chore: 删除未使用的 bge-base/bge-large/punct 模型
- bge-base-zh-v1.5、bge-large-zh-v1.5：只有 tokenizer/config 元数据， onnx 权重从未提交（README 一直写"需自行下载”），代码里也完全没有 任何地方引用这两个目录，纯占位死文件。 - punct（标点恢复模型，75M onnx）：electron-builder.config.ts 的 extraResources 从未把它纳入打包，装出来的应用里从来没有过这份模型， 语音识别的标点恢复功能实际一直是禁用状态。用户确认现阶段不需要， 直接删除源文件；代码逻辑保留（punctuationModelDir 为空时优雅跳过）， 以后要用可以自行准备模型再在设置里指定目录。 models/ 从 ~475M 降到 399M。README 目录说明与模型文件章节同步更新。 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
ssurui
parent
33ce4ec09c
commit
d686e600be
13
README.md
13
README.md
@ -42,11 +42,8 @@ elec_rag_ui/
|
||||
│ │ └── types/ # TypeScript 类型定义
|
||||
│ └── batch_questions.txt # 批量测试问题示例
|
||||
├── models/ # 本地模型（onnx 由 Git LFS 管理）
|
||||
│ ├── bge-small-zh-v1.5/ # 嵌入模型（512 维，~24MB）
|
||||
│ ├── bge-base-zh-v1.5/ # 嵌入模型（768 维，~98MB，onnx 需自行下载）
|
||||
│ ├── bge-large-zh-v1.5/ # 嵌入模型（1024 维，~312MB，onnx 需自行下载）
|
||||
│ ├── asr/ # Paraformer 流式语音识别模型（~376MB）
|
||||
│ └── punct/ # 标点恢复模型（~72MB）
|
||||
│ ├── bge-small-zh-v1.5/ # 嵌入模型（512 维，~24MB），本地 CPU 向量化回退用
|
||||
│ └── asr/ # Paraformer 流式语音识别模型（~376MB）
|
||||
├── docs/ # 示例文档（用于测试索引与问答）
|
||||
├── resources/ # 应用图标、TTS/ASR worker
|
||||
├── stubs/sharp/ # sharp 原生模块 stub
|
||||
@ -79,9 +76,9 @@ git lfs install
|
||||
GIT_SSL_NO_VERIFY=true git lfs pull
|
||||
```
|
||||
|
||||
**bge-base / bge-large** 的 onnx 文件需自行从 HuggingFace 下载：
|
||||
- [bge-base-zh-v1.5](https://huggingface.co/BAAI/bge-base-zh-v1.5)（下载 `onnx/model_quantized.onnx`，约 98MB）
|
||||
- [bge-large-zh-v1.5](https://huggingface.co/BAAI/bge-large-zh-v1.5)（下载 `onnx/model_quantized.onnx`，约 312MB）
|
||||
> 语音识别默认不带自动加标点（标点恢复模型未随包分发）；如需要可自行准备
|
||||
> [sherpa-onnx punct 模型](https://k2-fsa.github.io/sherpa/onnx/punctuation/models.html)，
|
||||
> 在设置里把"标点恢复模型目录"指过去即可，代码本身已支持该功能。
|
||||
|
||||
## 开发模式
|
||||
|
||||
|
||||
@ -1,39 +0,0 @@
|
||||
{
|
||||
"_name_or_path": "BAAI/bge-base-zh-v1.5",
|
||||
"architectures": [
|
||||
"BertModel"
|
||||
],
|
||||
"attention_probs_dropout_prob": 0.1,
|
||||
"bos_token_id": 0,
|
||||
"classifier_dropout": null,
|
||||
"directionality": "bidi",
|
||||
"eos_token_id": 2,
|
||||
"hidden_act": "gelu",
|
||||
"hidden_dropout_prob": 0.1,
|
||||
"hidden_size": 768,
|
||||
"id2label": {
|
||||
"0": "LABEL_0"
|
||||
},
|
||||
"initializer_range": 0.02,
|
||||
"intermediate_size": 3072,
|
||||
"label2id": {
|
||||
"LABEL_0": 0
|
||||
},
|
||||
"layer_norm_eps": 1e-12,
|
||||
"max_position_embeddings": 512,
|
||||
"model_type": "bert",
|
||||
"num_attention_heads": 12,
|
||||
"num_hidden_layers": 12,
|
||||
"output_past": true,
|
||||
"pad_token_id": 0,
|
||||
"pooler_fc_size": 768,
|
||||
"pooler_num_attention_heads": 12,
|
||||
"pooler_num_fc_layers": 3,
|
||||
"pooler_size_per_head": 128,
|
||||
"pooler_type": "first_token_transform",
|
||||
"position_embedding_type": "absolute",
|
||||
"transformers_version": "4.34.0.dev0",
|
||||
"type_vocab_size": 2,
|
||||
"use_cache": true,
|
||||
"vocab_size": 21128
|
||||
}
|
||||
@ -1,7 +0,0 @@
|
||||
{
|
||||
"cls_token": "[CLS]",
|
||||
"mask_token": "[MASK]",
|
||||
"pad_token": "[PAD]",
|
||||
"sep_token": "[SEP]",
|
||||
"unk_token": "[UNK]"
|
||||
}
|
||||
File diff suppressed because it is too large
Load Diff
@ -1,15 +0,0 @@
|
||||
{
|
||||
"clean_up_tokenization_spaces": true,
|
||||
"cls_token": "[CLS]",
|
||||
"do_basic_tokenize": true,
|
||||
"do_lower_case": true,
|
||||
"mask_token": "[MASK]",
|
||||
"model_max_length": 512,
|
||||
"never_split": null,
|
||||
"pad_token": "[PAD]",
|
||||
"sep_token": "[SEP]",
|
||||
"strip_accents": null,
|
||||
"tokenize_chinese_chars": true,
|
||||
"tokenizer_class": "BertTokenizer",
|
||||
"unk_token": "[UNK]"
|
||||
}
|
||||
@ -1,39 +0,0 @@
|
||||
{
|
||||
"_name_or_path": "BAAI/bge-large-zh-v1.5",
|
||||
"architectures": [
|
||||
"BertModel"
|
||||
],
|
||||
"attention_probs_dropout_prob": 0.1,
|
||||
"bos_token_id": 0,
|
||||
"classifier_dropout": null,
|
||||
"directionality": "bidi",
|
||||
"eos_token_id": 2,
|
||||
"hidden_act": "gelu",
|
||||
"hidden_dropout_prob": 0.1,
|
||||
"hidden_size": 1024,
|
||||
"id2label": {
|
||||
"0": "LABEL_0"
|
||||
},
|
||||
"initializer_range": 0.02,
|
||||
"intermediate_size": 4096,
|
||||
"label2id": {
|
||||
"LABEL_0": 0
|
||||
},
|
||||
"layer_norm_eps": 1e-12,
|
||||
"max_position_embeddings": 512,
|
||||
"model_type": "bert",
|
||||
"num_attention_heads": 16,
|
||||
"num_hidden_layers": 24,
|
||||
"output_past": true,
|
||||
"pad_token_id": 0,
|
||||
"pooler_fc_size": 768,
|
||||
"pooler_num_attention_heads": 12,
|
||||
"pooler_num_fc_layers": 3,
|
||||
"pooler_size_per_head": 128,
|
||||
"pooler_type": "first_token_transform",
|
||||
"position_embedding_type": "absolute",
|
||||
"transformers_version": "4.34.0.dev0",
|
||||
"type_vocab_size": 2,
|
||||
"use_cache": true,
|
||||
"vocab_size": 21128
|
||||
}
|
||||
@ -1,7 +0,0 @@
|
||||
{
|
||||
"cls_token": "[CLS]",
|
||||
"mask_token": "[MASK]",
|
||||
"pad_token": "[PAD]",
|
||||
"sep_token": "[SEP]",
|
||||
"unk_token": "[UNK]"
|
||||
}
|
||||
File diff suppressed because it is too large
Load Diff
@ -1,15 +0,0 @@
|
||||
{
|
||||
"clean_up_tokenization_spaces": true,
|
||||
"cls_token": "[CLS]",
|
||||
"do_basic_tokenize": true,
|
||||
"do_lower_case": true,
|
||||
"mask_token": "[MASK]",
|
||||
"model_max_length": 1000000000000000019884624838656,
|
||||
"never_split": null,
|
||||
"pad_token": "[PAD]",
|
||||
"sep_token": "[SEP]",
|
||||
"strip_accents": null,
|
||||
"tokenize_chinese_chars": true,
|
||||
"tokenizer_class": "BertTokenizer",
|
||||
"unk_token": "[UNK]"
|
||||
}
|
||||
@ -1,56 +0,0 @@
|
||||
# Introduction
|
||||
|
||||
This model is converted from
|
||||
https://modelscope.cn/models/iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch/summary
|
||||
|
||||
The commands to generate the onnx model are given below:
|
||||
|
||||
```
|
||||
pip install funasr modelscope
|
||||
pip install kaldi-native-fbank torchaudio onnx onnxruntime
|
||||
|
||||
mkdir -p /tmp/models
|
||||
cd /tmp/models
|
||||
|
||||
git clone https://www.modelscope.cn/iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch.git
|
||||
cd punc_ct-transformer_zh-cn-common-vocab272727-pytorch
|
||||
git lfs pull --include model.pt
|
||||
|
||||
cd /tmp
|
||||
git clone https://github.com/alibaba-damo-academy/FunASR
|
||||
cd FunASR/runtime/python/onnxruntime
|
||||
|
||||
cat >export-onnx.py <<EOF
|
||||
|
||||
from funasr_onnx import CT_Transformer
|
||||
model_dir = "/tmp/punc_ct-transformer_zh-cn-common-vocab272727-pytorch" # model = CT_Transformer(model_dir, quantize=True) model = CT_Transformer(model_dir)
|
||||
EOF
|
||||
|
||||
chmod +x export-onnx.py
|
||||
|
||||
./export-onnx.py
|
||||
```
|
||||
|
||||
You will find the exported
|
||||
`model.onnx` file inside
|
||||
`/tmp/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch`.
|
||||
|
||||
Now you can use ./add-model-metadata.py in this repo to add metadata to the generated
|
||||
`model.onnx`.
|
||||
|
||||
You can use
|
||||
```
|
||||
from onnxruntime.quantization import QuantType, quantize_dynamic
|
||||
|
||||
quantize_dynamic(
|
||||
model_input="./model.onnx",
|
||||
model_output="./model.int8.onnx",
|
||||
weight_type=QuantType.QUInt8,
|
||||
)
|
||||
```
|
||||
to get the `int8` quantized model
|
||||
|
||||
```
|
||||
-rw-r--r-- 1 fangjun staff 72M Jun 18 10:33 model.int8.onnx
|
||||
-rw-r--r-- 1 fangjun staff 281M Apr 12 2024 model.onnx
|
||||
```
|
||||
@ -1,46 +0,0 @@
|
||||
model: CTTransformer
|
||||
model_conf:
|
||||
ignore_id: 0
|
||||
embed_unit: 256
|
||||
att_unit: 256
|
||||
dropout_rate: 0.1
|
||||
punc_list:
|
||||
- <unk>
|
||||
- _
|
||||
- ，
|
||||
- 。
|
||||
- ？
|
||||
- 、
|
||||
punc_weight:
|
||||
- 1.0
|
||||
- 1.0
|
||||
- 1.0
|
||||
- 1.0
|
||||
- 1.0
|
||||
- 1.0
|
||||
sentence_end_id: 3
|
||||
|
||||
encoder: SANMEncoder
|
||||
encoder_conf:
|
||||
input_size: 256
|
||||
output_size: 256
|
||||
attention_heads: 8
|
||||
linear_units: 1024
|
||||
num_blocks: 4
|
||||
dropout_rate: 0.1
|
||||
positional_dropout_rate: 0.1
|
||||
attention_dropout_rate: 0.0
|
||||
input_layer: pe
|
||||
pos_enc_class: SinusoidalPositionEncoder
|
||||
normalize_before: true
|
||||
kernel_size: 11
|
||||
sanm_shfit: 0
|
||||
selfattention_layer_type: sanm
|
||||
padding_idx: 0
|
||||
|
||||
tokenizer: CharTokenizer
|
||||
tokenizer_conf:
|
||||
unk_symbol: <unk>
|
||||
|
||||
|
||||
|
||||
BIN
models/punct/sherpa-onnx-punct-ct-transformer-zh-en-vocab272727-2024-04-12-int8/model.int8.onnx
(Stored with Git LFS)
BIN
models/punct/sherpa-onnx-punct-ct-transformer-zh-en-vocab272727-2024-04-12-int8/model.int8.onnx
(Stored with Git LFS)
Binary file not shown.
File diff suppressed because it is too large
Load Diff
Loading…
x
Reference in New Issue
Block a user