responsible-prompting-demo / control /recommendation_handler.py
santanavagner's picture
Upload 72 files
11276e2 verified
#!/usr/bin/env python
# coding: utf-8
# Copyright 2021, IBM Corporation.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Python lib to recommend prompts.
"""
__author__ = "Vagner Santana, Melina Alberio, Cassia Sanctos and Tiago Machado"
__copyright__ = "IBM Corporation 2024"
__credits__ = ["Vagner Santana, Melina Alberio, Cassia Sanctos, Tiago Machado"]
__license__ = "Apache 2.0"
__version__ = "0.0.1"
import requests
import json
import math
import re
import warnings
import pandas as pd
import numpy as np
from sklearn.metrics.pairwise import cosine_similarity
import os
#os.environ['TRANSFORMERS_CACHE'] ="./models/allmini/cache"
import os.path
from sentence_transformers import SentenceTransformer
from umap import UMAP
import tensorflow as tf
from umap.parametric_umap import ParametricUMAP, load_ParametricUMAP
from sentence_transformers import SentenceTransformer
def populate_json(json_file_path = './prompt-sentences-main/prompt_sentences-all-minilm-l6-v2.json',
existing_json_populated_file_path = './prompt-sentences-main/prompt_sentences-all-minilm-l6-v2.json'):
"""
Function that receives a default json file with
empty embeddings and checks whether there is a
partially populated json file.
Args:
json_file_path: Path to json default file with
empty embeddings.
existing_json_populated_file_path: Path to partially
populated json file.
Returns:
A json.
Raises:
Exception when json file can't be loaded.
"""
json_file = json_file_path
if(os.path.isfile(existing_json_populated_file_path)):
json_file = existing_json_populated_file_path
try:
prompt_json = json.load(open(json_file))
json_error = None
return prompt_json, json_error
except Exception as e:
json_error = e
print(f'Error when loading sentences json file: {json_error}')
prompt_json = None
return prompt_json, json_error
def query(texts, api_url, headers):
"""
Function that requests embeddings for a given sentence.
Args:
texts: The sentence or entered prompt text.
api_url: API url for HF request.
headers: Content headers for HF request.
Returns:
A json with the sentence embeddings.
Raises:
Warning: Warns about sentences that have more
than 256 words.
"""
for t in texts:
n_words = len(re.split(r"\s+", t))
if(n_words > 256):
# warning in case of prompts longer than 256 words
warnings.warn("Warning: Sentence provided is longer than 256 words. Model all-MiniLM-L6-v2 expects sentences up to 256 words.")
warnings.warn("Word count:{}".format(n_words))
if('sentence-transformers/all-MiniLM-L6-v2' in api_url):
model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
out = model.encode(texts).tolist()
else:
response = requests.post(api_url, headers=headers, json={"inputs": texts, "options":{"wait_for_model":True}})
out = response.json()
return out
def split_into_sentences(prompt):
"""
Function that splits the input text into sentences based
on punctuation (.!?). The regular expression pattern
'(?<=[.!?]) +' ensures that we split after a sentence-ending
punctuation followed by one or more spaces.
Args:
prompt: The entered prompt text.
Returns:
A list of extracted sentences.
Raises:
Nothing.
"""
sentences = re.split(r'(?<=[.!?]) +', prompt)
return sentences
def get_similarity(embedding1, embedding2):
"""
Function that returns cosine similarity between
two embeddings.
Args:
embedding1: first embedding.
embedding2: second embedding.
Returns:
The similarity value.
Raises:
Nothing.
"""
v1 = np.array( embedding1 ).reshape( 1, -1 )
v2 = np.array( embedding2 ).reshape( 1, -1 )
similarity = cosine_similarity( v1, v2 )
return similarity[0, 0]
def get_distance(embedding1, embedding2):
"""
Function that returns euclidean distance between
two embeddings.
Args:
embedding1: first embedding.
embedding2: second embedding.
Returns:
The euclidean distance value.
Raises:
Nothing.
"""
total = 0
if(len(embedding1) != len(embedding2)):
return math.inf
for i, obj in enumerate(embedding1):
total += math.pow(embedding2[0][i] - embedding1[0][i], 2)
return(math.sqrt(total))
def sort_by_similarity(e):
"""
Function that sorts by similarity.
Args:
e:
Returns:
The sorted similarity value.
Raises:
Nothing.
"""
return e['similarity']
def recommend_prompt(prompt, prompt_json, api_url, headers, add_lower_threshold = 0.3,
add_upper_threshold = 0.5, remove_lower_threshold = 0.1,
remove_upper_threshold = 0.5, model_id = 'sentence-transformers/all-minilm-l6-v2'):
"""
Function that recommends prompts additions or removals.
Args:
prompt: The entered prompt text.
prompt_json: Json file populated with embeddings.
api_url: API url for HF request.
headers: Content headers for HF request.
add_lower_threshold: Lower threshold for sentence addition,
the default value is 0.3.
add_upper_threshold: Upper threshold for sentence addition,
the default value is 0.5.
remove_lower_threshold: Lower threshold for sentence removal,
the default value is 0.3.
remove_upper_threshold: Upper threshold for sentence removal,
the default value is 0.5.
model_id: Id of the model, the default value is all-minilm-l6-v2 movel.
Returns:
Prompt values to add or remove.
Raises:
Nothing.
"""
if(model_id == 'baai/bge-large-en-v1.5' ):
json_file = './prompt-sentences-main/prompt_sentences-bge-large-en-v1.5.json'
umap_model = load_ParametricUMAP('./models/umap/BAAI/bge-large-en-v1.5/')
elif(model_id == 'intfloat/multilingual-e5-large'):
json_file = './prompt-sentences-main/prompt_sentences-multilingual-e5-large.json'
umap_model = load_ParametricUMAP('./models/umap/intfloat/multilingual-e5-large/')
else: # fall back to all-minilm as default
json_file = './prompt-sentences-main/prompt_sentences-all-minilm-l6-v2.json'
umap_model = load_ParametricUMAP('./models/umap/sentence-transformers/all-MiniLM-L6-v2/')
prompt_json = json.load(open(json_file))
# Output initialization
out, out['input'], out['add'], out['remove'] = {}, {}, {}, {}
input_items, items_to_add, items_to_remove = [], [], []
# Spliting prompt into sentences
input_sentences = split_into_sentences(prompt)
# TODO: Request embeddings for input an d store in a input_embeddingS
# Recommendation of values to add to the current prompt
# Using only the last sentence for the add recommendation
input_embedding = query(input_sentences[-1], api_url, headers)
for v in prompt_json['positive_values']:
# Dealing with values without prompts and makinig sure they have the same dimensions
if(len(v['centroid']) == len(input_embedding)):
if(get_similarity(pd.DataFrame(input_embedding), pd.DataFrame(v['centroid'])) > add_lower_threshold):
closer_prompt = -1
for p in v['prompts']:
d_prompt = get_similarity(pd.DataFrame(input_embedding), pd.DataFrame(p['embedding']))
# The sentence_threshold is being used as a ceiling meaning that for high similarities the sentence/value might already be presente in the prompt
# So, we don't want to recommend adding something that is already there
if(d_prompt > closer_prompt and d_prompt > add_lower_threshold and d_prompt < add_upper_threshold):
closer_prompt = d_prompt
items_to_add.append({
'value': v['label'],
'prompt': p['text'],
'similarity': d_prompt,
'x': p['x'],
'y': p['y']})
out['add'] = items_to_add
# Recommendation of values to remove from the current prompt
i = 0
# Recommendation of values to remove from the current prompt
for sentence in input_sentences:
input_embedding = query(sentence, api_url, headers) # remote
# Obtaining XY coords for input sentences from a parametric UMAP model
if(len(prompt_json['negative_values'][0]['centroid']) == len(input_embedding) and sentence != ''):
embeddings_umap = umap_model.transform(tf.expand_dims(pd.DataFrame(input_embedding), axis=0))
input_items.append({
'sentence': sentence,
'x': str(embeddings_umap[0][0]),
'y': str(embeddings_umap[0][1])
})
for v in prompt_json['negative_values']:
# Dealing with values without prompts and makinig sure they have the same dimensions
if(len(v['centroid']) == len(input_embedding)):
if(get_similarity(pd.DataFrame(input_embedding), pd.DataFrame(v['centroid'])) > remove_lower_threshold):
closer_prompt = -1
for p in v['prompts']:
d_prompt = get_similarity(pd.DataFrame(input_embedding), pd.DataFrame(p['embedding']))
# A more restrict threshold is used here to prevent false positives
# The sentence_threshold is being used to indicate that there must be a sentence in the prompt that is similiar to one of our adversarial prompts
# So, yes, we want to recommend the removal of something adversarial we've found
if(d_prompt > closer_prompt and d_prompt > remove_upper_threshold):
closer_prompt = d_prompt
items_to_remove.append({
'value': v['label'],
'sentence': sentence,
'sentence_index': i,
'closest_harmful_sentence': p['text'],
'similarity': d_prompt,
'x': p['x'],
'y': p['y']})
out['remove'] = items_to_remove
i += 1
out['input'] = input_items
out['add'] = sorted(out['add'], key=sort_by_similarity, reverse=True)
values_map = {}
for item in out['add'][:]:
if(item['value'] in values_map):
out['add'].remove(item)
else:
values_map[item['value']] = item['similarity']
out['add'] = out['add'][0:5]
out['remove'] = sorted(out['remove'], key=sort_by_similarity, reverse=True)
values_map = {}
for item in out['remove'][:]:
if(item['value'] in values_map):
out['remove'].remove(item)
else:
values_map[item['value']] = item['similarity']
out['remove'] = out['remove'][0:5]
return out
def get_thresholds(prompts, prompt_json, api_url, headers, model_id = 'sentence-transformers/all-minilm-l6-v2'):
"""
Function that recommends thresholds given an array of prompts.
Args:
prompts: The array with samples of prompts to be used in the system.
prompt_json: Sentences to be forwarded to the recommendation endpoint.
model_id: Id of the model, the default value is all-minilm-l6-v2 model.
Returns:
A map with thresholds for the sample prompts and the informed model.
Raises:
Nothing.
"""
# Array limits for retrieving the thresholds
# if( len( prompts ) < 10 or len( prompts ) > 30 ):
# return -1
add_similarities = []
remove_similarities = []
for p_id, p in enumerate(prompts):
out = recommend_prompt(p, prompt_json, api_url, headers, 0, 1, 0, 0, model_id) # Wider possible range
for r in out['add']:
add_similarities.append(r['similarity'])
for r in out['remove']:
remove_similarities.append(r['similarity'])
add_similarities_df = pd.DataFrame({'similarity': add_similarities})
remove_similarities_df = pd.DataFrame({'similarity': remove_similarities})
thresholds = {}
thresholds['add_lower_threshold'] = round(add_similarities_df.describe([.1]).loc['10%', 'similarity'], 1)
thresholds['add_higher_threshold'] = round(add_similarities_df.describe([.9]).loc['90%', 'similarity'], 1)
thresholds['remove_lower_threshold'] = round(remove_similarities_df.describe([.1]).loc['10%', 'similarity'], 1)
thresholds['remove_higher_threshold'] = round(remove_similarities_df.describe([.9]).loc['90%', 'similarity'], 1)
return thresholds
def recommend_local(prompt, prompt_json, model_id, model_path = './models/all-MiniLM-L6-v2/', add_lower_threshold = 0.3,
add_upper_threshold = 0.5, remove_lower_threshold = 0.1,
remove_upper_threshold = 0.5):
"""
Function that recommends prompts additions or removals
using a local model.
Args:
prompt: The entered prompt text.
prompt_json: Json file populated with embeddings.
model_id: Id of the local model.
model_path: Path to the local model.
Returns:
Prompt values to add or remove.
Raises:
Nothing.
"""
if(model_id == 'baai/bge-large-en-v1.5' ):
json_file = './prompt-sentences-main/prompt_sentences-bge-large-en-v1.5.json'
umap_model = load_ParametricUMAP('./models/umap/BAAI/bge-large-en-v1.5/')
elif(model_id == 'intfloat/multilingual-e5-large'):
json_file = './prompt-sentences-main/prompt_sentences-multilingual-e5-large.json'
umap_model = load_ParametricUMAP('./models/umap/intfloat/multilingual-e5-large/')
else: # fall back to all-minilm as default
json_file = './prompt-sentences-main/prompt_sentences-all-minilm-l6-v2.json'
umap_model = load_ParametricUMAP('./models/umap/sentence-transformers/all-MiniLM-L6-v2/')
prompt_json = json.load(open(json_file))
# Output initialization
out, out['input'], out['add'], out['remove'] = {}, {}, {}, {}
input_items, items_to_add, items_to_remove = [], [], []
# Spliting prompt into sentences
input_sentences = split_into_sentences(prompt)
# Recommendation of values to add to the current prompt
# Using only the last sentence for the add recommendation
model = SentenceTransformer(model_path)
input_embedding = model.encode(input_sentences[-1])
for v in prompt_json['positive_values']:
# Dealing with values without prompts and makinig sure they have the same dimensions
if(len(v['centroid']) == len(input_embedding)):
if(get_similarity(pd.DataFrame(input_embedding), pd.DataFrame(v['centroid'])) > add_lower_threshold):
closer_prompt = -1
for p in v['prompts']:
d_prompt = get_similarity(pd.DataFrame(input_embedding), pd.DataFrame(p['embedding']))
# The sentence_threshold is being used as a ceiling meaning that for high similarities the sentence/value might already be presente in the prompt
# So, we don't want to recommend adding something that is already there
if(d_prompt > closer_prompt and d_prompt > add_lower_threshold and d_prompt < add_upper_threshold):
closer_prompt = d_prompt
items_to_add.append({
'value': v['label'],
'prompt': p['text'],
'similarity': d_prompt,
'x': p['x'],
'y': p['y']})
out['add'] = items_to_add
# Recommendation of values to remove from the current prompt
i = 0
# Recommendation of values to remove from the current prompt
for sentence in input_sentences:
input_embedding = model.encode(sentence) # local
# Obtaining XY coords for input sentences from a parametric UMAP model
if(len(prompt_json['negative_values'][0]['centroid']) == len(input_embedding) and sentence != ''):
embeddings_umap = umap_model.transform(tf.expand_dims(pd.DataFrame(input_embedding), axis=0))
input_items.append({
'sentence': sentence,
'x': str(embeddings_umap[0][0]),
'y': str(embeddings_umap[0][1])
})
for v in prompt_json['negative_values']:
# Dealing with values without prompts and makinig sure they have the same dimensions
if(len(v['centroid']) == len(input_embedding)):
if(get_similarity(pd.DataFrame(input_embedding), pd.DataFrame(v['centroid'])) > remove_lower_threshold):
closer_prompt = -1
for p in v['prompts']:
d_prompt = get_similarity(pd.DataFrame(input_embedding), pd.DataFrame(p['embedding']))
# A more restrict threshold is used here to prevent false positives
# The sentence_threhold is being used to indicate that there must be a sentence in the prompt that is similiar to one of our adversarial prompts
# So, yes, we want to recommend the revolval of something adversarial we've found
if(d_prompt > closer_prompt and d_prompt > remove_upper_threshold):
closer_prompt = d_prompt
items_to_remove.append({
'value': v['label'],
'sentence': sentence,
'sentence_index': i,
'closest_harmful_sentence': p['text'],
'similarity': d_prompt,
'x': p['x'],
'y': p['y']})
out['remove'] = items_to_remove
i += 1
out['input'] = input_items
out['add'] = sorted(out['add'], key=sort_by_similarity, reverse=True)
values_map = {}
for item in out['add'][:]:
if(item['value'] in values_map):
out['add'].remove(item)
else:
values_map[item['value']] = item['similarity']
out['add'] = out['add'][0:5]
out['remove'] = sorted(out['remove'], key=sort_by_similarity, reverse=True)
values_map = {}
for item in out['remove'][:]:
if(item['value'] in values_map):
out['remove'].remove(item)
else:
values_map[item['value']] = item['similarity']
out['remove'] = out['remove'][0:5]
return out