Skip to content

Commit 5c12bce

Browse files
committed
docs: remove emoji symbols from documentation content
1 parent dc7613c commit 5c12bce

8 files changed

Lines changed: 63 additions & 63 deletions

File tree

doc/DATA_FORMAT_TEMPLATE.md

Lines changed: 14 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -4,18 +4,18 @@
44
55
---
66

7-
## 📋 列名规范总览
7+
## 列名规范总览
88

99
| 用途 | 列名 | 是否必需 | 说明 |
1010
|------|------|---------|------|
11-
| **文本** | `text` | 必需 | 文档文本内容 |
11+
| **文本** | `text` | 必需 | 文档文本内容 |
1212
| **时间** | `timestamp` | DTM 必需 | 支持年份/日期/时间戳格式 |
1313
| **协变量** | `cov_*` | STM 必需 | 必须以 `cov_` 前缀命名 |
1414
| **标签** | `label` | 有监督必需 | 文档分类标签 |
1515

1616
---
1717

18-
## 1️⃣ 文本列: `text`
18+
## 1. 文本列: `text`
1919

2020
**必须命名为 `text`**
2121

@@ -27,7 +27,7 @@ text
2727

2828
---
2929

30-
## 2️⃣ 时间列: `timestamp`
30+
## 2. 时间列: `timestamp`
3131

3232
**必须命名为 `timestamp`**,用于 DTM (Dynamic Topic Model) 时序分析。
3333

@@ -48,15 +48,15 @@ text,timestamp
4848
"2022年的文档内容...",2022-03-20 10:30:00
4949
```
5050

51-
### ⚠️ 重要说明
51+
### 重要说明
5252

5353
- **DTM 最终只使用年份**,不支持月/日级别的时间切片
5454
- 无论输入 `2026-10-17` 还是 `2026`,最终都会转换为年份 `2026`
5555
- 至少需要 **2 个不同的年份** 才能运行 DTM
5656

5757
---
5858

59-
## 3️⃣ 协变量列: `cov_*` 前缀
59+
## 3. 协变量列: `cov_*` 前缀
6060

6161
**必须以 `cov_` 前缀命名**,用于 STM (Structural Topic Model) 分析。
6262

@@ -87,7 +87,7 @@ text,timestamp,cov_province,cov_category,cov_source
8787

8888
---
8989

90-
## 4️⃣ 标签列: `label`
90+
## 4. 标签列: `label`
9191

9292
**必须命名为 `label`**,用于有监督主题建模。
9393

@@ -100,7 +100,7 @@ text,label
100100

101101
---
102102

103-
## 📝 完整数据模板
103+
## 完整数据模板
104104

105105
### 模板 1: 基础主题建模 (LDA/CTM/ETM)
106106

@@ -149,7 +149,7 @@ text,timestamp,cov_province,cov_category,cov_source,label
149149

150150
---
151151

152-
## 🔍 系统检测输出示例
152+
## 系统检测输出示例
153153

154154
运行 `prepare_data.py` 时会显示:
155155

@@ -164,11 +164,11 @@ Column Auto-Detection Results (Strict Mode)
164164
- Covariate columns: 'cov_<name>' (e.g., cov_province, cov_category)
165165
- Label column: 'label'
166166
167-
[Time Column] 'timestamp' (type: year)
167+
[Time Column] OK 'timestamp' (type: year)
168168
Sample values: [2020, 2021, 2022, 2023, 2024]
169169
Note: All formats will be converted to YEAR for DTM analysis
170170
171-
[Covariate Columns] Detected 3 columns:
171+
[Covariate Columns] OK Detected 3 columns:
172172
- cov_province: 34 unique values, e.g., ['北京', '上海', '广东']
173173
- cov_category: 5 unique values, e.g., ['政策', '新闻', '报告']
174174
- cov_source: 8 unique values, e.g., ['政府网站', '媒体', '学术期刊']
@@ -177,7 +177,7 @@ Column Auto-Detection Results (Strict Mode)
177177

178178
---
179179

180-
## ⚠️ 遗留列名兼容
180+
## 遗留列名兼容
181181

182182
为保持向后兼容,系统仍会识别以下遗留列名,但会显示警告:
183183

@@ -195,7 +195,7 @@ Column Auto-Detection Results (Strict Mode)
195195

196196
---
197197

198-
## 🛠️ 数据准备命令
198+
## 数据准备命令
199199

200200
```bash
201201
cd THETA/src/models
@@ -212,7 +212,7 @@ python prepare_data.py --dataset my_data --model baseline
212212

213213
---
214214

215-
## 数据质量检查清单
215+
## 数据质量检查清单
216216

217217
- [ ] CSV 文件使用 UTF-8 编码
218218
- [ ] 文本列命名为 `text`

doc/appendix/hardware-requirements.md

Lines changed: 12 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -21,9 +21,9 @@ This appendix provides detailed performance benchmarks and hardware requirements
2121
- 8B: 35 docs/s (**117× faster than CPU**)
2222

2323
3. **Maximum Configurations on 24GB GPU**
24-
- 0.6B: batch_size=64, seq_len=512
25-
- 4B: batch_size=64, seq_len=512
26-
- 8B: batch_size=64, seq_len=256 ✅ (OOM at seq_len=512)
24+
- 0.6B: batch_size=64, seq_len=512 (supported)
25+
- 4B: batch_size=64, seq_len=512 (supported)
26+
- 8B: batch_size=64, seq_len=256 (supported; OOM at seq_len=512)
2727

2828
4. **Recommended Production Configurations** (with 20% memory headroom)
2929
- 0.6B: batch_size=32, seq_len=256 (peak memory 2.2 GB)
@@ -162,19 +162,19 @@ Tight memory (<8GB) → batch_size=8 or use gradient accumulation
162162
### Recommended Configurations by GPU Memory
163163

164164
#### 8GB GPU
165-
- 0.6B: batch_size=64
166-
- 4B: batch_size=4-8 ⚠️
167-
- 8B: Not recommended
165+
- 0.6B: batch_size=64 (supported)
166+
- 4B: batch_size=4-8 (limited)
167+
- 8B: Not recommended
168168

169169
#### 16GB GPU
170-
- 0.6B: batch_size=64
171-
- 4B: batch_size=32
172-
- 8B: batch_size=8 ⚠️
170+
- 0.6B: batch_size=64 (supported)
171+
- 4B: batch_size=32 (supported)
172+
- 8B: batch_size=8 (limited)
173173

174174
#### 24GB GPU
175-
- 0.6B: batch_size=64
176-
- 4B: batch_size=64
177-
- 8B: batch_size=32
175+
- 0.6B: batch_size=64 (supported)
176+
- 4B: batch_size=64 (supported)
177+
- 8B: batch_size=32 (supported)
178178

179179
---
180180

doc/appendix/hardware-requirements.zh.md

Lines changed: 12 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -23,9 +23,9 @@
2323
- 8B: 35 docs/s(**117倍于 CPU**
2424

2525
3. **24GB GPU 可运行的最大配置**
26-
- 0.6B: batch_size=64, seq_len=512
27-
- 4B: batch_size=64, seq_len=512
28-
- 8B: batch_size=64, seq_len=256 ✅(seq_len=512 时 OOM)
26+
- 0.6B: batch_size=64, seq_len=512(支持)
27+
- 4B: batch_size=64, seq_len=512(支持)
28+
- 8B: batch_size=64, seq_len=256(支持;seq_len=512 时 OOM)
2929

3030
4. **推荐生产配置**(留 20% 显存余量)
3131
- 0.6B: batch_size=32, seq_len=256(峰值显存 2.2 GB)
@@ -164,19 +164,19 @@
164164
### 不同显存容量的推荐配置
165165

166166
#### 8GB GPU
167-
- 0.6B: batch_size=64
168-
- 4B: batch_size=4-8 ⚠️
169-
- 8B: 不推荐
167+
- 0.6B: batch_size=64(支持)
168+
- 4B: batch_size=4-8(受限)
169+
- 8B: 不推荐
170170

171171
#### 16GB GPU
172-
- 0.6B: batch_size=64
173-
- 4B: batch_size=32
174-
- 8B: batch_size=8 ⚠️
172+
- 0.6B: batch_size=64(支持)
173+
- 4B: batch_size=32(支持)
174+
- 8B: batch_size=8(受限)
175175

176176
#### 24GB GPU
177-
- 0.6B: batch_size=64
178-
- 4B: batch_size=64
179-
- 8B: batch_size=32
177+
- 0.6B: batch_size=64(支持)
178+
- 4B: batch_size=64(支持)
179+
- 8B: batch_size=32(支持)
180180

181181
---
182182

doc/architecture/script_mapping.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -352,7 +352,7 @@ def _generate_sbert_embeddings(self):
352352

353353
### 5.2 CTM 模型 SBERT 使用确认
354354

355-
**结论**: **CTM 已完全使用 .env 中的 SBERT_MODEL_PATH**
355+
**结论**: **CTM 已完全使用 .env 中的 SBERT_MODEL_PATH**
356356

357357
1. **旧逻辑已移除**: CTM 不再从预计算的 `sbert_embeddings.npy` 文件读取
358358
2. **自动生成**: 如果 workspace 中没有 SBERT 嵌入,`train_ctm()` 会自动调用 `_generate_sbert_embeddings()`

doc/user-guide/preprocessing.md

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -204,10 +204,10 @@ python prepare_data.py \
204204
Expected output:
205205
```
206206
Checking preprocessed files for dataset: my_dataset
207-
BOW data: ./result/0.6B/my_dataset/bow/
208-
Embeddings: qwen_embeddings_zeroshot.npy (1024 dims)
209-
Vocabulary: vocab.pkl (5000 words)
210-
Document indices: doc_indices.npy
207+
OK BOW data: ./result/0.6B/my_dataset/bow/
208+
OK Embeddings: qwen_embeddings_zeroshot.npy (1024 dims)
209+
OK Vocabulary: vocab.pkl (5000 words)
210+
OK Document indices: doc_indices.npy
211211
All required files present.
212212
```
213213

doc/user-guide/preprocessing.zh.md

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -208,10 +208,10 @@ python prepare_data.py \
208208
预期输出:
209209
```
210210
检查数据集预处理文件:my_dataset
211-
BOW数据:./result/0.6B/my_dataset/bow/
212-
嵌入:qwen_embeddings_zeroshot.npy(1024维)
213-
词汇表:vocab.pkl(5000词)
214-
文档索引:doc_indices.npy
211+
OK BOW数据:./result/0.6B/my_dataset/bow/
212+
OK 嵌入:qwen_embeddings_zeroshot.npy(1024维)
213+
OK 词汇表:vocab.pkl(5000词)
214+
OK 文档索引:doc_indices.npy
215215
所有必需文件都存在。
216216
```
217217

doc/user-guide/scripts.md

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -61,9 +61,9 @@ bash scripts/02_clean_data.sh \
6161

6262
| Parameter | Required | Description | Default |
6363
|-----------|----------|-------------|---------|
64-
| `--input` | | Input CSV file or directory (docx/txt) | - |
65-
| `--language` | (not for preview) | Data language: english, chinese, german, spanish | - |
66-
| `--text_column` | (CSV mode) | Name of the text column to clean | - |
64+
| `--input` | Yes | Input CSV file or directory (docx/txt) | - |
65+
| `--language` | Yes (not for preview) | Data language: english, chinese, german, spanish | - |
66+
| `--text_column` | Yes (CSV mode) | Name of the text column to clean | - |
6767
| `--label_columns` | | Comma-separated label/metadata columns to keep as-is | - |
6868
| `--keep_all` | | Keep ALL original columns (only text column is cleaned) | false |
6969
| `--preview` | | Show CSV columns and sample rows, then exit | false |
@@ -149,8 +149,8 @@ bash scripts/03_prepare_data.sh --dataset mydata \
149149

150150
| Parameter | Required | Description | Default |
151151
|-----------|----------|-------------|---------|
152-
| `--dataset` | | Dataset name | - |
153-
| `--model` | | Target model: lda, hdp, stm (requires covariates), btm, nvdm, gsm, prodlda, ctm, etm, dtm, bertopic, theta | - |
152+
| `--dataset` | Yes | Dataset name | - |
153+
| `--model` | Yes | Target model: lda, hdp, stm (requires covariates), btm, nvdm, gsm, prodlda, ctm, etm, dtm, bertopic, theta | - |
154154
| `--model_size` | | Qwen model size (theta only): 0.6B, 4B, 8B | 0.6B |
155155
| `--mode` | | Embedding mode (theta only): zero_shot, unsupervised, supervised | zero_shot |
156156
| `--vocab_size` | | Vocabulary size | 5000 |
@@ -235,7 +235,7 @@ bash scripts/04_train_theta.sh \
235235

236236
| Parameter | Required | Description | Default |
237237
|-----------|----------|-------------|---------|
238-
| `--dataset` | | Dataset name | - |
238+
| `--dataset` | Yes | Dataset name | - |
239239
| `--model_size` | | Qwen model size: 0.6B, 4B, 8B | 0.6B |
240240
| `--mode` | | Embedding mode: zero_shot, unsupervised, supervised | zero_shot |
241241
| `--num_topics` | | Number of topics K | 20 |
@@ -467,8 +467,8 @@ bash scripts/05_train_baseline.sh \
467467

468468
| Parameter | Required | Description | Default |
469469
|-----------|----------|-------------|---------|
470-
| `--dataset` | | Dataset name | - |
471-
| `--models` | | Model list (comma-separated) | - |
470+
| `--dataset` | Yes | Dataset name | - |
471+
| `--models` | Yes | Model list (comma-separated) | - |
472472
| `--num_topics` | | Number of topics (ignored for hdp/bertopic) | 20 |
473473
| `--vocab_size` | | Vocabulary size | 5000 |
474474
| `--epochs` | | Training epochs (neural models) | 100 |

doc/user-guide/scripts.zh.md

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -61,9 +61,9 @@ bash scripts/02_clean_data.sh \
6161

6262
| 参数 | 必需 | 描述 | 默认值 |
6363
|-----------|----------|-------------|---------|
64-
| `--input` | | 输入的 CSV 文件或目录(docx/txt) | - |
65-
| `--language` | (预览模式不适用) | 数据语言:english, chinese, german, spanish | - |
66-
| `--text_column` | (CSV 模式) | 需要清洗的文本列名 | - |
64+
| `--input` | | 输入的 CSV 文件或目录(docx/txt) | - |
65+
| `--language` | (预览模式不适用) | 数据语言:english, chinese, german, spanish | - |
66+
| `--text_column` | (CSV 模式) | 需要清洗的文本列名 | - |
6767
| `--label_columns` | | 需要原样保留的标签/元数据列,逗号分隔 | - |
6868
| `--keep_all` | | 保留所有原始列(仅清洗文本列) | false |
6969
| `--preview` | | 显示 CSV 列和示例行后退出 | false |
@@ -149,8 +149,8 @@ bash scripts/03_prepare_data.sh --dataset mydata \
149149

150150
| 参数 | 必需 | 描述 | 默认值 |
151151
|-----------|----------|-------------|---------|
152-
| `--dataset` | | 数据集名称 | - |
153-
| `--model` | | 目标模型:lda, hdp, stm(需要协变量), btm, nvdm, gsm, prodlda, ctm, etm, dtm, bertopic, theta | - |
152+
| `--dataset` | | 数据集名称 | - |
153+
| `--model` | | 目标模型:lda, hdp, stm(需要协变量), btm, nvdm, gsm, prodlda, ctm, etm, dtm, bertopic, theta | - |
154154
| `--model_size` | | 通义千问模型规模(仅 theta):0.6B, 4B, 8B | 0.6B |
155155
| `--mode` | | 嵌入模式(仅 theta):zero_shot, unsupervised, supervised | zero_shot |
156156
| `--vocab_size` | | 词汇表大小 | 5000 |
@@ -235,7 +235,7 @@ bash scripts/04_train_theta.sh \
235235

236236
| 参数 | 必需 | 描述 | 默认值 |
237237
|-----------|----------|-------------|---------|
238-
| `--dataset` | | 数据集名称 | - |
238+
| `--dataset` | | 数据集名称 | - |
239239
| `--model_size` | | 通义千问模型规模:0.6B, 4B, 8B | 0.6B |
240240
| `--mode` | | 嵌入模式:zero_shot, unsupervised, supervised | zero_shot |
241241
| `--num_topics` | | 主题数量 K | 20 |
@@ -465,8 +465,8 @@ bash scripts/05_train_baseline.sh \
465465

466466
| 参数 | 必需 | 描述 | 默认值 |
467467
|-----------|----------|-------------|---------|
468-
| `--dataset` | | 数据集名称 | - |
469-
| `--models` | | 模型列表(逗号分隔) | - |
468+
| `--dataset` | | 数据集名称 | - |
469+
| `--models` | | 模型列表(逗号分隔) | - |
470470
| `--num_topics` | | 主题数量(hdp/bertopic 忽略此参数) | 20 |
471471
| `--vocab_size` | | 词汇表大小 | 5000 |
472472
| `--epochs` | | 训练轮数(神经模型) | 100 |

0 commit comments

Comments
 (0)