| 45 |
+
if "/" not in repo_id:
|
| 46 |
+
raise RuntimeError(
|
| 47 |
+
f"Invalid --repo_id '{args.repo_id}'. Expected format: <namespace>/<repo_name>, "
|
| 48 |
+
"for example: DesonDai/UIPress"
|
| 49 |
+
)
|
| 50 |
+
|
| 51 |
+
token = os.environ.get(args.token_env)
|
| 52 |
+
if not token:
|
| 53 |
+
raise RuntimeError(
|
| 54 |
+
f"Missing token: please set {args.token_env} first, then rerun."
|
| 55 |
+
)
|
| 56 |
+
|
| 57 |
+
api = HfApi(token=token)
|
| 58 |
+
|
| 59 |
+
if args.mode == "large":
|
| 60 |
+
# Recommended for large folders: more robust and resumable uploads.
|
| 61 |
+
api.upload_large_folder(
|
| 62 |
+
repo_id=repo_id,
|
| 63 |
+
folder_path=args.folder_path,
|
| 64 |
+
repo_type=args.repo_type,
|
| 65 |
+
ignore_patterns=args.ignore,
|
| 66 |
+
num_workers=args.num_workers,
|
| 67 |
+
print_report=True,
|
| 68 |
+
print_report_every=30,
|
| 69 |
+
)
|
| 70 |
+
else:
|
| 71 |
+
api.upload_folder(
|
| 72 |
+
repo_id=repo_id,
|
| 73 |
+
folder_path=args.folder_path,
|
| 74 |
+
repo_type=args.repo_type,
|
| 75 |
+
path_in_repo=args.path_in_repo,
|
| 76 |
+
commit_message=args.commit_message,
|
| 77 |
+
ignore_patterns=args.ignore,
|
| 78 |
+
token=token,
|
| 79 |
+
)
|
| 80 |
+
|
| 81 |
+
print(f"Sync completed: {args.repo_type}/{repo_id}")
|
| 82 |
+
|
| 83 |
+
|
| 84 |
+
if __name__ == "__main__":
|
| 85 |
+
main()
|