Skip to content

Commit e1acc38

Browse files
author
litongjava
committed
add document
1 parent 76df083 commit e1acc38

File tree

2 files changed

+103
-85
lines changed

2 files changed

+103
-85
lines changed

‎README.md

Lines changed: 21 additions & 85 deletions
Original file line numberDiff line numberDiff line change
@@ -11,93 +11,14 @@ options:
1111
-m FNAME, --model FNAME [models/ggml-base.en.bin] model path
1212
-di, --diarize [false ] stereo audio diarization
1313
```
14-
## service
14+
## whisper_http_server_base_httplib
1515

16-
Simple http service. WAV Files are passed to the inference model via http requests.
16+
Simple http service. WAV mp4 and m4a Files are passed to the inference model via http requests.
1717

1818
```
19-
./cmake-build-debug/service -m models/ggml-base.en.bin
20-
```
21-
22-
```shell
23-
whisper_init_from_file_with_params_no_state: loading model from 'models/ggml-base.en.bin'
24-
whisper_model_load: loading model
25-
whisper_model_load: n_vocab = 51864
26-
whisper_model_load: n_audio_ctx = 1500
27-
whisper_model_load: n_audio_state = 512
28-
whisper_model_load: n_audio_head = 8
29-
whisper_model_load: n_audio_layer = 6
30-
whisper_model_load: n_text_ctx = 448
31-
whisper_model_load: n_text_state = 512
32-
whisper_model_load: n_text_head = 8
33-
whisper_model_load: n_text_layer = 6
34-
whisper_model_load: n_mels = 80
35-
whisper_model_load: ftype = 1
36-
whisper_model_load: qntvr = 0
37-
whisper_model_load: type = 2 (base)
38-
whisper_model_load: adding 1607 extra tokens
39-
whisper_model_load: n_langs = 99
40-
whisper_backend_init: using Metal backend
41-
ggml_metal_init: allocating
42-
ggml_metal_init: found device: Apple M2
43-
ggml_metal_init: picking default device: Apple M2
44-
ggml_metal_init: default.metallib not found, loading from source
45-
ggml_metal_init: error: could not use bundle path to find ggml-metal.metal, falling back to trying cwd
46-
ggml_metal_init: loading 'ggml-metal.metal'
47-
ggml_metal_init: GPU name: Apple M2
48-
ggml_metal_init: GPU family: MTLGPUFamilyApple8 (1008)
49-
ggml_metal_init: hasUnifiedMemory = true
50-
ggml_metal_init: recommendedMaxWorkingSetSize = 11453.25 MB
51-
ggml_metal_init: maxTransferRate = built-in GPU
52-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 156.68 MB, ( 157.20 / 11453.25)
53-
whisper_model_load: Metal buffer size = 156.67 MB
54-
whisper_model_load: model size = 156.58 MB
55-
whisper_backend_init: using Metal backend
56-
ggml_metal_init: allocating
57-
ggml_metal_init: found device: Apple M2
58-
ggml_metal_init: picking default device: Apple M2
59-
ggml_metal_init: default.metallib not found, loading from source
60-
ggml_metal_init: error: could not use bundle path to find ggml-metal.metal, falling back to trying cwd
61-
ggml_metal_init: loading 'ggml-metal.metal'
62-
ggml_metal_init: GPU name: Apple M2
63-
ggml_metal_init: GPU family: MTLGPUFamilyApple8 (1008)
64-
ggml_metal_init: hasUnifiedMemory = true
65-
ggml_metal_init: recommendedMaxWorkingSetSize = 11453.25 MB
66-
ggml_metal_init: maxTransferRate = built-in GPU
67-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 16.52 MB, ( 173.72 / 11453.25)
68-
whisper_init_state: kv self size = 16.52 MB
69-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 18.43 MB, ( 192.15 / 11453.25)
70-
whisper_init_state: kv cross size = 18.43 MB
71-
whisper_init_state: loading Core ML model from 'models/ggml-base.en-encoder.mlmodelc'
72-
whisper_init_state: first run on a device may take a while ...
73-
whisper_init_state: Core ML model loaded
74-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 0.02 MB, ( 196.51 / 11453.25)
75-
whisper_init_state: compute buffer (conv) = 5.67 MB
76-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 0.02 MB, ( 196.53 / 11453.25)
77-
whisper_init_state: compute buffer (cross) = 4.71 MB
78-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 0.02 MB, ( 196.54 / 11453.25)
79-
whisper_init_state: compute buffer (decode) = 96.41 MB
80-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 4.05 MB, ( 200.59 / 11453.25)
81-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 3.08 MB, ( 203.67 / 11453.25)
82-
ggml_metal_add_buffer: allocated 'backend ' buffer, size = 94.78 MB, ( 298.45 / 11453.25)
83-
84-
whisper service listening at http://0.0.0.0:8080
85-
86-
Received request: jfk.wav
87-
Successfully loaded jfk.wav
88-
89-
system_info: n_threads = 4 / 8 | AVX = 0 | AVX2 = 0 | AVX512 = 0 | FMA = 0 | NEON = 1 | ARM_FMA = 1 | METAL = 1 | F16C = 0 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | SSSE3 = 0 | VSX = 0 | CUDA = 0 | COREML = 1 | OPENVINO = 0 |
90-
91-
handleInference: processing 'jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 processors, lang = en, task = transcribe, timestamps = 1 ...
92-
93-
Running whisper.cpp inference on jfk.wav
94-
95-
[00:00:00.000 --> 00:00:11.000] And so my fellow Americans, ask not what your country can do for you, ask what you can do for your country.
96-
```
97-
```
98-
./service -h
19+
./whisper_http_server_base_httplib -h
9920
100-
usage: ./bin/service [options]
21+
usage: ./bin/whisper_http_server_base_httplib [options]
10122
10223
options:
10324
-h, --help [default] show this help message and exit
@@ -131,7 +52,12 @@ options:
13152
--host HOST, [127.0.0.1] Hostname/ip-adress for the service
13253
--port PORT, [8080 ] Port number for the service
13354
```
134-
55+
## start whisper_http_server_base_httplib
56+
```
57+
./cmake-build-debug/whisper_http_server_base_httplib -m models/ggml-base.en.bin
58+
```
59+
Test server
60+
see request doc in [doc](doc)
13561
## request examples
13662

13763
**/inference**
@@ -140,11 +66,21 @@ curl --location --request POST http://127.0.0.1:8080/inference \
14066
--form file=@"./samples/jfk.wav" \
14167
--form temperature="0.2" \
14268
--form response-format="json"
69+
--form audio_format="wav"
14370
```
14471

14572
**/load**
14673
```
14774
curl 127.0.0.1:8080/load \
14875
-H "Content-Type: multipart/form-data" \
14976
-F model="<path-to-model-file>"
150-
```
77+
```
78+
79+
## whisper_server_base_on_uwebsockets
80+
web socket server
81+
start server
82+
```
83+
./cmake-build-debug/whisper_server_base_on_uwebsockets -m models/ggml-base.en.bin
84+
```
85+
Test server
86+
see python [client](client)

‎doc/whiser-cpp-server.md

Lines changed: 82 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,82 @@
1+
---
2+
title: whiser-cpp-server v1.0.0
3+
language_tabs:
4+
- shell: Shell
5+
- http: HTTP
6+
- javascript: JavaScript
7+
- ruby: Ruby
8+
- python: Python
9+
- php: PHP
10+
- java: Java
11+
- go: Go
12+
toc_footers: []
13+
includes: []
14+
search: true
15+
code_clipboard: true
16+
highlight_theme: darkula
17+
headingLevel: 2
18+
generator: "@tarslib/widdershins v4.0.17"
19+
20+
---
21+
22+
# whiser-cpp-server
23+
24+
> v1.0.0
25+
26+
Base URLs:
27+
28+
# Authentication
29+
30+
# Default
31+
32+
## POST /inference
33+
34+
POST /inference
35+
36+
> Body 请求参数
37+
38+
```yaml
39+
file: file://1.m4a
40+
temperature: "0.2"
41+
response-format: json
42+
audio_format: m4a
43+
44+
```
45+
46+
### 请求参数
47+
48+
|名称|位置|类型|必选|说明|
49+
|---|---|---|---|---|
50+
|body|body|object| 否 |none|
51+
|» file|body|string(binary)| 否 |filename|
52+
|» temperature|body|string| 否 |none|
53+
|» response-format|body|string| 否 |none|
54+
|» audio_format|body|string| 否 |audio format,support m4a,mp3,and wav|
55+
56+
> 返回示例
57+
58+
> 成功
59+
60+
```json
61+
{
62+
"code": 0,
63+
"data": [
64+
{
65+
"sentence": " And so my fellow Americans ask not what your country can do for you, ask what you can do for your country.",
66+
"t0": 0,
67+
"t1": 1100
68+
}
69+
]
70+
}
71+
```
72+
73+
### 返回结果
74+
75+
|状态码|状态码含义|说明|数据模型|
76+
|---|---|---|---|
77+
|200|[OK](https://tools.ietf.org/html/rfc7231#section-6.3.1)|成功|Inline|
78+
79+
### 返回数据结构
80+
81+
# 数据模型
82+

0 commit comments

Comments
 (0)