-
Notifications
You must be signed in to change notification settings - Fork 1
/
flask_app.py
375 lines (293 loc) · 12.8 KB
/
flask_app.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
"""The app main."""
import uuid
from datetime import datetime
from os.path import isfile, join
import gin
import json
import logging
import os
import traceback
import random
from flask import Flask
from flask import render_template, request, Blueprint
from logging.config import dictConfig
from actions.prediction.predict import convert_str_to_options
from actions.util_functions import text2int, get_current_prompt
from logic.core import ExplainBot
from logic.sample_prompts_by_action import sample_prompt_for_action
import easyocr
import numpy as np
from scipy.io.wavfile import read
import librosa
import soundfile as sf
import torch
from transformers import Speech2TextProcessor, Speech2TextForConditionalGeneration
my_uuid = uuid.uuid4()
# gunicorn doesn't have command line flags, using a gin file to pass command line args
@gin.configurable
class GlobalArgs:
def __init__(self, config, baseurl):
self.config = config
self.baseurl = baseurl
# Parse gin global config
gin.parse_config_file("global_config.gin")
# Get args
args = GlobalArgs()
bp = Blueprint('host', __name__, template_folder='templates')
dictConfig({
'version': 1,
'formatters': {'default': {
'format': '[%(asctime)s] %(levelname)s in %(module)s: %(message)s',
}},
'handlers': {'wsgi': {
'class': 'logging.StreamHandler',
'stream': 'ext://flask.logging.wsgi_errors_stream',
'formatter': 'default'
}},
'root': {
'level': 'INFO',
'handlers': ['wsgi']
}
})
os.environ["TOKENIZERS_PARALLELISM"] = "false"
# Try to use all fragmented GPU memory
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "max_split_size_mb:21"
# Parse application level configs
gin.parse_config_file(args.config)
# Setup the explainbot
BOT = ExplainBot()
@bp.route('/')
def home():
"""Load the explanation interface."""
app.logger.info("Loaded Login")
objective = BOT.conversation.describe.get_dataset_objective()
BOT.conversation.build_temp_dataset()
df = BOT.conversation.temp_dataset.contents['X']
f_names = list(BOT.conversation.temp_dataset.contents['X'].columns)
dataset = BOT.conversation.describe.get_dataset_name()
entries = []
for j in range(10):
temp = {}
for f in f_names:
if dataset == "ECQA" and f == "choices":
temp[f] = convert_str_to_options(df[f][j])
else:
temp[f] = df[f][j]
entries.append(temp)
return render_template("index.html", currentUserId="user", datasetObjective=objective, entries=entries,
dataset=dataset)
@bp.route("/log_feedback", methods=['POST'])
def log_feedback():
"""Logs feedback"""
feedback = request.data.decode("utf-8")
app.logger.info(feedback)
split_feedback = feedback.split(" || ")
message = f"Feedback formatted improperly. Got: {split_feedback}"
assert split_feedback[0].startswith("MessageID: "), message
assert split_feedback[1].startswith("Feedback: "), message
assert split_feedback[2].startswith("Username: "), message
assert split_feedback[3].startswith("Answer: "), message
message_id = split_feedback[0][len("MessageID: "):]
feedback_text = split_feedback[1][len("Feedback: "):]
username = split_feedback[2][len("Username: "):]
answer = split_feedback[3][len("Answer: "):]
current_time = datetime.now()
time_stamp = current_time.timestamp()
date_time = datetime.fromtimestamp(time_stamp)
str_time = date_time.strftime("%d-%m-%Y, %H:%M:%S")
logging_info = {
"id": message_id,
"feedback_text": feedback_text,
"username": username,
"answer": answer,
"dataset": BOT.conversation.describe.get_dataset_name(),
"parsed_text": BOT.parsed_text,
"user_text": BOT.user_text,
"timestamp": str_time
}
BOT.log(logging_info)
return ""
@bp.route("/export_history", methods=["Post"])
def export_history():
BOT.export_history()
return ""
@bp.route("/sample_prompt", methods=["Post"])
def sample_prompt():
"""Samples a prompt"""
data = json.loads(request.data)
action = data["action"]
username = data["thisUserName"]
prompt = sample_prompt_for_action(action,
BOT.prompts.filename_to_prompt_id,
BOT.prompts.final_prompt_set,
BOT.conversation)
logging_info = {
"username": username,
"requested_action_generation": action,
"generated_prompt": prompt
}
BOT.log(logging_info)
return prompt
@bp.route("/get_response", methods=['POST'])
def get_bot_response():
"""Load the box response."""
if request.method == "POST":
response = ""
try:
flag = None
audio = None
try:
# Receive the uploaded image
img = request.files["image"]
flag = "img"
except:
pass
try:
data = json.loads(request.data)
flag = "text"
except:
pass
try:
audio = request.files["audio"]
audio.save("recording.wav")
flag = "audio"
except:
pass
if flag == "img":
# Save image locally
img.save(f"./{img.filename}")
app.logger.info(f"Image uploaded!")
if torch.cuda.is_available():
gpu = True
else:
gpu = False
reader = easyocr.Reader(['en'], gpu=gpu) # this needs to run only once to load the model into memory
result = reader.readtext(f"{img.filename}", detail=0)
if len(result) < 2:
raise ValueError("Only one sentence is recognized. Please try other images!")
else:
temp = {'first_input': result[0], 'second_input': result[1]}
BOT.conversation.custom_input = temp
BOT.conversation.used = False
app.logger.info(f"[CUSTOM INPUT] {temp}")
response = "You have given a custom input via uploaded image. " \
"Please enter a follow-up question or prompt! <br><br>" + "Entered custom input: <br>"
if BOT.conversation.describe.get_dataset_name() == "covid_fact":
response += f"Claim: {temp['first_input']} <br>Evidence: {temp['second_input']} <>"
else:
response += f"Question: {temp['first_input']} <br>Choices: {temp['second_input']} <>"
elif flag == "audio":
model = Speech2TextForConditionalGeneration.from_pretrained("facebook/s2t-small-librispeech-asr")
processor = Speech2TextProcessor.from_pretrained("facebook/s2t-small-librispeech-asr")
# Load wav in array
x, _ = librosa.load('./recording.wav', sr=16000)
sf.write('tmp.wav', x, 16000)
a = read("tmp.wav")
temp = np.array(a[1], dtype=np.float64)
inputs = processor(temp, sampling_rate=16000, return_tensors="pt")
generated_ids = model.generate(inputs["input_features"], attention_mask=inputs["attention_mask"])
transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)
# Convert words to digits if available
user_text = text2int(transcription[0])
BOT.user_text = user_text
conversation = BOT.conversation
# Remove generated wav files
os.remove("./recording.wav")
os.remove("./tmp.wav")
app.logger.info("generating the bot response")
response = f"<b>Recorded text</b>: {user_text}<br><br>"
elif flag == "text":
# Change level for QA
level = data["qalevel"]
prompt_type = data["prompt_type"]
BOT.conversation.qa_level = level
BOT.conversation.prompt_type = prompt_type
app.logger.info(f"Prompt type: {prompt_type}")
# Normal user input
if data['custom_input'] == '0':
# Normal user input
user_text = data["userInput"]
BOT.user_text = user_text
conversation = BOT.conversation
app.logger.info("generating the bot response")
response = BOT.update_state(user_text, conversation)
elif data['custom_input'] == '1':
# custom input
user_text = data["userInput"]
if BOT.conversation.describe.get_dataset_name() == "ECQA":
if len(user_text["second_input"].split("-")) != 5:
return "5 choices should be provided and concatenated by '-'!"
BOT.conversation.custom_input = user_text
BOT.conversation.used = False
app.logger.info(f"[CUSTOM INPUT] {user_text}")
response = "You have given a custom input. " \
"Please enter a follow-up question or prompt! <br><br>" + "Entered custom input: <br>"
if BOT.conversation.describe.get_dataset_name() == "covid_fact":
response += f"Claim: {user_text['first_input']} <br>Evidence: {user_text['second_input']} <>"
else:
response += f"Question: {user_text['first_input']} <br>Choices: {user_text['second_input']} <>"
BOT.conversation.store_last_parse(f"custominput '{user_text}'")
else:
# custom input removal
app.logger.info(f"[CUSTOM INPUT] Custom input is removed!")
BOT.conversation.custom_input = None
BOT.conversation.used = True
response = "Entered custom input is now removed! <>"
BOT.write_to_history(BOT.user_text, response)
except torch.cuda.OutOfMemoryError:
response = "I recognized a CUDA out of memory. I suggest to choose a smaller " \
"model for your hardware configuration or with GPTQ/4bit quantization. You can do that by " \
"opening the global_config.gin file and editing the value of GlobalArgs.config to an " \
"equivalent with a model of smaller parameter " \
"size, e.g. \"ecqa_llama_gptq.gin\" or \"ecqa_pythia.gin\"."
except Exception as ext:
app.logger.info(f"Traceback getting bot response: {traceback.format_exc()}")
app.logger.info(f"Exception getting bot response: {ext}")
response = random.choice(BOT.dialogue_flow_map["sorry"])
return response
@bp.route("/custom_input", methods=["Post"])
def custom_input():
data = json.loads(request.data)
custom_input = data["custom_input"]
username = data["thisUserName"]
BOT.conversation.custom_input = custom_input
BOT.conversation.used = False
app.logger.info("custom_input: " + custom_input)
return custom_input
@bp.route("/filter_dataset", methods=["POST"])
def filter_dataset():
filter_text = json.loads(request.data)["filterMsgText"]
df = BOT.conversation.stored_vars["dataset"].contents["X"]
if len(filter_text) > 0:
filtered_df = df[df[BOT.text_fields].apply(lambda row: row.str.contains(filter_text)).any(axis=1)]
BOT.conversation.temp_dataset.contents["X"] = filtered_df
app.logger.info(f"{len(filtered_df)} instances of {BOT.conversation.describe.dataset_name} include the filter "
f"string '{filter_text}'")
final_df = filtered_df
else:
final_df = df
return {
'jsonData': final_df.to_json(orient="index"),
'totalDataLen': len(df)
}
@bp.route("/reset_temp_dataset", methods=["Post"])
def reset_temp_dataset():
data = json.loads(request.data)
username = data["thisUserName"]
# Reset the tempdataset
BOT.conversation.build_temp_dataset()
app.logger.info("Reset temp dataset successfully!")
return "reset temp_dataset"
@bp.route("/get_prompt", methods=["Post"])
def get_prompt():
return get_current_prompt(BOT.parsed_text, BOT.conversation)
app = Flask(__name__)
app.register_blueprint(bp, url_prefix=args.baseurl)
if __name__ != '__main__':
gunicorn_logger = logging.getLogger('gunicorn.error')
app.logger.handlers = gunicorn_logger.handlers
app.logger.setLevel(gunicorn_logger.level)
if __name__ == "__main__":
# clean up storage file on restart
app.logger.info(f"Launching app from config: {args.config}")
app.run(debug=False, port=4455, host="localhost")