Spaces:
Build error
Build error
Upload folder using huggingface_hub
Browse files- .gitignore +2 -0
- .vscode/launch.json +17 -0
- 0000.jpeg +0 -0
- 0001.jpg +0 -0
- README.md +3 -9
- gradio_cached_examples/15/log.csv +2 -0
- main.py +83 -0
- multi_main.py +105 -0
- requirements.txt +78 -0
- run.sh +9 -0
- setup.sh +2 -0
- static/0000.jpeg +0 -0
- static/0001.jpg +0 -0
- static/image_1149d78e-43f6-4d5b-8e63-e24c2641012b.jpg +0 -0
- static/image_11bc11b6-57a9-42d3-81fc-97d957e62f28.jpg +0 -0
- static/image_33685dfc-b96f-401c-847d-3dd537186ec2.jpg +0 -0
- static/image_4a883844-b1b2-4154-bbfe-0b108d93dca6.jpg +0 -0
- static/image_4e43dfc4-9362-4e8d-b20f-d9816a9b684d.jpg +0 -0
- static/image_51e314f6-0935-4430-a53f-bfdc22ca7cc4.jpg +0 -0
- static/image_6c63b8a2-8bfc-485e-a238-94e76f22e7db.jpg +0 -0
- static/image_82c514bf-bdc1-4d60-b073-a97b5816ea56.jpg +0 -0
- static/image_845cd65a-30a7-43c3-be8c-819bc04de98d.jpg +0 -0
- static/image_86804064-1f0e-4130-bcc3-30dd839a3c0a.jpg +0 -0
- static/image_9afab29f-77d9-4451-8a52-092a5c37625d.jpg +0 -0
- static/image_9c5c694a-acef-4545-8759-75ccffa39f6d.jpg +0 -0
- static/image_b73c1efa-8112-404c-ab9e-eddb870f43af.jpg +0 -0
- static/image_bbacb8a4-89d9-45ac-9da2-28717009e750.jpg +0 -0
- static/image_bd087d82-9ed6-4b68-8ac1-c1417cfbb995.jpg +0 -0
- static/image_ca9e5436-bc9b-4242-9a02-261d2209feb3.jpg +0 -0
- static/image_dd018613-a2be-4eef-8183-a01929c835fd.jpg +0 -0
- static/image_e63e5e5c-41cc-4f42-b896-95644d6b28f3.jpg +0 -0
- static/image_e6c53092-7b49-4ba5-96dc-38e9479d3c6a.jpg +0 -0
- static/image_f9bef0de-bc72-415d-b527-6c068594af02.jpg +0 -0
- test_hd.py +187 -0
- try_demo.py +224 -0
- try_demo_demo.py +206 -0
- try_grpc.py +139 -0
- try_hd.py +137 -0
- try_hd_v2.py +218 -0
.gitignore
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
flagged
|
2 |
+
*.log
|
.vscode/launch.json
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
// 使用 IntelliSense 了解相关属性。
|
3 |
+
// 悬停以查看现有属性的描述。
|
4 |
+
// 欲了解更多信息,请访问: https://go.microsoft.com/fwlink/?linkid=830387
|
5 |
+
"version": "0.2.0",
|
6 |
+
"configurations": [
|
7 |
+
{
|
8 |
+
"name": "Python: 当前文件",
|
9 |
+
"type": "python",
|
10 |
+
"request": "launch",
|
11 |
+
"program": "test_hd.py",
|
12 |
+
"console": "integratedTerminal",
|
13 |
+
"python" :"/home/hadoop-automl/tianrunhe/anaconda3/envs/gradio_env_3.9/bin/python"
|
14 |
+
// "python":"/home/hadoop-automl/tianrunhe/anaconda3/envs/gradio/bin/python"
|
15 |
+
}
|
16 |
+
]
|
17 |
+
}
|
0000.jpeg
ADDED
0001.jpg
ADDED
README.md
CHANGED
@@ -1,12 +1,6 @@
|
|
1 |
---
|
2 |
-
title:
|
3 |
-
|
4 |
-
colorFrom: indigo
|
5 |
-
colorTo: indigo
|
6 |
sdk: gradio
|
7 |
-
sdk_version: 4.31.
|
8 |
-
app_file: app.py
|
9 |
-
pinned: false
|
10 |
---
|
11 |
-
|
12 |
-
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
1 |
---
|
2 |
+
title: gradio_demo
|
3 |
+
app_file: try_grpc.py
|
|
|
|
|
4 |
sdk: gradio
|
5 |
+
sdk_version: 4.31.4
|
|
|
|
|
6 |
---
|
|
|
|
gradio_cached_examples/15/log.csv
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
output,flag,username,timestamp
|
2 |
+
,,,2024-05-21 21:35:52.230145
|
main.py
ADDED
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from TritonServerClient import TritonServerClient, InferInput, InferRequestedOutput
|
3 |
+
from TritonServerClient.utils import np_to_triton_dtype
|
4 |
+
from functools import wraps
|
5 |
+
import numpy as np
|
6 |
+
|
7 |
+
def prepare_params(query):
|
8 |
+
bs=1
|
9 |
+
# query = "#角色:直播间文案写作专家。#任务:根据用户输入的商品信息写出用于推广营销的直播文案。#目标:引导用户购买直播间商品,提高销量。#要求:- 如果没有输入商品信息,则邀请输入。- 文案写作 - 按照“引出话题”、“产品讲解”、“购买方式”三个方面写作文案。 - 在“产品讲解”部分详细讲解产品细节: - 【重要】充分扩充内容,吸引用户购买。 - 详细描述商品的优惠活动信息(如用户输入没有,不要询问,直接补充)。 - 在“购买方式”部分: - 仅包括引导用户点击链接购买的内容。 - 不要包括快递物流承诺。 - 【重要】每个部分多于5句话。 - 使用口语化的文字。 - 表达方式有亲和力和感染力。 - 对用户们使用“宝子们”、“宝宝们”、“家人们”等有亲和力的称呼。 - 多使用自问自答的语句。- 当你准备好了回复“请输入商品描述和优惠信息”我们就开始。#输出:- 包含在""里。- 写出每一部分内容的标题,包括【引出话题】、【产品讲解】、【购买方式】。- 示例:""【引出话题】【产品讲解】【购买方式】""- 输入商品信息:%s" % prompt
|
10 |
+
# query = '请你扮演直播间直播卖货的主播角色,根据用户提出的问题,检索知识库中的信息,如果查找到相关的信息,使用口语化的文字,用非常简短的一句话回答用户的问题;如果查找不到相关的信息,直接回答没有,并引导用户关注本场直播商品。你的目的是激发用户的购买欲、提高直播商品销量。回答对用户们使用“宝子”、“宝宝”、“亲”等有亲和力的称呼。\n\n# 用户提出的问题:\n这里面有可乐吗\n\n# 知识库:\n商品名称:外送欢乐桶(券)兑换券\n直播链接序号:64号链接号链接\n商品现价:66元\n商品原价:142元\n折扣:5\n品牌名称:麦当劳麦乐送\n券包张数:1\n套餐详情:["麦辣鸡腿汉堡1个","板烧鸡腿堡1个","中可乐3杯","四拼小食桶A"]\n补充:四拼小食桶A(麦麦脆汁鸡(鸡腿)1块+中薯条1份+麦辣鸡翅2块+麦乐鸡5块)+麦辣鸡腿堡1个+板烧鸡腿堡1个+中可乐3杯 原材料:面包\n\n\n# 输出格式:\n答:\n\n# 要求\n对于用户问到了知识库中未提及的信息不要编造,直接不回答'
|
11 |
+
|
12 |
+
title_text = np.array([query.encode('utf-8')], dtype=np.string_)
|
13 |
+
title_text = np.tile(title_text, (bs, 1))
|
14 |
+
|
15 |
+
data_batch = {}
|
16 |
+
data_batch['query'] = title_text
|
17 |
+
|
18 |
+
inputs = [
|
19 |
+
InferInput("query", data_batch['query'].shape,
|
20 |
+
np_to_triton_dtype(data_batch['query'].dtype)),
|
21 |
+
]
|
22 |
+
|
23 |
+
inputs[0].set_data_from_numpy(data_batch['query'])
|
24 |
+
|
25 |
+
return inputs
|
26 |
+
|
27 |
+
|
28 |
+
def make_a_try(inputs, outputs='response', model_name='qwen', model_version='1'):
|
29 |
+
outputs_list = []
|
30 |
+
ori_outputs_list = outputs.strip().split(",")
|
31 |
+
for out_ele in ori_outputs_list:
|
32 |
+
outputs_list.append(out_ele.strip())
|
33 |
+
outputs = [InferRequestedOutput(x) for x in outputs_list]
|
34 |
+
|
35 |
+
response = my_client.predict(model_name=model_name, inputs=inputs, model_version=model_version, outputs=outputs)
|
36 |
+
|
37 |
+
rsp_info = {}
|
38 |
+
if outputs_list == []:
|
39 |
+
for out_name_ele in response._result.outputs:
|
40 |
+
outputs_list.append(out_name_ele.name)
|
41 |
+
for output_name in outputs_list:
|
42 |
+
res = response.as_numpy(output_name)
|
43 |
+
response = np.expand_dims(res, axis=0)
|
44 |
+
response = response[0].decode('utf-8')
|
45 |
+
rsp_info[output_name] = response
|
46 |
+
|
47 |
+
return rsp_info['response']
|
48 |
+
|
49 |
+
|
50 |
+
|
51 |
+
def greet(prompt):
|
52 |
+
"""Greet someone."""
|
53 |
+
# print(prompt)
|
54 |
+
inputs = prepare_params(prompt)
|
55 |
+
result = make_a_try(inputs)
|
56 |
+
|
57 |
+
return result
|
58 |
+
|
59 |
+
|
60 |
+
if __name__ == "__main__":
|
61 |
+
param_info = {}
|
62 |
+
param_info['appkey'] = "com.sankuai.automl.serving"
|
63 |
+
# param_info['appkey'] = "com.sankuai.automl.streamvlm"
|
64 |
+
|
65 |
+
param_info['remote_appkey'] = "com.sankuai.automl.chat3"
|
66 |
+
param_info['model_name'] = "qwen"
|
67 |
+
param_info['model_version'] = "1"
|
68 |
+
param_info['time_out'] = 60000
|
69 |
+
param_info['server_targets'] = []
|
70 |
+
param_info['outputs'] = 'response'
|
71 |
+
|
72 |
+
appkey, remote_appkey, model_name, model_version, time_out, server_targets = param_info['appkey'], param_info['remote_appkey'], param_info['model_name'], param_info['model_version'], param_info['time_out'], param_info['server_targets']
|
73 |
+
|
74 |
+
# my_client = TritonServerClient(appkey=appkey, remote_appkey=remote_appkey, time_out=time_out, server_targets=server_targets)
|
75 |
+
|
76 |
+
# 以上部分,用户可以修改代码
|
77 |
+
demo = gr.Interface(
|
78 |
+
fn=greet,
|
79 |
+
inputs=["textbox"],
|
80 |
+
outputs=["textbox"],
|
81 |
+
)
|
82 |
+
|
83 |
+
demo.launch(server_name="0.0.0.0", server_port=8088, debug=True, share=True)
|
multi_main.py
ADDED
@@ -0,0 +1,105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from TritonServerClient import TritonServerClient, InferInput, InferRequestedOutput
|
3 |
+
from TritonServerClient.utils import np_to_triton_dtype
|
4 |
+
from functools import wraps
|
5 |
+
import numpy as np
|
6 |
+
|
7 |
+
|
8 |
+
def prepare_params(query):
|
9 |
+
bs=1
|
10 |
+
# query = "#角色:直播间文案写作专家。#任务:根据用户输入的商品信息写出用于推广营销的直播文案。#目标:引导用户购买直播间商品,提高销量。#要求:- 如果没有输入商品信息,则邀请输入。- 文案写作 - 按照“引出话题”、“产品讲解”、“购买方式”三个方面写作文案。 - 在“产品讲解”部分详细讲解产品细节: - 【重要】充分扩充内容,吸引用户购买。 - 详细描述商品的优惠活动信息(如用户输入没有,不要询问,直接补充)。 - 在“购买方式”部分: - 仅包括引导用户点击链接购买的内容。 - 不要包括快递物流承诺。 - 【重要】每个部分多于5句话。 - 使用口语化的文字。 - 表达方式有亲和力和感染力。 - 对用户们使用“宝子们”、“宝宝们”、“家人们”等有亲和力的称呼。 - 多使用自问自答的语句。- 当你准备好了回复“请输入商品描述和优惠信息”我们就开始。#输出:- 包含在""里。- 写出每一部分内容的标题,包括【引出话题】、【产品讲解】、【购买方式】。- 示例:""【引出话题】【产品讲解】【购买方式】""- 输入商品信息:%s" % prompt
|
11 |
+
# query = '请你扮演直播间直播卖货的主播角色,根据用户提出的问题,检索知识库中的信息,如果查找到相关的信息,使用口语化的文字,用非常简短的一句话回答用户的问题;如果查找不到相关的信息,直接回答没有,并引导用户关注本场直播商品。你的目的是激发用户的购买欲、提高直播商品销量。回答对用户们使用“宝子”、“宝宝”、“亲”等有亲和力的称呼。\n\n# 用户提出的问题:\n这里面有可乐吗\n\n# 知识库:\n商品名称:外送欢乐桶(券)兑换券\n直播链接序号:64号链接号链接\n商品现价:66元\n商品原价:142元\n折扣:5\n品牌名称:麦当劳麦乐送\n券包张数:1\n套餐详情:["麦辣鸡腿汉堡1个","板烧鸡腿堡1个","中可乐3杯","四拼小食桶A"]\n补充:四拼小食桶A(麦麦脆汁鸡(鸡腿)1块+中薯条1份+麦辣鸡翅2块+麦乐鸡5块)+麦辣鸡腿堡1个+板烧鸡腿堡1个+中可乐3杯 原材料:面包\n\n\n# 输出格式:\n答:\n\n# 要求\n对于用户问到了知识库中未提及的信息不要编造,直接不回答'
|
12 |
+
|
13 |
+
title_text = np.array([query.encode('utf-8')], dtype=np.string_)
|
14 |
+
title_text = np.tile(title_text, (bs, 1))
|
15 |
+
|
16 |
+
data_batch = {}
|
17 |
+
data_batch['query'] = title_text
|
18 |
+
|
19 |
+
inputs = [
|
20 |
+
InferInput("query", data_batch['query'].shape,
|
21 |
+
np_to_triton_dtype(data_batch['query'].dtype)),
|
22 |
+
]
|
23 |
+
|
24 |
+
inputs[0].set_data_from_numpy(data_batch['query'])
|
25 |
+
|
26 |
+
return inputs
|
27 |
+
|
28 |
+
|
29 |
+
def make_a_try(inputs, outputs='response', model_name='qwen', model_version='1'):
|
30 |
+
outputs_list = []
|
31 |
+
ori_outputs_list = outputs.strip().split(",")
|
32 |
+
for out_ele in ori_outputs_list:
|
33 |
+
outputs_list.append(out_ele.strip())
|
34 |
+
outputs = [InferRequestedOutput(x) for x in outputs_list]
|
35 |
+
|
36 |
+
response = my_client.predict(model_name=model_name, inputs=inputs, model_version=model_version, outputs=outputs)
|
37 |
+
|
38 |
+
rsp_info = {}
|
39 |
+
if outputs_list == []:
|
40 |
+
for out_name_ele in response._result.outputs:
|
41 |
+
outputs_list.append(out_name_ele.name)
|
42 |
+
for output_name in outputs_list:
|
43 |
+
res = response.as_numpy(output_name)
|
44 |
+
response = np.expand_dims(res, axis=0)
|
45 |
+
response = response[0].decode('utf-8')
|
46 |
+
rsp_info[output_name] = response
|
47 |
+
print("response:",rsp_info)
|
48 |
+
return rsp_info['response']
|
49 |
+
|
50 |
+
|
51 |
+
|
52 |
+
def greet(prompt):
|
53 |
+
"""Greet someone."""
|
54 |
+
# print(prompt)
|
55 |
+
print("prompt:",prompt)
|
56 |
+
inputs = prepare_params(prompt)
|
57 |
+
print(inputs)
|
58 |
+
result = make_a_try(inputs)
|
59 |
+
|
60 |
+
return result
|
61 |
+
|
62 |
+
def clear_input():
|
63 |
+
|
64 |
+
return ""
|
65 |
+
|
66 |
+
if __name__ == "__main__":
|
67 |
+
param_info = {}
|
68 |
+
# param_info['appkey'] = "com.sankuai.automl.serving"
|
69 |
+
param_info['appkey'] = "com.sankuai.automl.streamvlm"
|
70 |
+
|
71 |
+
param_info['remote_appkey'] = "com.sankuai.automl.chat3"
|
72 |
+
param_info['model_name'] = "qwen"
|
73 |
+
param_info['model_version'] = "1"
|
74 |
+
param_info['time_out'] = 60000
|
75 |
+
param_info['server_targets'] = []
|
76 |
+
param_info['outputs'] = 'response'
|
77 |
+
|
78 |
+
appkey, remote_appkey, model_name, model_version, time_out, server_targets = param_info['appkey'], param_info['remote_appkey'], param_info['model_name'], param_info['model_version'], param_info['time_out'], param_info['server_targets']
|
79 |
+
|
80 |
+
my_client = TritonServerClient(appkey=appkey, remote_appkey=remote_appkey, time_out=time_out, server_targets=server_targets)
|
81 |
+
|
82 |
+
# # 以上部分,用户可以��改代码
|
83 |
+
# demo = gr.Interface(
|
84 |
+
# fn=greet,
|
85 |
+
# inputs=["textbox"],
|
86 |
+
# outputs=["textbox"],
|
87 |
+
# )
|
88 |
+
with gr.Blocks(title='demo') as demo:
|
89 |
+
with gr.Row():
|
90 |
+
with gr.Column():
|
91 |
+
promptbox = gr.Textbox(label = "prompt")
|
92 |
+
|
93 |
+
with gr.Column():
|
94 |
+
output = gr.Textbox(label = "output")
|
95 |
+
with gr.Row():
|
96 |
+
submit = gr.Button("submit")
|
97 |
+
clear = gr.Button("clear")
|
98 |
+
|
99 |
+
submit.click(fn=greet,inputs=[promptbox],outputs=[output])
|
100 |
+
clear.click(fn=clear_input, inputs=[], outputs=[output])
|
101 |
+
|
102 |
+
# demo.launch(server_name="0.0.0.0", server_port=8088, debug=True, share=True)
|
103 |
+
# demo.launch(server_name="0.0.0.0", server_port=8080, debug=True, share=True)
|
104 |
+
|
105 |
+
#http://10.99.5.48:8080/
|
requirements.txt
ADDED
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
aiofiles==23.2.1
|
2 |
+
aiohttp==3.8.6
|
3 |
+
aiosignal==1.3.1
|
4 |
+
altair==5.0.1
|
5 |
+
annotated-types==0.5.0
|
6 |
+
anyio==3.7.1
|
7 |
+
async-timeout==4.0.3
|
8 |
+
asynctest==0.13.0
|
9 |
+
attrs==23.2.0
|
10 |
+
certifi @ file:///croot/certifi_1671487769961/work/certifi
|
11 |
+
charset-normalizer==3.3.2
|
12 |
+
click==8.1.7
|
13 |
+
cycler==0.11.0
|
14 |
+
exceptiongroup==1.2.1
|
15 |
+
fastapi==0.99.1
|
16 |
+
ffmpy==0.3.2
|
17 |
+
filelock==3.12.2
|
18 |
+
fonttools==4.38.0
|
19 |
+
frozenlist==1.3.3
|
20 |
+
fsspec==2023.1.0
|
21 |
+
gradio==3.34.0
|
22 |
+
gradio_client==0.2.6
|
23 |
+
grpcio==1.62.2
|
24 |
+
h11==0.14.0
|
25 |
+
httpcore==0.17.3
|
26 |
+
httpx==0.24.1
|
27 |
+
huggingface-hub==0.16.4
|
28 |
+
idna==3.7
|
29 |
+
importlib-metadata==6.7.0
|
30 |
+
importlib-resources==5.12.0
|
31 |
+
Jinja2==3.1.3
|
32 |
+
jsonschema==4.17.3
|
33 |
+
kiwisolver==1.4.5
|
34 |
+
linkify-it-py==2.0.3
|
35 |
+
markdown-it-py==2.2.0
|
36 |
+
MarkupSafe==2.1.5
|
37 |
+
matplotlib==3.5.3
|
38 |
+
mdit-py-plugins==0.3.3
|
39 |
+
mdurl==0.1.2
|
40 |
+
multidict==6.0.5
|
41 |
+
numpy==1.21.6
|
42 |
+
octo-rpc==0.4.7
|
43 |
+
orjson==3.9.7
|
44 |
+
packaging==24.0
|
45 |
+
pandas==1.3.5
|
46 |
+
Pillow==9.5.0
|
47 |
+
pkgutil_resolve_name==1.3.10
|
48 |
+
ply==3.11
|
49 |
+
protobuf==3.20.1
|
50 |
+
psutil==5.9.8
|
51 |
+
pydantic==1.10.11
|
52 |
+
pydantic_core==2.14.6
|
53 |
+
pydub==0.25.1
|
54 |
+
Pygments==2.17.2
|
55 |
+
pyparsing==3.1.2
|
56 |
+
pyrsistent==0.19.3
|
57 |
+
python-cat==0.0.11
|
58 |
+
python-dateutil==2.9.0.post0
|
59 |
+
python-multipart==0.0.8
|
60 |
+
pytz==2024.1
|
61 |
+
PyYAML==6.0.1
|
62 |
+
requests==2.31.0
|
63 |
+
semantic-version==2.10.0
|
64 |
+
six==1.16.0
|
65 |
+
sniffio==1.3.1
|
66 |
+
starlette==0.27.0
|
67 |
+
thrift==0.20.0
|
68 |
+
thriftpy2==0.4.20
|
69 |
+
toolz==0.12.1
|
70 |
+
tqdm==4.66.2
|
71 |
+
TritonServerClient==0.0.7
|
72 |
+
typing_extensions==4.7.1
|
73 |
+
uc-micro-py==1.0.3
|
74 |
+
urllib3==2.0.7
|
75 |
+
uvicorn==0.22.0
|
76 |
+
websockets==11.0.3
|
77 |
+
yarl==1.9.4
|
78 |
+
zipp==3.15.0
|
run.sh
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
source /workdir/yanghandi/gradio_demo/setup.sh
|
2 |
+
# source setup.sh
|
3 |
+
python main.py &> gradio.log &
|
4 |
+
|
5 |
+
python try_hd_v2.py &> multi_gradio.log &
|
6 |
+
|
7 |
+
python try_hd_v2.py &> try_hd_v2.log &
|
8 |
+
|
9 |
+
python test_hd.py &> test_hd.log &
|
setup.sh
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
conda deactivate
|
2 |
+
conda activate gradio
|
static/0000.jpeg
ADDED
static/0001.jpg
ADDED
static/image_1149d78e-43f6-4d5b-8e63-e24c2641012b.jpg
ADDED
static/image_11bc11b6-57a9-42d3-81fc-97d957e62f28.jpg
ADDED
static/image_33685dfc-b96f-401c-847d-3dd537186ec2.jpg
ADDED
static/image_4a883844-b1b2-4154-bbfe-0b108d93dca6.jpg
ADDED
static/image_4e43dfc4-9362-4e8d-b20f-d9816a9b684d.jpg
ADDED
static/image_51e314f6-0935-4430-a53f-bfdc22ca7cc4.jpg
ADDED
static/image_6c63b8a2-8bfc-485e-a238-94e76f22e7db.jpg
ADDED
static/image_82c514bf-bdc1-4d60-b073-a97b5816ea56.jpg
ADDED
static/image_845cd65a-30a7-43c3-be8c-819bc04de98d.jpg
ADDED
static/image_86804064-1f0e-4130-bcc3-30dd839a3c0a.jpg
ADDED
static/image_9afab29f-77d9-4451-8a52-092a5c37625d.jpg
ADDED
static/image_9c5c694a-acef-4545-8759-75ccffa39f6d.jpg
ADDED
static/image_b73c1efa-8112-404c-ab9e-eddb870f43af.jpg
ADDED
static/image_bbacb8a4-89d9-45ac-9da2-28717009e750.jpg
ADDED
static/image_bd087d82-9ed6-4b68-8ac1-c1417cfbb995.jpg
ADDED
static/image_ca9e5436-bc9b-4242-9a02-261d2209feb3.jpg
ADDED
static/image_dd018613-a2be-4eef-8183-a01929c835fd.jpg
ADDED
static/image_e63e5e5c-41cc-4f42-b896-95644d6b28f3.jpg
ADDED
static/image_e6c53092-7b49-4ba5-96dc-38e9479d3c6a.jpg
ADDED
static/image_f9bef0de-bc72-415d-b527-6c068594af02.jpg
ADDED
test_hd.py
ADDED
@@ -0,0 +1,187 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import argparse
|
2 |
+
import queue
|
3 |
+
import sys
|
4 |
+
import uuid
|
5 |
+
from functools import partial
|
6 |
+
|
7 |
+
import numpy as np
|
8 |
+
import tritonclient.grpc as grpcclient
|
9 |
+
from tritonclient.utils import InferenceServerException
|
10 |
+
import gradio as gr
|
11 |
+
from functools import wraps
|
12 |
+
|
13 |
+
####
|
14 |
+
from PIL import Image
|
15 |
+
import base64
|
16 |
+
import io
|
17 |
+
#####
|
18 |
+
from http.server import HTTPServer, SimpleHTTPRequestHandler
|
19 |
+
import socket
|
20 |
+
####
|
21 |
+
import os
|
22 |
+
import uuid
|
23 |
+
####
|
24 |
+
|
25 |
+
class UserData:
|
26 |
+
def __init__(self):
|
27 |
+
self._completed_requests = queue.Queue()
|
28 |
+
|
29 |
+
def callback(user_data, result, error):
|
30 |
+
if error:
|
31 |
+
user_data._completed_requests.put(error)
|
32 |
+
else:
|
33 |
+
user_data._completed_requests.put(result)
|
34 |
+
|
35 |
+
def make_a_try(img_url,text):
|
36 |
+
model_name = 'ensemble_mllm'
|
37 |
+
user_data = UserData()
|
38 |
+
sequence_id = 100
|
39 |
+
int_sequence_id0 = sequence_id
|
40 |
+
result_list=[]
|
41 |
+
try:
|
42 |
+
triton_client = grpcclient.InferenceServerClient(
|
43 |
+
url="10.95.163.43:8001",
|
44 |
+
# verbose=FLAGS.verbose,
|
45 |
+
verbose = True, #False
|
46 |
+
ssl=False,
|
47 |
+
root_certificates=None,
|
48 |
+
private_key=None,
|
49 |
+
certificate_chain=None,
|
50 |
+
)
|
51 |
+
except Exception as e:
|
52 |
+
print("channel creation failed: " + str(e))
|
53 |
+
return ""
|
54 |
+
# Infer
|
55 |
+
inputs = []
|
56 |
+
img_url_bytes = img_url.encode("utf-8")
|
57 |
+
img_url_bytes = np.array(img_url_bytes, dtype=bytes)
|
58 |
+
img_url_bytes = img_url_bytes.reshape([1, -1])
|
59 |
+
|
60 |
+
inputs.append(grpcclient.InferInput('IMAGE_URL', img_url_bytes.shape, "BYTES"))
|
61 |
+
inputs[0].set_data_from_numpy(img_url_bytes)
|
62 |
+
|
63 |
+
text_bytes = text.encode("utf-8")
|
64 |
+
text_bytes = np.array(text_bytes, dtype=bytes)
|
65 |
+
text_bytes = text_bytes.reshape([1, -1])
|
66 |
+
# text_input = np.expand_dims(text_bytes, axis=0)
|
67 |
+
text_input = text_bytes
|
68 |
+
|
69 |
+
inputs.append(grpcclient.InferInput('TEXT', text_input.shape, "BYTES"))
|
70 |
+
inputs[1].set_data_from_numpy(text_input)
|
71 |
+
|
72 |
+
outputs = []
|
73 |
+
outputs.append(grpcclient.InferRequestedOutput("OUTPUT"))
|
74 |
+
# Test with outputs
|
75 |
+
results = triton_client.infer(
|
76 |
+
model_name=model_name,
|
77 |
+
inputs=inputs,
|
78 |
+
outputs=outputs,
|
79 |
+
client_timeout=None, #FLAGS.client_timeout,
|
80 |
+
# headers={"test": "1"},
|
81 |
+
compression_algorithm=None, #FLAGS.grpc_compression_algorithm,
|
82 |
+
)
|
83 |
+
|
84 |
+
statistics = triton_client.get_inference_statistics(model_name=model_name)
|
85 |
+
print(statistics)
|
86 |
+
if len(statistics.model_stats) != 1:
|
87 |
+
print("FAILED: Inference Statistics")
|
88 |
+
return ""
|
89 |
+
|
90 |
+
# Get the output arrays from the results
|
91 |
+
output_data = results.as_numpy("OUTPUT")
|
92 |
+
result_str = output_data[0][0].decode('utf-8')
|
93 |
+
|
94 |
+
print("OUTPUT: "+ result_str)
|
95 |
+
return result_str
|
96 |
+
|
97 |
+
def greet(image, text):
|
98 |
+
###save img
|
99 |
+
static_path = f"/workdir/yanghandi/gradio_demo/static"
|
100 |
+
# 将图片转换为字节流
|
101 |
+
img_byte_arr = io.BytesIO()
|
102 |
+
try:
|
103 |
+
image.save(img_byte_arr, format='JPEG')
|
104 |
+
except Exception:
|
105 |
+
return ""
|
106 |
+
img_byte_arr = img_byte_arr.getvalue()
|
107 |
+
|
108 |
+
# 为图片生成一个唯一的文件名
|
109 |
+
# filename = "image_" + str(os.getpid()) + ".jpg" #uuid
|
110 |
+
unique_id = uuid.uuid4()
|
111 |
+
filename = f"image_{unique_id}.jpg"
|
112 |
+
filepath = os.path.join(static_path, filename)
|
113 |
+
|
114 |
+
# 将字节流写入文件
|
115 |
+
with open(filepath, 'wb') as f:
|
116 |
+
f.write(img_byte_arr)
|
117 |
+
|
118 |
+
|
119 |
+
img_url = f"http://10.99.5.48:8080/file=static/" + filename
|
120 |
+
# img_url = PIL_to_URL(img_url)
|
121 |
+
# img_url = "http://10.99.5.48:8080/file=static/0000.jpeg"
|
122 |
+
result = make_a_try(img_url,text)
|
123 |
+
# print(result)
|
124 |
+
return result
|
125 |
+
|
126 |
+
|
127 |
+
def clear_output():
|
128 |
+
|
129 |
+
return ""
|
130 |
+
|
131 |
+
def get_example():
|
132 |
+
return [
|
133 |
+
[f"/workdir/yanghandi/gradio_demo/static/0001.jpg", f"图中的人物是谁"]
|
134 |
+
]
|
135 |
+
if __name__ == "__main__":
|
136 |
+
|
137 |
+
param_info = {}
|
138 |
+
# param_info['appkey'] = "com.sankuai.automl.serving"
|
139 |
+
param_info['appkey'] = "10.199.14.151:8001"
|
140 |
+
|
141 |
+
# param_info['remote_appkey'] = "com.sankuai.automl.chat3"
|
142 |
+
param_info['remote_appkey'] = "10.199.14.151:8001"
|
143 |
+
param_info['model_name'] = 'ensemble_mllm'
|
144 |
+
param_info['model_version'] = "1"
|
145 |
+
param_info['time_out'] = 60000
|
146 |
+
param_info['server_targets'] = []
|
147 |
+
param_info['outputs'] = 'response'
|
148 |
+
|
149 |
+
|
150 |
+
gr.set_static_paths(paths=["static/"])
|
151 |
+
|
152 |
+
with gr.Blocks(title='demo') as demo:
|
153 |
+
gr.Markdown("# 自研模型测试demo")
|
154 |
+
gr.Markdown("尝试使用该demo,上传图片并开始讨论它,或者尝试下面的例子")
|
155 |
+
|
156 |
+
with gr.Row():
|
157 |
+
with gr.Column():
|
158 |
+
# imagebox = gr.Image(value="static/0000.jpeg",type="pil")
|
159 |
+
imagebox = gr.Image(type="pil")
|
160 |
+
promptbox = gr.Textbox(label = "prompt")
|
161 |
+
|
162 |
+
with gr.Column():
|
163 |
+
output = gr.Textbox(label = "output")
|
164 |
+
with gr.Row():
|
165 |
+
submit = gr.Button("submit")
|
166 |
+
clear = gr.Button("clear")
|
167 |
+
|
168 |
+
submit.click(fn=greet,inputs=[imagebox, promptbox],outputs=[output])
|
169 |
+
clear.click(fn=clear_output, inputs=[], outputs=[output])
|
170 |
+
|
171 |
+
gr.Markdown("# example")
|
172 |
+
|
173 |
+
gr.Examples(
|
174 |
+
examples = get_example(),
|
175 |
+
fn = greet,
|
176 |
+
inputs=[imagebox, promptbox],
|
177 |
+
outputs = [output],
|
178 |
+
cache_examples = True
|
179 |
+
)
|
180 |
+
|
181 |
+
demo.launch(server_name="0.0.0.0", server_port=8080, debug=True, share=True)
|
182 |
+
|
183 |
+
|
184 |
+
# img_url = f"https://s3plus.sankuai.com/automl-pkgs/0000.jpeg"
|
185 |
+
# # img_url = f"http://10.99.5.48:8080/file=static/static/image_cff7077b-3506-4253-82b7-b6547f2f63c1.jpg"
|
186 |
+
# text = f"talk about this women"
|
187 |
+
# greet(img_url,text)
|
try_demo.py
ADDED
@@ -0,0 +1,224 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python
|
2 |
+
# Copyright (c) 2020, NVIDIA CORPORATION. All rights reserved.
|
3 |
+
#
|
4 |
+
# Redistribution and use in source and binary forms, with or without
|
5 |
+
# modification, are permitted provided that the following conditions
|
6 |
+
# are met:
|
7 |
+
# * Redistributions of source code must retain the above copyright
|
8 |
+
# notice, this list of conditions and the following disclaimer.
|
9 |
+
# * Redistributions in binary form must reproduce the above copyright
|
10 |
+
# notice, this list of conditions and the following disclaimer in the
|
11 |
+
# documentation and/or other materials provided with the distribution.
|
12 |
+
# * Neither the name of NVIDIA CORPORATION nor the names of its
|
13 |
+
# contributors may be used to endorse or promote products derived
|
14 |
+
# from this software without specific prior written permission.
|
15 |
+
#
|
16 |
+
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
17 |
+
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
18 |
+
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
19 |
+
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
20 |
+
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
21 |
+
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
22 |
+
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
23 |
+
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
24 |
+
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
25 |
+
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
26 |
+
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
27 |
+
|
28 |
+
import argparse
|
29 |
+
import queue
|
30 |
+
import sys
|
31 |
+
import uuid
|
32 |
+
from functools import partial
|
33 |
+
|
34 |
+
import numpy as np
|
35 |
+
import tritonclient.grpc as grpcclient
|
36 |
+
from tritonclient.utils import InferenceServerException
|
37 |
+
|
38 |
+
##
|
39 |
+
import time
|
40 |
+
import threading
|
41 |
+
###
|
42 |
+
|
43 |
+
FLAGS = None
|
44 |
+
|
45 |
+
|
46 |
+
class UserData:
|
47 |
+
def __init__(self):
|
48 |
+
self._completed_requests = queue.Queue()
|
49 |
+
|
50 |
+
|
51 |
+
# Define the callback function. Note the last two parameters should be
|
52 |
+
# result and error. InferenceServerClient would povide the results of an
|
53 |
+
# inference as grpcclient.InferResult in result. For successful
|
54 |
+
# inference, error will be None, otherwise it will be an object of
|
55 |
+
# tritonclientutils.InferenceServerException holding the error details
|
56 |
+
def callback(user_data, result, error):
|
57 |
+
if error:
|
58 |
+
user_data._completed_requests.put(error)
|
59 |
+
else:
|
60 |
+
user_data._completed_requests.put(result)
|
61 |
+
|
62 |
+
|
63 |
+
if __name__ == "__main__":
|
64 |
+
parser = argparse.ArgumentParser()
|
65 |
+
parser.add_argument(
|
66 |
+
"-v",
|
67 |
+
"--verbose",
|
68 |
+
action="store_true",
|
69 |
+
required=False,
|
70 |
+
default=False,
|
71 |
+
help="Enable verbose output",
|
72 |
+
)
|
73 |
+
# parser.add_argument(
|
74 |
+
# "-u",
|
75 |
+
# "--url",
|
76 |
+
# type=str,
|
77 |
+
# required=False,
|
78 |
+
# default="localhost:8001",
|
79 |
+
# help="Inference server URL and it gRPC port. Default is localhost:8001.",
|
80 |
+
# )
|
81 |
+
parser.add_argument(
|
82 |
+
"-u",
|
83 |
+
"--url",
|
84 |
+
type=str,
|
85 |
+
required=False,
|
86 |
+
default="10.199.14.151:8001",
|
87 |
+
help="Inference server URL and it gRPC port. Default is localhost:8001.",
|
88 |
+
)
|
89 |
+
parser.add_argument(
|
90 |
+
"-t",
|
91 |
+
"--stream-timeout",
|
92 |
+
type=float,
|
93 |
+
required=False,
|
94 |
+
default=None,
|
95 |
+
help="Stream timeout in seconds. Default is None.",
|
96 |
+
)
|
97 |
+
# parser.add_argument(
|
98 |
+
# "-d",
|
99 |
+
# "--dyna",
|
100 |
+
# action="store_true",
|
101 |
+
# required=False,
|
102 |
+
# default=False,
|
103 |
+
# help="Assume dynamic sequence model",
|
104 |
+
# )
|
105 |
+
# parser.add_argument(
|
106 |
+
# "-o",
|
107 |
+
# "--offset",
|
108 |
+
# type=int,
|
109 |
+
# required=False,
|
110 |
+
# default=0,
|
111 |
+
# help="Add offset to sequence ID used",
|
112 |
+
# )
|
113 |
+
|
114 |
+
FLAGS = parser.parse_args()
|
115 |
+
|
116 |
+
# # We use custom "sequence" models which take 1 input
|
117 |
+
# # value. The output is the accumulated value of the inputs. See
|
118 |
+
# # src/custom/sequence.
|
119 |
+
# int_sequence_model_name = (
|
120 |
+
# "simple_dyna_sequence" if FLAGS.dyna else "simple_sequence"
|
121 |
+
# )
|
122 |
+
# string_sequence_model_name = (
|
123 |
+
# "simple_string_dyna_sequence" if FLAGS.dyna else "simple_sequence"
|
124 |
+
# )
|
125 |
+
model_name = 'ensemble_mllm'
|
126 |
+
model_version = ""
|
127 |
+
batch_size = 1
|
128 |
+
# img_url = f"https://s3plus.sankuai.com/automl-pkgs/0000.jpeg"
|
129 |
+
img_url = "/workdir/yanghandi/gradio_demo/static/0000.jpeg"
|
130 |
+
# img_url = f"https://s3plus.sankuai.com/automl-pkgs/0003.jpeg"
|
131 |
+
text = f"详细描述一下这张图片"
|
132 |
+
sequence_id = 100
|
133 |
+
int_sequence_id0 = sequence_id
|
134 |
+
|
135 |
+
result_list = []
|
136 |
+
user_data = UserData()
|
137 |
+
|
138 |
+
# It is advisable to use client object within with..as clause
|
139 |
+
# when sending streaming requests. This ensures the client
|
140 |
+
# is closed when the block inside with exits.
|
141 |
+
with grpcclient.InferenceServerClient(
|
142 |
+
url=FLAGS.url, verbose=FLAGS.verbose
|
143 |
+
) as triton_client:
|
144 |
+
try:
|
145 |
+
# Establish stream
|
146 |
+
triton_client.start_stream(
|
147 |
+
callback=partial(callback, user_data),
|
148 |
+
stream_timeout=FLAGS.stream_timeout,
|
149 |
+
)
|
150 |
+
|
151 |
+
# Create the tensor for INPUT
|
152 |
+
inputs = []
|
153 |
+
img_url_bytes = img_url.encode("utf-8")
|
154 |
+
img_url_bytes = np.array(img_url_bytes, dtype=bytes)
|
155 |
+
img_url_bytes = img_url_bytes.reshape([1, -1])
|
156 |
+
|
157 |
+
inputs.append(grpcclient.InferInput('IMAGE_URL', img_url_bytes.shape, "BYTES"))
|
158 |
+
inputs[0].set_data_from_numpy(img_url_bytes)
|
159 |
+
|
160 |
+
text_bytes = text.encode("utf-8")
|
161 |
+
text_bytes = np.array(text_bytes, dtype=bytes)
|
162 |
+
text_bytes = text_bytes.reshape([1, -1])
|
163 |
+
# text_input = np.expand_dims(text_bytes, axis=0)
|
164 |
+
text_input = text_bytes
|
165 |
+
|
166 |
+
inputs.append(grpcclient.InferInput('TEXT', text_input.shape, "BYTES"))
|
167 |
+
inputs[1].set_data_from_numpy(text_input)
|
168 |
+
|
169 |
+
outputs = []
|
170 |
+
outputs.append(grpcclient.InferRequestedOutput("OUTPUT"))
|
171 |
+
# Issue the asynchronous sequence inference.
|
172 |
+
triton_client.async_stream_infer(
|
173 |
+
model_name=model_name,
|
174 |
+
inputs=inputs,
|
175 |
+
outputs=outputs,
|
176 |
+
request_id="{}".format(sequence_id),
|
177 |
+
sequence_id=sequence_id,
|
178 |
+
sequence_start=True,
|
179 |
+
sequence_end=True,
|
180 |
+
)
|
181 |
+
|
182 |
+
except InferenceServerException as error:
|
183 |
+
print(error)
|
184 |
+
sys.exit(1)
|
185 |
+
|
186 |
+
# Retrieve results...
|
187 |
+
recv_count = 0
|
188 |
+
|
189 |
+
#####
|
190 |
+
|
191 |
+
####
|
192 |
+
|
193 |
+
while True:
|
194 |
+
|
195 |
+
# if len(result_list) == 80:
|
196 |
+
# print("1")
|
197 |
+
data_item = user_data._completed_requests.get()
|
198 |
+
# try:
|
199 |
+
# data_item = user_data._completed_requests.get(timeout=5)
|
200 |
+
# except Exception as e:
|
201 |
+
# print("queue wrong")
|
202 |
+
# break
|
203 |
+
if type(data_item) == InferenceServerException:
|
204 |
+
print('InferenceServerException: ', data_item)
|
205 |
+
sys.exit(1)
|
206 |
+
this_id = data_item.get_response().id.split("_")[0]
|
207 |
+
if int(this_id) != int_sequence_id0:
|
208 |
+
print("unexpected sequence id returned by the server: {}".format(this_id))
|
209 |
+
sys.exit(1)
|
210 |
+
|
211 |
+
result = data_item.as_numpy("OUTPUT")
|
212 |
+
if len(result[0][0])==0:
|
213 |
+
break
|
214 |
+
|
215 |
+
result_list.append(data_item.as_numpy("OUTPUT"))
|
216 |
+
|
217 |
+
recv_count = recv_count + 1
|
218 |
+
result_str = ''.join([item[0][0].decode('utf-8') for item in result_list])
|
219 |
+
print(f"{len(result_list)}: {result_str}")
|
220 |
+
|
221 |
+
|
222 |
+
print("hd",result_str)
|
223 |
+
print("PASS: Sequence")
|
224 |
+
print("hd",result_str)
|
try_demo_demo.py
ADDED
@@ -0,0 +1,206 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python
|
2 |
+
# Copyright (c) 2020, NVIDIA CORPORATION. All rights reserved.
|
3 |
+
#
|
4 |
+
# Redistribution and use in source and binary forms, with or without
|
5 |
+
# modification, are permitted provided that the following conditions
|
6 |
+
# are met:
|
7 |
+
# * Redistributions of source code must retain the above copyright
|
8 |
+
# notice, this list of conditions and the following disclaimer.
|
9 |
+
# * Redistributions in binary form must reproduce the above copyright
|
10 |
+
# notice, this list of conditions and the following disclaimer in the
|
11 |
+
# documentation and/or other materials provided with the distribution.
|
12 |
+
# * Neither the name of NVIDIA CORPORATION nor the names of its
|
13 |
+
# contributors may be used to endorse or promote products derived
|
14 |
+
# from this software without specific prior written permission.
|
15 |
+
#
|
16 |
+
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
17 |
+
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
18 |
+
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
19 |
+
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
20 |
+
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
21 |
+
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
22 |
+
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
23 |
+
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
24 |
+
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
25 |
+
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
26 |
+
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
27 |
+
|
28 |
+
import argparse
|
29 |
+
import sys
|
30 |
+
|
31 |
+
import numpy as np
|
32 |
+
import tritonclient.grpc as grpcclient
|
33 |
+
|
34 |
+
if __name__ == "__main__":
|
35 |
+
parser = argparse.ArgumentParser()
|
36 |
+
parser.add_argument(
|
37 |
+
"-v",
|
38 |
+
"--verbose",
|
39 |
+
action="store_true",
|
40 |
+
required=False,
|
41 |
+
default=False,
|
42 |
+
help="Enable verbose output",
|
43 |
+
)
|
44 |
+
parser.add_argument(
|
45 |
+
"-u",
|
46 |
+
"--url",
|
47 |
+
type=str,
|
48 |
+
required=False,
|
49 |
+
default="10.95.163.43:8001",
|
50 |
+
help="Inference server URL. Default is localhost:8001.",
|
51 |
+
)
|
52 |
+
parser.add_argument(
|
53 |
+
"-s",
|
54 |
+
"--ssl",
|
55 |
+
action="store_true",
|
56 |
+
required=False,
|
57 |
+
default=False,
|
58 |
+
help="Enable SSL encrypted channel to the server",
|
59 |
+
)
|
60 |
+
parser.add_argument(
|
61 |
+
"-t",
|
62 |
+
"--client-timeout",
|
63 |
+
type=float,
|
64 |
+
required=False,
|
65 |
+
default=None,
|
66 |
+
help="Client timeout in seconds. Default is None.",
|
67 |
+
)
|
68 |
+
parser.add_argument(
|
69 |
+
"-r",
|
70 |
+
"--root-certificates",
|
71 |
+
type=str,
|
72 |
+
required=False,
|
73 |
+
default=None,
|
74 |
+
help="File holding PEM-encoded root certificates. Default is None.",
|
75 |
+
)
|
76 |
+
parser.add_argument(
|
77 |
+
"-p",
|
78 |
+
"--private-key",
|
79 |
+
type=str,
|
80 |
+
required=False,
|
81 |
+
default=None,
|
82 |
+
help="File holding PEM-encoded private key. Default is None.",
|
83 |
+
)
|
84 |
+
parser.add_argument(
|
85 |
+
"-x",
|
86 |
+
"--certificate-chain",
|
87 |
+
type=str,
|
88 |
+
required=False,
|
89 |
+
default=None,
|
90 |
+
help="File holding PEM-encoded certificate chain. Default is None.",
|
91 |
+
)
|
92 |
+
parser.add_argument(
|
93 |
+
"-C",
|
94 |
+
"--grpc-compression-algorithm",
|
95 |
+
type=str,
|
96 |
+
required=False,
|
97 |
+
default=None,
|
98 |
+
help="The compression algorithm to be used when sending request to server. Default is None.",
|
99 |
+
)
|
100 |
+
|
101 |
+
FLAGS = parser.parse_args()
|
102 |
+
try:
|
103 |
+
# triton_client = grpcclient.InferenceServerClient(
|
104 |
+
# url=FLAGS.url,
|
105 |
+
# verbose=FLAGS.verbose,
|
106 |
+
# ssl=FLAGS.ssl,
|
107 |
+
# root_certificates=FLAGS.root_certificates,
|
108 |
+
# private_key=FLAGS.private_key,
|
109 |
+
# certificate_chain=FLAGS.certificate_chain,
|
110 |
+
# )
|
111 |
+
triton_client = grpcclient.InferenceServerClient(
|
112 |
+
url=FLAGS.url,
|
113 |
+
# verbose=FLAGS.verbose,
|
114 |
+
verbose = True,
|
115 |
+
ssl=FLAGS.ssl,
|
116 |
+
root_certificates=None,
|
117 |
+
private_key=None,
|
118 |
+
certificate_chain=None,
|
119 |
+
)
|
120 |
+
except Exception as e:
|
121 |
+
print("channel creation failed: " + str(e))
|
122 |
+
sys.exit()
|
123 |
+
|
124 |
+
model_name = "ensemble_mllm"
|
125 |
+
img_url = f"https://s3plus.sankuai.com/automl-pkgs/0000.jpeg"
|
126 |
+
# img_url = f"https://s3plus.sankuai.com/automl-pkgs/0003.jpeg"
|
127 |
+
text = f"详细描述一下这张图片"
|
128 |
+
|
129 |
+
# Infer
|
130 |
+
inputs = []
|
131 |
+
img_url_bytes = img_url.encode("utf-8")
|
132 |
+
img_url_bytes = np.array(img_url_bytes, dtype=bytes)
|
133 |
+
img_url_bytes = img_url_bytes.reshape([1, -1])
|
134 |
+
|
135 |
+
inputs.append(grpcclient.InferInput('IMAGE_URL', img_url_bytes.shape, "BYTES"))
|
136 |
+
inputs[0].set_data_from_numpy(img_url_bytes)
|
137 |
+
|
138 |
+
text_bytes = text.encode("utf-8")
|
139 |
+
text_bytes = np.array(text_bytes, dtype=bytes)
|
140 |
+
text_bytes = text_bytes.reshape([1, -1])
|
141 |
+
# text_input = np.expand_dims(text_bytes, axis=0)
|
142 |
+
text_input = text_bytes
|
143 |
+
|
144 |
+
inputs.append(grpcclient.InferInput('TEXT', text_input.shape, "BYTES"))
|
145 |
+
inputs[1].set_data_from_numpy(text_input)
|
146 |
+
|
147 |
+
outputs = []
|
148 |
+
outputs.append(grpcclient.InferRequestedOutput("OUTPUT"))
|
149 |
+
|
150 |
+
# Test with outputs
|
151 |
+
results = triton_client.infer(
|
152 |
+
model_name=model_name,
|
153 |
+
inputs=inputs,
|
154 |
+
outputs=outputs,
|
155 |
+
client_timeout=None, #FLAGS.client_timeout,
|
156 |
+
# headers={"test": "1"},
|
157 |
+
compression_algorithm=None, #FLAGS.grpc_compression_algorithm,
|
158 |
+
)
|
159 |
+
|
160 |
+
statistics = triton_client.get_inference_statistics(model_name=model_name)
|
161 |
+
print(statistics)
|
162 |
+
if len(statistics.model_stats) != 1:
|
163 |
+
print("FAILED: Inference Statistics")
|
164 |
+
sys.exit(1)
|
165 |
+
|
166 |
+
# Get the output arrays from the results
|
167 |
+
output_data = results.as_numpy("OUTPUT")
|
168 |
+
result_str = output_data[0][0].decode('utf-8')
|
169 |
+
|
170 |
+
print("OUTPUT: "+ result_str)
|
171 |
+
|
172 |
+
# # Test with no outputs
|
173 |
+
# results = triton_client.infer(
|
174 |
+
# model_name=model_name,
|
175 |
+
# inputs=inputs,
|
176 |
+
# outputs=None,
|
177 |
+
# compression_algorithm=FLAGS.grpc_compression_algorithm,
|
178 |
+
# )
|
179 |
+
|
180 |
+
# # Get the output arrays from the results
|
181 |
+
# output0_data = results.as_numpy("OUTPUT0")
|
182 |
+
# output1_data = results.as_numpy("OUTPUT1")
|
183 |
+
|
184 |
+
# for i in range(16):
|
185 |
+
# print(
|
186 |
+
# str(input0_data[0][i])
|
187 |
+
# + " + "
|
188 |
+
# + str(input1_data[0][i])
|
189 |
+
# + " = "
|
190 |
+
# + str(output0_data[0][i])
|
191 |
+
# )
|
192 |
+
# print(
|
193 |
+
# str(input0_data[0][i])
|
194 |
+
# + " - "
|
195 |
+
# + str(input1_data[0][i])
|
196 |
+
# + " = "
|
197 |
+
# + str(output1_data[0][i])
|
198 |
+
# )
|
199 |
+
# if (input0_data[0][i] + input1_data[0][i]) != output0_data[0][i]:
|
200 |
+
# print("sync infer error: incorrect sum")
|
201 |
+
# sys.exit(1)
|
202 |
+
# if (input0_data[0][i] - input1_data[0][i]) != output1_data[0][i]:
|
203 |
+
# print("sync infer error: incorrect difference")
|
204 |
+
# sys.exit(1)
|
205 |
+
|
206 |
+
print("PASS: infer")
|
try_grpc.py
ADDED
@@ -0,0 +1,139 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
# from TritonServerClient import TritonServerClient, InferInput, InferRequestedOutput
|
3 |
+
# from TritonServerClient.utils import np_to_triton_dtype
|
4 |
+
from functools import wraps
|
5 |
+
import numpy as np
|
6 |
+
|
7 |
+
import tritonclient.grpc as grpcclient
|
8 |
+
from tritonclient.utils import InferenceServerException
|
9 |
+
|
10 |
+
# def prepare_params(query):
|
11 |
+
# bs=1
|
12 |
+
# # query = "#角色:直播间文案写作专家。#任务:根据用户输入的商品信息写出用于推广营销的直播文案。#目标:引导用户购买直播间商品,提高销量。#要求:- 如果没有输入商品信息,则邀请输入。- 文案写作 - 按照“引出话题”、“产品讲解”、“购买方式”三个方面写作文案。 - 在“产品讲解”部分详细讲解产品细节: - 【重要】充分扩充内容,吸引用户购买。 - 详细描述商品的优惠活动信息(如用户输入没有,不要询问,直接补充)。 - 在“购买方式”部分: - 仅包括引导用户点击链接购买的内容。 - 不要包括快递物流承诺。 - 【重要】每个部分多于5句话。 - 使用口语化的文字。 - 表达方式有亲和力和感染力。 - 对用户们使用“宝子们”、“宝宝们”、“家人们”等有亲和力的称呼。 - 多使用自问自答的语句。- 当你准备好了回复“请输入商品描述和优惠信息”我们就开始。#输出:- 包含在""里。- 写出每一部分内容的标题,包括【引出话题】、【产品讲解】、【购买方式】。- 示例:""【引出话题】【产品讲解】【购买方式】""- 输入商品信息:%s" % prompt
|
13 |
+
# # query = '请你扮演直播间直播卖货的主播角色,根据用户提出的问题,检索知识库中的信息,如果查找到相关的信息,使用口语化的文字,用非常简短的一句话回答用户的问题;如果查找不到相关的信息,直接回答没有,并引导用户关注本场直播商品。你的目的是激发用户的购买欲、提高直播商品销量。回答对用户们使用“宝子”、“宝宝”、“亲”等有亲和力的称呼。\n\n# 用户提出的问题:\n这里面有可乐吗\n\n# 知识库:\n商品名称:外送欢乐桶(券)兑换券\n直播链接序号:64号链接号链接\n商品现价:66元\n商品原价:142元\n折扣:5\n品牌名称:麦当劳麦乐送\n券包张数:1\n套餐详情:["麦辣鸡腿汉堡1个","板烧鸡腿堡1个","中可乐3杯","四拼小食桶A"]\n补充:四拼小食桶A(麦麦脆汁鸡(鸡腿)1块+中薯条1份+麦辣鸡翅2块+麦乐鸡5块)+麦辣鸡腿堡1个+板烧鸡腿堡1个+中可乐3杯 原材料:面包\n\n\n# 输出格式:\n答:\n\n# 要求\n对于用户问到了知识库中未提及的信息不要编造,直接不回答'
|
14 |
+
# ######
|
15 |
+
# #####
|
16 |
+
# title_text = np.array([query.encode('utf-8')], dtype=np.string_)
|
17 |
+
# title_text = np.tile(title_text, (bs, 1))
|
18 |
+
|
19 |
+
# data_batch = {}
|
20 |
+
# data_batch['query'] = title_text
|
21 |
+
|
22 |
+
# inputs = [
|
23 |
+
# InferInput("query", data_batch['query'].shape,
|
24 |
+
# np_to_triton_dtype(data_batch['query'].dtype)),
|
25 |
+
# ]
|
26 |
+
|
27 |
+
# inputs[0].set_data_from_numpy(data_batch['query'])
|
28 |
+
|
29 |
+
# return inputs
|
30 |
+
def prepare_params(query,img):
|
31 |
+
bs=1
|
32 |
+
# query = "#角色:直播间文案写作专家。#任务:根据用户输入的商品信息写出用于推广营销的直播文案。#目标:引导用户购买直播间商品,提高销量。#要求:- 如果没有输入商品信息,则邀请输入。- 文案写作 - 按照“引出话题”、“产品讲解”、“购买方式”三个方面写作文案。 - 在“产品讲解”部分详细讲解产品细节: - 【重要】充分扩充内容,吸引用户购买。 - 详细描述商品的优惠活动信息(如用户输入没有,不要询问,直接补充)。 - 在“购买方式”部分: - 仅包括引导用户点击链接购买的内容。 - 不要包括快递物流承诺。 - 【重要】每个部分多于5句话。 - 使用口语化的文字。 - 表达方式有亲和力和感染力。 - 对用户们使用“宝子们”、“宝宝们”、“家人们”等有亲和力的称呼。 - 多使用自问自答的语句。- 当你准备好了回复“请输入商品描述和优惠信息”我们就开始。#输出:- 包含在""里。- 写出每一部分内容的标题,包括【引出话题】、【产品讲解】、【购买方式】。- 示例:""【引出话题】【产品讲解】【购买方式】""- 输入商品信息:%s" % prompt
|
33 |
+
# query = '请你扮演直播间直播卖货的主播角色,根据用户提出的问题,检索知识库中的信息,如果查找到相关的信息,使用口语化的文字,用非常简短的一句话回答用户的问题;如果查找不到相关的信息,直接回答没有,并引导用户关注本场直播商品。你的目的是激发用户的购买欲、提高直播商品销量。回答对用户们使用“宝子”、“宝宝”、“亲��等有亲和力的称呼。\n\n# 用户提出的问题:\n这里面有可乐吗\n\n# 知识库:\n商品名称:外送欢乐桶(券)兑换券\n直播链接序号:64号链接号链接\n商品现价:66元\n商品原价:142元\n折扣:5\n品牌名称:麦当劳麦乐送\n券包张数:1\n套餐详情:["麦辣鸡腿汉堡1个","板烧鸡腿堡1个","中可乐3杯","四拼小食桶A"]\n补充:四拼小食桶A(麦麦脆汁鸡(鸡腿)1块+中薯条1份+麦辣鸡翅2块+麦乐鸡5块)+麦辣鸡腿堡1个+板烧鸡腿堡1个+中可乐3杯 原材料:面包\n\n\n# 输出格式:\n答:\n\n# 要求\n对于用户问到了知识库中未提及的信息不要编造,直接不回答'
|
34 |
+
inputs = []
|
35 |
+
######
|
36 |
+
img_info = np.array([img.encode('utf-8')], dtype=np.string_)
|
37 |
+
img_info = np.tile(img_info, (bs,1))
|
38 |
+
|
39 |
+
|
40 |
+
inputs.append(InferInput("img", img_info.shape,
|
41 |
+
np_to_triton_dtype(img_info.dtype)),)
|
42 |
+
|
43 |
+
|
44 |
+
inputs[0].set_data_from_numpy(img_info)
|
45 |
+
#####
|
46 |
+
title_text = np.array([query.encode('utf-8')], dtype=np.string_)
|
47 |
+
title_text = np.tile(title_text, (bs, 1))
|
48 |
+
|
49 |
+
data_batch = {}
|
50 |
+
data_batch['query'] = title_text
|
51 |
+
|
52 |
+
inputs.append( InferInput("query", data_batch['query'].shape,
|
53 |
+
np_to_triton_dtype(data_batch['query'].dtype))
|
54 |
+
)
|
55 |
+
|
56 |
+
|
57 |
+
inputs[1].set_data_from_numpy(data_batch['query'])
|
58 |
+
|
59 |
+
return inputs
|
60 |
+
|
61 |
+
def make_a_try(inputs, outputs='response', model_name='ensemble_mllm', model_version=''): # qwen 1
|
62 |
+
outputs_list = []
|
63 |
+
ori_outputs_list = outputs.strip().split(",")
|
64 |
+
for out_ele in ori_outputs_list:
|
65 |
+
outputs_list.append(out_ele.strip())
|
66 |
+
outputs = [InferRequestedOutput(x) for x in outputs_list]
|
67 |
+
|
68 |
+
response = my_client.predict(model_name=model_name, inputs=inputs, model_version=model_version, outputs=outputs)
|
69 |
+
|
70 |
+
rsp_info = {}
|
71 |
+
if outputs_list == []:
|
72 |
+
for out_name_ele in response._result.outputs:
|
73 |
+
outputs_list.append(out_name_ele.name)
|
74 |
+
for output_name in outputs_list:
|
75 |
+
res = response.as_numpy(output_name)
|
76 |
+
response = np.expand_dims(res, axis=0)
|
77 |
+
response = response[0].decode('utf-8')
|
78 |
+
rsp_info[output_name] = response
|
79 |
+
print("response:",rsp_info)
|
80 |
+
return rsp_info['response']
|
81 |
+
|
82 |
+
|
83 |
+
|
84 |
+
# def greet(prompt):
|
85 |
+
# """Greet someone."""
|
86 |
+
# # print(prompt)
|
87 |
+
# print("prompt:",prompt)
|
88 |
+
# inputs = prepare_params(prompt)
|
89 |
+
# print(inputs)
|
90 |
+
# result = make_a_try(inputs)
|
91 |
+
|
92 |
+
# return result
|
93 |
+
def greet(prompt,img):
|
94 |
+
"""Greet someone."""
|
95 |
+
# print(prompt)
|
96 |
+
print("prompt:",prompt)
|
97 |
+
inputs = prepare_params(prompt,img)
|
98 |
+
print(inputs)
|
99 |
+
result = make_a_try(inputs)
|
100 |
+
|
101 |
+
return result
|
102 |
+
|
103 |
+
def clear_input():
|
104 |
+
|
105 |
+
return ""
|
106 |
+
|
107 |
+
|
108 |
+
if __name__ == "__main__":
|
109 |
+
param_info = {}
|
110 |
+
# param_info['appkey'] = "com.sankuai.automl.serving"
|
111 |
+
param_info['appkey'] = "com.sankuai.automl.streamvlm"
|
112 |
+
|
113 |
+
# param_info['remote_appkey'] = "com.sankuai.automl.chat3"
|
114 |
+
param_info['remote_appkey'] = "com.sankuai.automl.streamvlm"
|
115 |
+
|
116 |
+
param_info['model_name'] = "ensemble_mllm"
|
117 |
+
param_info['model_version'] = "1"
|
118 |
+
param_info['time_out'] = 60000
|
119 |
+
param_info['server_targets'] = []
|
120 |
+
param_info['outputs'] = 'response'
|
121 |
+
|
122 |
+
appkey, remote_appkey, model_name, model_version, time_out, server_targets = param_info['appkey'], param_info['remote_appkey'], param_info['model_name'], param_info['model_version'], param_info['time_out'], param_info['server_targets']
|
123 |
+
|
124 |
+
my_client = TritonServerClient(appkey=appkey, remote_appkey=remote_appkey, time_out=time_out, server_targets=server_targets)
|
125 |
+
# triton_client.async_stream_infer(
|
126 |
+
# model_name=model_name,
|
127 |
+
# inputs=inputs,
|
128 |
+
# outputs=outputs,
|
129 |
+
# request_id="{}".format(sequence_id),
|
130 |
+
# sequence_id=sequence_id,
|
131 |
+
# sequence_start=True,
|
132 |
+
# sequence_end=True,
|
133 |
+
# )
|
134 |
+
|
135 |
+
img_url = f"https://s3plus.sankuai.com/automl-pkgs/0000.jpeg"
|
136 |
+
|
137 |
+
greet("nihao",img_url)
|
138 |
+
# greet("nihao")
|
139 |
+
print("描述这张图片")
|
try_hd.py
ADDED
@@ -0,0 +1,137 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from TritonServerClient import TritonServerClient, InferInput, InferRequestedOutput
|
3 |
+
from TritonServerClient.utils import np_to_triton_dtype
|
4 |
+
from functools import wraps
|
5 |
+
import numpy as np
|
6 |
+
|
7 |
+
|
8 |
+
# def prepare_params(query):
|
9 |
+
# bs=1
|
10 |
+
# # query = "#角色:直播间文案写作专家。#任务:根据用户输入的商品信息写出用于推广营销的直播文案。#目标:引导用户购买直播间商品,提高销量。#要求:- 如果没有输入商品信息,则邀请输入。- 文案写作 - 按照“引出话题”、“产品讲解”、“购买方式”三个方面写作文案。 - 在“产品讲解”部分详细讲解产品细节: - 【重要】充分扩充内容,吸引用户购买。 - 详细描述商品的优惠活动信息(如用户输入没有,不要询问,直接补充)。 - 在“购买方式”部分: - 仅包括引导用户点击链接购买的内容。 - 不要包括快递物流承诺。 - 【重要】每个部分多于5句话。 - 使用口语化的文字。 - 表达方式有亲和力和感染力。 - 对用户们使用“宝子们”、“宝宝们”、“家人们”等有亲和力的称呼。 - 多使用自问自答的语句。- 当你准备好了回复“请输入商品描述和优惠信息”我们就开始。#输出:- 包含在""里。- 写出每一部分内容的标题,包括【引出话题】、【产品讲解】、【购买方式】。- 示例:""【引出话题】【产品讲解】【购买方式】""- 输入商品信息:%s" % prompt
|
11 |
+
# # query = '请你扮演直播间直播卖货的主播角色,根据用户提出的问题,检索知识库中的信息,如果查找到相关的信息,使用口语化的文字,用非常简短的一句话回答用户的问题;如果查找不到相关的信息,直接回答没有,并引导用户关注本场直播商品。你的目的是激发用户的购买欲、提高直播商品销量。回答对用户们使用“宝子”、“宝宝”、“亲”等有亲和力的称呼。\n\n# 用户提出的问题:\n这里面有可乐吗\n\n# 知识库:\n商品名称:外送欢乐桶(券)兑换券\n直播链接序号:64号链接号链接\n商品现价:66元\n商品原价:142元\n折扣:5\n品牌名称:麦当劳麦乐送\n券包张数:1\n套餐详情:["麦辣鸡腿汉堡1个","板烧鸡腿堡1个","中可乐3杯","四拼小食桶A"]\n补充:四拼小食桶A(麦麦脆汁鸡(鸡腿)1块+中薯条1份+麦辣鸡翅2块+麦乐鸡5块)+麦辣鸡腿堡1个+板烧鸡腿堡1个+中可乐3杯 原材料:面包\n\n\n# 输出格式:\n答:\n\n# 要求\n对于用户问到了知识库中未提及的信息不要编造,直接不回答'
|
12 |
+
# ######
|
13 |
+
# #####
|
14 |
+
# title_text = np.array([query.encode('utf-8')], dtype=np.string_)
|
15 |
+
# title_text = np.tile(title_text, (bs, 1))
|
16 |
+
|
17 |
+
# data_batch = {}
|
18 |
+
# data_batch['query'] = title_text
|
19 |
+
|
20 |
+
# inputs = [
|
21 |
+
# InferInput("query", data_batch['query'].shape,
|
22 |
+
# np_to_triton_dtype(data_batch['query'].dtype)),
|
23 |
+
# ]
|
24 |
+
|
25 |
+
# inputs[0].set_data_from_numpy(data_batch['query'])
|
26 |
+
|
27 |
+
# return inputs
|
28 |
+
def prepare_params(query,img):
|
29 |
+
bs=1
|
30 |
+
# query = "#角色:直播间文案写作专家。#任务:根据用户输入的商品信息写出用于推广营销的直播文案。#目标:引导用户购买直播间商品,提高销量。#要求:- 如果没有输入商品信息,则邀请输入。- 文案写作 - 按照“引出话题”、“产品讲解”、“购买方式”三个方面写作文案。 - 在“产品讲解”部分详细讲解产品细节: - 【重要】充分扩充内容,吸引用户购买。 - 详细描述商品的优惠活动信息(如用户输入没有,不要询问,直接补充)。 - 在“购买方式”部分: - 仅包括引导用户点击链接购买的内容。 - 不要包括快递物流承诺。 - 【重要】每个部分多于5句话。 - 使用口语化的文字。 - 表达方式有亲和力和感染力。 - 对用户们使用“宝子们”、“宝宝们”、“家人们”等有亲和力的称呼。 - 多使用自问自答的语句。- 当你准备好了回复“请输入商品描述和优惠信息”我们就开始。#输出:- 包含在""里。- 写出每一部分内容的标题,包括【引出话题】、【产品讲解】、【购买方式】。- 示例:""【引出话题】【产品讲解】【购买方式】""- 输入商品信息:%s" % prompt
|
31 |
+
# query = '请你扮演直播间直播卖货的主播角色,根据用户提出的问题,检索知识库中的信息,如果查找到相关的信息,使用口语化的文字,用非常简短的一句话回答用户的问题;如果查找不到相关的信息,直接回答没有,并引导用户关注本场直播商品。你的目的是激发用户的购买欲、提高直播商品销量。回答对用户们使用“宝子”、“宝宝”、“亲”等有亲和力的称呼。\n\n# 用户提出的问题:\n这里面有可乐吗\n\n# 知识库:\n商品名称:外送欢乐桶(券)兑换券\n直播链接序号:64号链接号链接\n商品现价:66元\n商品原价:142元\n折扣:5\n品牌名称:麦当劳麦乐送\n券包张数:1\n套餐详情:["麦辣鸡腿汉堡1个","板烧鸡腿堡1个","中可乐3杯","四拼小食桶A"]\n补充:四拼小食桶A(麦麦脆汁鸡(鸡腿)1块+中薯条1份+麦辣鸡翅2块+麦乐鸡5块)+麦辣鸡腿堡1个+板烧鸡腿堡1个+中可乐3杯 原材料:面包\n\n\n# 输出格式:\n答:\n\n# 要求\n对于用户问到了知识库中未提及的信息不要编造,直接不回答'
|
32 |
+
inputs = []
|
33 |
+
######
|
34 |
+
img_info = np.array([img.encode('utf-8')], dtype=np.string_)
|
35 |
+
img_info = np.tile(img_info, (bs,1))
|
36 |
+
|
37 |
+
|
38 |
+
inputs.append(InferInput("img", img_info.shape,
|
39 |
+
np_to_triton_dtype(img_info.dtype)),)
|
40 |
+
|
41 |
+
|
42 |
+
inputs[0].set_data_from_numpy(img_info)
|
43 |
+
#####
|
44 |
+
title_text = np.array([query.encode('utf-8')], dtype=np.string_)
|
45 |
+
title_text = np.tile(title_text, (bs, 1))
|
46 |
+
|
47 |
+
data_batch = {}
|
48 |
+
data_batch['query'] = title_text
|
49 |
+
|
50 |
+
inputs.append( InferInput("query", data_batch['query'].shape,
|
51 |
+
np_to_triton_dtype(data_batch['query'].dtype))
|
52 |
+
)
|
53 |
+
|
54 |
+
|
55 |
+
inputs[1].set_data_from_numpy(data_batch['query'])
|
56 |
+
|
57 |
+
return inputs
|
58 |
+
|
59 |
+
def make_a_try(inputs, outputs='response', model_name='ensemble_mllm', model_version=''): # qwen 1
|
60 |
+
outputs_list = []
|
61 |
+
ori_outputs_list = outputs.strip().split(",")
|
62 |
+
for out_ele in ori_outputs_list:
|
63 |
+
outputs_list.append(out_ele.strip())
|
64 |
+
outputs = [InferRequestedOutput(x) for x in outputs_list]
|
65 |
+
|
66 |
+
response = my_client.predict(model_name=model_name, inputs=inputs, model_version=model_version, outputs=outputs)
|
67 |
+
|
68 |
+
rsp_info = {}
|
69 |
+
if outputs_list == []:
|
70 |
+
for out_name_ele in response._result.outputs:
|
71 |
+
outputs_list.append(out_name_ele.name)
|
72 |
+
for output_name in outputs_list:
|
73 |
+
res = response.as_numpy(output_name)
|
74 |
+
response = np.expand_dims(res, axis=0)
|
75 |
+
response = response[0].decode('utf-8')
|
76 |
+
rsp_info[output_name] = response
|
77 |
+
print("response:",rsp_info)
|
78 |
+
return rsp_info['response']
|
79 |
+
|
80 |
+
|
81 |
+
|
82 |
+
# def greet(prompt):
|
83 |
+
# """Greet someone."""
|
84 |
+
# # print(prompt)
|
85 |
+
# print("prompt:",prompt)
|
86 |
+
# inputs = prepare_params(prompt)
|
87 |
+
# print(inputs)
|
88 |
+
# result = make_a_try(inputs)
|
89 |
+
|
90 |
+
# return result
|
91 |
+
def greet(prompt,img):
|
92 |
+
"""Greet someone."""
|
93 |
+
# print(prompt)
|
94 |
+
print("prompt:",prompt)
|
95 |
+
inputs = prepare_params(prompt,img)
|
96 |
+
print(inputs)
|
97 |
+
result = make_a_try(inputs)
|
98 |
+
|
99 |
+
return result
|
100 |
+
|
101 |
+
def clear_input():
|
102 |
+
|
103 |
+
return ""
|
104 |
+
|
105 |
+
|
106 |
+
if __name__ == "__main__":
|
107 |
+
param_info = {}
|
108 |
+
# param_info['appkey'] = "com.sankuai.automl.serving"
|
109 |
+
param_info['appkey'] = "com.sankuai.automl.streamvlm"
|
110 |
+
|
111 |
+
# param_info['remote_appkey'] = "com.sankuai.automl.chat3"
|
112 |
+
param_info['remote_appkey'] = "com.sankuai.automl.streamvlm"
|
113 |
+
|
114 |
+
param_info['model_name'] = "ensemble_mllm"
|
115 |
+
param_info['model_version'] = "1"
|
116 |
+
param_info['time_out'] = 60000
|
117 |
+
param_info['server_targets'] = []
|
118 |
+
param_info['outputs'] = 'response'
|
119 |
+
|
120 |
+
appkey, remote_appkey, model_name, model_version, time_out, server_targets = param_info['appkey'], param_info['remote_appkey'], param_info['model_name'], param_info['model_version'], param_info['time_out'], param_info['server_targets']
|
121 |
+
|
122 |
+
my_client = TritonServerClient(appkey=appkey, remote_appkey=remote_appkey, time_out=time_out, server_targets=server_targets)
|
123 |
+
# triton_client.async_stream_infer(
|
124 |
+
# model_name=model_name,
|
125 |
+
# inputs=inputs,
|
126 |
+
# outputs=outputs,
|
127 |
+
# request_id="{}".format(sequence_id),
|
128 |
+
# sequence_id=sequence_id,
|
129 |
+
# sequence_start=True,
|
130 |
+
# sequence_end=True,
|
131 |
+
# )
|
132 |
+
|
133 |
+
img_url = f"https://s3plus.sankuai.com/automl-pkgs/0000.jpeg"
|
134 |
+
|
135 |
+
greet("nihao",img_url)
|
136 |
+
# greet("nihao")
|
137 |
+
print("描述这张图片")
|
try_hd_v2.py
ADDED
@@ -0,0 +1,218 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import argparse
|
2 |
+
import queue
|
3 |
+
import sys
|
4 |
+
import uuid
|
5 |
+
from functools import partial
|
6 |
+
|
7 |
+
import numpy as np
|
8 |
+
import tritonclient.grpc as grpcclient
|
9 |
+
from tritonclient.utils import InferenceServerException
|
10 |
+
import gradio as gr
|
11 |
+
from functools import wraps
|
12 |
+
|
13 |
+
####
|
14 |
+
from PIL import Image
|
15 |
+
import base64
|
16 |
+
import io
|
17 |
+
#####
|
18 |
+
from http.server import HTTPServer, SimpleHTTPRequestHandler
|
19 |
+
import socket
|
20 |
+
####
|
21 |
+
import os
|
22 |
+
import uuid
|
23 |
+
####
|
24 |
+
|
25 |
+
class UserData:
|
26 |
+
def __init__(self):
|
27 |
+
self._completed_requests = queue.Queue()
|
28 |
+
|
29 |
+
def callback(user_data, result, error):
|
30 |
+
if error:
|
31 |
+
user_data._completed_requests.put(error)
|
32 |
+
else:
|
33 |
+
user_data._completed_requests.put(result)
|
34 |
+
|
35 |
+
def make_a_try(img_url,text):
|
36 |
+
model_name = 'ensemble_mllm'
|
37 |
+
user_data = UserData()
|
38 |
+
sequence_id = 100
|
39 |
+
int_sequence_id0 = sequence_id
|
40 |
+
result_list=[]
|
41 |
+
with grpcclient.InferenceServerClient(
|
42 |
+
url="10.199.14.151:8001", verbose = False
|
43 |
+
) as triton_client:
|
44 |
+
try:
|
45 |
+
# Establish stream
|
46 |
+
triton_client.start_stream(
|
47 |
+
callback=partial(callback, user_data),
|
48 |
+
stream_timeout=None,
|
49 |
+
)
|
50 |
+
# Create the tensor for INPUT
|
51 |
+
inputs = []
|
52 |
+
img_url_bytes = img_url.encode("utf-8")
|
53 |
+
img_url_bytes = np.array(img_url_bytes, dtype=bytes)
|
54 |
+
img_url_bytes = img_url_bytes.reshape([1, -1])
|
55 |
+
|
56 |
+
inputs.append(grpcclient.InferInput('IMAGE_URL', img_url_bytes.shape, "BYTES"))
|
57 |
+
inputs[0].set_data_from_numpy(img_url_bytes)
|
58 |
+
|
59 |
+
text_bytes = text.encode("utf-8")
|
60 |
+
text_bytes = np.array(text_bytes, dtype=bytes)
|
61 |
+
text_bytes = text_bytes.reshape([1, -1])
|
62 |
+
# text_input = np.expand_dims(text_bytes, axis=0)
|
63 |
+
text_input = text_bytes
|
64 |
+
|
65 |
+
inputs.append(grpcclient.InferInput('TEXT', text_input.shape, "BYTES"))
|
66 |
+
inputs[1].set_data_from_numpy(text_input)
|
67 |
+
|
68 |
+
outputs = []
|
69 |
+
outputs.append(grpcclient.InferRequestedOutput("OUTPUT"))
|
70 |
+
# Issue the asynchronous sequence inference.
|
71 |
+
triton_client.async_stream_infer(
|
72 |
+
model_name=model_name,
|
73 |
+
inputs=inputs,
|
74 |
+
outputs=outputs,
|
75 |
+
request_id="{}".format(sequence_id),
|
76 |
+
sequence_id=sequence_id,
|
77 |
+
sequence_start=True,
|
78 |
+
sequence_end=True,
|
79 |
+
)
|
80 |
+
######hd
|
81 |
+
except InferenceServerException as error:
|
82 |
+
print(error)
|
83 |
+
# sys.exit(1)
|
84 |
+
# continue
|
85 |
+
return ""
|
86 |
+
|
87 |
+
|
88 |
+
# Retrieve results...
|
89 |
+
recv_count = 0
|
90 |
+
|
91 |
+
while True:
|
92 |
+
try:
|
93 |
+
data_item = user_data._completed_requests.get(timeout=5)
|
94 |
+
except Exception as e:
|
95 |
+
break
|
96 |
+
# data_item = user_data._completed_requests.get()
|
97 |
+
if type(data_item) == InferenceServerException:
|
98 |
+
print('InferenceServerException: ', data_item)
|
99 |
+
# sys.exit(1)
|
100 |
+
return ""
|
101 |
+
this_id = data_item.get_response().id.split("_")[0]
|
102 |
+
if int(this_id) != int_sequence_id0:
|
103 |
+
print("unexpected sequence id returned by the server: {}".format(this_id))
|
104 |
+
# sys.exit(1)
|
105 |
+
return ""
|
106 |
+
####
|
107 |
+
result = data_item.as_numpy("OUTPUT")
|
108 |
+
if len(result[0][0])==0:
|
109 |
+
break
|
110 |
+
####
|
111 |
+
result_list.append(data_item.as_numpy("OUTPUT"))
|
112 |
+
|
113 |
+
recv_count = recv_count + 1
|
114 |
+
result_str = ''.join([item[0][0].decode('utf-8') for item in result_list])
|
115 |
+
return result_str
|
116 |
+
|
117 |
+
|
118 |
+
def greet(image, text):
|
119 |
+
###save img
|
120 |
+
static_path = f"/workdir/yanghandi/gradio_demo/static"
|
121 |
+
# 将图片转换为字节流
|
122 |
+
img_byte_arr = io.BytesIO()
|
123 |
+
try:
|
124 |
+
image.save(img_byte_arr, format='JPEG')
|
125 |
+
except Exception:
|
126 |
+
return ""
|
127 |
+
img_byte_arr = img_byte_arr.getvalue()
|
128 |
+
|
129 |
+
# 为图片生成一个唯一的文件名
|
130 |
+
# filename = "image_" + str(os.getpid()) + ".jpg" #uuid
|
131 |
+
unique_id = uuid.uuid4()
|
132 |
+
filename = f"image_{unique_id}.jpg"
|
133 |
+
filepath = os.path.join(static_path, filename)
|
134 |
+
|
135 |
+
# 将字节流写入文件
|
136 |
+
with open(filepath, 'wb') as f:
|
137 |
+
f.write(img_byte_arr)
|
138 |
+
|
139 |
+
|
140 |
+
img_url = f"http://10.99.5.48:8080/file=static/" + filename
|
141 |
+
# img_url = PIL_to_URL(img_url)
|
142 |
+
# img_url = "http://10.99.5.48:8080/file=static/0000.jpeg"
|
143 |
+
result = make_a_try(img_url,text)
|
144 |
+
# print(result)
|
145 |
+
return result
|
146 |
+
|
147 |
+
# def greet_example(image, text):
|
148 |
+
# ###save img
|
149 |
+
# # filename = image
|
150 |
+
# # static_path = "/workdir/yanghandi/gradio_demo/static"
|
151 |
+
# img_url = "http://10.99.5.48:8080/file=static/0000.jpeg"
|
152 |
+
|
153 |
+
# # img_url = PIL_to_URL(img_url)
|
154 |
+
# # img_url = "http://10.99.5.48:8080/file=static/0000.jpeg"
|
155 |
+
# result = make_a_try(img_url,text)
|
156 |
+
# # print(result)
|
157 |
+
# return result
|
158 |
+
|
159 |
+
def clear_output():
|
160 |
+
|
161 |
+
return ""
|
162 |
+
|
163 |
+
def get_example():
|
164 |
+
return [
|
165 |
+
[f"/workdir/yanghandi/gradio_demo/static/0001.jpg", f"图中的人物是谁"]
|
166 |
+
]
|
167 |
+
if __name__ == "__main__":
|
168 |
+
|
169 |
+
param_info = {}
|
170 |
+
# param_info['appkey'] = "com.sankuai.automl.serving"
|
171 |
+
param_info['appkey'] = "10.199.14.151:8001"
|
172 |
+
|
173 |
+
# param_info['remote_appkey'] = "com.sankuai.automl.chat3"
|
174 |
+
param_info['remote_appkey'] = "10.199.14.151:8001"
|
175 |
+
param_info['model_name'] = 'ensemble_mllm'
|
176 |
+
param_info['model_version'] = "1"
|
177 |
+
param_info['time_out'] = 60000
|
178 |
+
param_info['server_targets'] = []
|
179 |
+
param_info['outputs'] = 'response'
|
180 |
+
|
181 |
+
|
182 |
+
gr.set_static_paths(paths=["static/"])
|
183 |
+
|
184 |
+
with gr.Blocks(title='demo') as demo:
|
185 |
+
gr.Markdown("# 自研模型测试demo")
|
186 |
+
gr.Markdown("尝试使用该demo,上传图片并开始讨论它,或者尝试下面的例子")
|
187 |
+
|
188 |
+
with gr.Row():
|
189 |
+
with gr.Column():
|
190 |
+
# imagebox = gr.Image(value="static/0000.jpeg",type="pil")
|
191 |
+
imagebox = gr.Image(type="pil")
|
192 |
+
promptbox = gr.Textbox(label = "prompt")
|
193 |
+
|
194 |
+
with gr.Column():
|
195 |
+
output = gr.Textbox(label = "output")
|
196 |
+
with gr.Row():
|
197 |
+
submit = gr.Button("submit")
|
198 |
+
clear = gr.Button("clear")
|
199 |
+
|
200 |
+
submit.click(fn=greet,inputs=[imagebox, promptbox],outputs=[output])
|
201 |
+
clear.click(fn=clear_output, inputs=[], outputs=[output])
|
202 |
+
|
203 |
+
gr.Markdown("# example")
|
204 |
+
|
205 |
+
gr.Examples(
|
206 |
+
examples = get_example(),
|
207 |
+
fn = greet,
|
208 |
+
inputs=[imagebox, promptbox],
|
209 |
+
outputs = [output],
|
210 |
+
cache_examples = True
|
211 |
+
)
|
212 |
+
|
213 |
+
demo.launch(server_name="0.0.0.0", server_port=8080, debug=True, share=True)
|
214 |
+
|
215 |
+
|
216 |
+
# img_url = f"https://s3plus.sankuai.com/automl-pkgs/0000.jpeg"
|
217 |
+
# text = f"详细描述一下这张图片"
|
218 |
+
# greet(img_url,text)
|