File size: 11,052 Bytes
d195d4f |
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 |
import json
import yaml
import os
import time
import math
import pdb
from typing import List, Dict
from factool.knowledge_qa.tool import google_search
from factool.knowledge_qa.tool import local_search
from factool.utils.base.pipeline import pipeline
class knowledge_qa_pipeline(pipeline):
def __init__(self, foundation_model, snippet_cnt, search_type, data_link=None, Embed_link=None):
super().__init__('knowledge_qa', foundation_model)
if(search_type == 'online'):
self.tool = google_search(snippet_cnt = snippet_cnt)
elif(search_type == 'local'):
self.tool = local_search(snippet_cnt = snippet_cnt, data_link=data_link, embedding_link=Embed_link)
with open(os.path.join(self.prompts_path, "claim_extraction.yaml"), 'r') as file:
data = yaml.load(file, Loader=yaml.FullLoader)
self.claim_prompt = data['knowledge_qa']
with open(os.path.join(self.prompts_path, 'query_generation.yaml'), 'r') as file:
data = yaml.load(file, Loader=yaml.FullLoader)
self.query_prompt = data['knowledge_qa']
with open(os.path.join(self.prompts_path, 'agreement_verification.yaml'), 'r') as file:
data = yaml.load(file, Loader=yaml.FullLoader)
self.verification_prompt = data['knowledge_qa']
async def _claim_extraction(self, responses):
messages_list = [
[
{"role": "system", "content": self.claim_prompt['system']},
{"role": "user", "content": self.claim_prompt['user'].format(input=response)},
]
for response in responses
]
return await self.chat.async_run(messages_list, List)
async def _query_generation(self, claims):
if claims == None:
return ['None']
messages_list = [
[
{"role": "system", "content": self.query_prompt['system']},
{"role": "user", "content": self.query_prompt['user'].format(input=claim['claim'] if 'claim' in claim else '')},
]
for claim in claims
]
return await self.chat.async_run(messages_list, List)
async def _verification(self, claims, evidences):
messages_list = [
[
{"role": "system", "content": self.verification_prompt['system']},
{"role": "user", "content": self.verification_prompt['user'].format(claim=claim['claim'], evidence=str(evidence))},
]
for claim, evidence in zip(claims, evidences)
]
return await self.chat.async_run(messages_list, Dict)
async def run_with_tool_live(self, responses):
claims_in_responses = await self._claim_extraction(responses)
queries_in_responses = []
evidences_in_responses = []
sources_in_responses = []
verifications_in_responses = []
#pdb.set_trace()
for claims_in_response in claims_in_responses:
queries = await self._query_generation(claims_in_response)
queries_in_responses.append(queries)
search_outputs_for_claims = await self.tool.run(queries)
evidences = [output["content"] for search_outputs_for_claim in search_outputs_for_claims for output in search_outputs_for_claim]
evidences_in_responses.append(evidences)
sources = [output["source"] for search_outputs_for_claim in search_outputs_for_claims for output in search_outputs_for_claim]
sources_in_responses.append(sources)
verifications = await self._verification(claims_in_response, evidences)
verifications_in_responses.append(verifications)
return claims_in_responses, queries_in_responses, evidences_in_responses, sources_in_responses, verifications_in_responses
async def run_with_tool_live_without_claim_extraction(self, claims):
queries = await self._query_generation(claims)
evidences = await self.tool.run(queries)
final_response = await self._verification(claims, evidences)
for i in range(len(final_response)):
if final_response[i] != None:
final_response[i]['queries'] = queries[i]
final_response[i]['evidences'] = evidences[i]
return final_response
async def run_with_tool_api_call(self, prompts, responses):
batch_size = 5
num_batches = math.ceil(len(prompts) / batch_size)
self.sample_list = [{"prompt": prompt, "response": response, "category": 'kbqa'} for prompt, response in zip(prompts, responses)]
for i in range(num_batches):
print(i)
batch_start = i * batch_size
batch_end = min((i + 1) * batch_size, len(responses))
claims_in_responses, queries_in_responses, evidences_in_responses, sources_in_responses, verifications_in_responses = await self.run_with_tool_live(responses[batch_start:batch_end])
for j, (claims_in_response, queries_in_response, evidences_in_response, sources_in_response, verifications_in_response) in enumerate(zip(claims_in_responses, queries_in_responses, evidences_in_responses, sources_in_responses, verifications_in_responses)):
index = batch_start + j
if claims_in_response != None:
for k, claim in enumerate(claims_in_response):
if verifications_in_response[k] != None:
if claim != None:
verifications_in_response[k].update({'claim': claim['claim']})
else:
verifications_in_response[k].update({'claim': 'None'})
evidences_with_source = []
for evidence, source in zip(evidences_in_response, sources_in_response):
evidences_with_source.append({'evidence': evidence, 'source': source})
self.sample_list[index].update({
'claims': claims_in_response,
'queries': queries_in_response,
# 'evidences': evidences_in_response,
# 'sources': sources_in_response,
'evidences': evidences_with_source,
'claim_level_factuality': verifications_in_response,
'response_level_factuality': all([verification['factuality'] if verification != None else True for verification in verifications_in_response])
})
return self.sample_list
async def run_with_tool_dataset(self, annotated_dataset_path: str, with_tool_classified_dataset_path: str, rerun: bool = False, rerun_indices: list = []):
data_path = with_tool_classified_dataset_path if rerun else annotated_dataset_path
with open(data_path, 'r') as f:
data = [json.loads(line) for line in f]
self.sample_list = data if rerun else [claim for sample in data for claim in sample['claims']]
rerun_elements = self.sample_list if not rerun else [self.sample_list[i] for i in rerun_indices]
batch_size = 4
num_batches = math.ceil(len(rerun_elements) / batch_size) # 5
for i in range(num_batches):
print(i)
batch_start = i * batch_size
batch_end = min((i + 1) * batch_size, len(rerun_elements))
responses = await self.run_with_tool_live_without_claim_extraction(rerun_elements[batch_start:batch_end])
for j, response in enumerate(responses):
index = batch_start + j if rerun == False else rerun_indices[batch_start + j]
if response is None:
self.sample_list[index].update({
'with_tool_classification': 'None',
'with_tool_reasoning': 'None',
'queries': 'None',
'evidences': 'None'
})
else:
self.sample_list[index].update({
'with_tool_classification': response.get('factuality', 'None'),
'with_tool_reasoning': response.get('reasoning', 'None'),
'queries': response.get('queries', 'None'),
'evidences': response.get('evidences', 'None')
})
# save everything after each batch to prevent data loss
with open(with_tool_classified_dataset_path, 'w') as f:
for item in self.sample_list:
json_str = json.dumps(item)
f.write(json_str + '\n')
async def run_self_check_live(self, fewshot, batch):
user_prompt_key = 'user_3_shot_CoT' if fewshot else 'user_zero_shot_CoT'
messages_list = [
[
{"role": "system", "content": self.self_check_prompt['system']},
{"role": "user", "content": self.self_check_prompt[user_prompt_key].format(claim=response['claim'])},
]
for response in batch
]
return await self.chat.async_run(messages_list, Dict)
async def run_self_check_dataset(self, annotated_dataset_path: str, self_check_classified_dataset_path: str, fewshot: bool = False, rerun: bool = False, rerun_indices: list = []):
data_path = annotated_dataset_path if not rerun else self_check_classified_dataset_path
with open(data_path, 'r') as f:
data = [json.loads(line) for line in f]
self.sample_list = data if rerun else [claim for sample in data for claim in sample['claims']]
rerun_elements = self.sample_list if not rerun else [self.sample_list[i] for i in rerun_indices]
batch_size = 10
num_batches = math.ceil(len(rerun_elements) / batch_size)
for i in range(num_batches):
print(i)
batch_start = i * batch_size
batch_end = min((i + 1) * batch_size, len(rerun_elements))
batch = rerun_elements[batch_start:batch_end]
responses = await self.run_self_check_live(fewshot, batch)
for j, response in enumerate(responses):
index = batch_start + j if not rerun else rerun_indices[batch_start + j]
if response is None:
self.sample_list[index].update({
'self_check_classification': 'None',
'self_check_reasoning': 'None'
})
else:
self.sample_list[index].update({
'self_check_classification': response.get('factuality', 'None'),
'self_check_reasoning': response.get('reasoning', 'None')
})
# save everything after each batch to prevent data loss
with open(self_check_classified_dataset_path, 'w') as f:
for item in self.sample_list:
json_str = json.dumps(item)
f.write(json_str + '\n')
|