File size: 11,788 Bytes
232b620
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# https://www.mixedbread.ai/blog/mxbai-embed-large-v1
# https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1

import os
import time
import pandas as pd
import numpy as np
from typing import Dict

import torch
from transformers import AutoModel, AutoTokenizer
from sentence_transformers.util import cos_sim
from accelerate import Accelerator  # Import from accelerate
from scipy.stats import zscore

# Set up environment variables for Hugging Face caching
os.environ["HF_HUB_CACHE"] = "/eos/jeodpp/home/users/consose/cache/huggingface/hub"
os.environ["HUGGINGFACE_HUB_CACHE"] = "/eos/jeodpp/home/users/consose/cache/huggingface/hub"
os.environ["HF_HOME"] = "/eos/jeodpp/home/users/consose/cache/huggingface/hub"

# Initialize the Accelerator
accelerator = Accelerator()

# Use the device managed by Accelerator
device = accelerator.device
print("Using accelerator device =", device)


# 1. Load the model and tokenizer
model_id_Retriever = 'mixedbread-ai/mxbai-embed-large-v1'
tokenizer_Retriever = AutoTokenizer.from_pretrained(model_id_Retriever)
modelRetriever = AutoModel.from_pretrained(model_id_Retriever)

# Accelerate prepares the model (e.g., moves to the appropriate device)
modelRetriever = accelerator.prepare(modelRetriever)




# Define the transform_query function
def transform_query(queryText: str) -> str:
    """For retrieval, add the prompt for queryText (not for documents)."""
    return f'Represent this sentence for searching relevant passages: {queryText}'

# Define the pooling function
def pooling(outputs: torch.Tensor, inputs: Dict, strategy: str = 'cls') -> np.ndarray:
    if strategy == 'cls':
        outputs = outputs[:, 0]
    elif strategy == 'mean':
        outputs = torch.sum(
            outputs * inputs["attention_mask"][:, :, None], dim=1
        ) / torch.sum(inputs["attention_mask"], dim=1, keepdim=True)
    else:
        raise NotImplementedError
    return outputs.detach().cpu().numpy()


def retrievePassageSimilarities(queryText, passages):
    # Create the docs list by adding the transformed queryText and then the passages
    docs = [transform_query(queryText)] + passages

    # 2. Encode the inputs
    inputs = tokenizer_Retriever(docs, padding=True, return_tensors='pt')

    # Move inputs to the right device using accelerator
    inputs = {k: v.to(device) for k, v in inputs.items()}
    with torch.no_grad():
        outputs = modelRetriever(**inputs).last_hidden_state
    embeddings = pooling(outputs, inputs, 'cls')

    similarities = cos_sim(embeddings[0], embeddings[1:])

    #print('similarities:', similarities)

    return similarities



def RAG_retrieval_Base(queryText,passages, min_threshold=0.0, max_num_passages=None):

    try:
        similarities=retrievePassageSimilarities(queryText, passages)

        #Create a DataFrame
        df = pd.DataFrame({
            'Passage': passages,
            'Similarity': similarities.flatten()  # Flatten the similarity tensor/array to ensure compatibility
        })

        # Filter the DataFrame based on the similarity threshold
        df_filtered = df[df['Similarity'] >= min_threshold]

        # If max_num_passages is specified, limit the number of passages returned
        if max_num_passages is not None:
            df_filtered = df_filtered.nlargest(max_num_passages, 'Similarity')

        df_filtered = df_filtered.sort_values(by='Similarity', ascending=False)

        # Return the filtered DataFrame
        return df_filtered

    except Exception as e:
        # Log the exception message or handle it as needed
        print(f"An error occurred: {e}")
        return pd.DataFrame()  # Return an empty DataFrame in case of error



def RAG_retrieval_Z_scores(queryText, passages, z_threshold=1.0, max_num_passages=None, min_threshold=0.5):
    try:
        # Encoding and similarity computation remains the same

        similarities = retrievePassageSimilarities(queryText, passages)

        # Calculate z-scores for similarities
        z_scores = zscore(similarities.flatten())

        # Create a DataFrame with passages, similarities, and z-scores
        df = pd.DataFrame({
            'Passage': passages,
            'Similarity': similarities.flatten(),
            'Z-Score': z_scores
        })

        # Filter passages based on z-score threshold
        df_filtered = df[df['Z-Score'] >= z_threshold]

        if min_threshold:
            # Filter the DataFrame also on min similarity threshold
            df_filtered = df[df['Similarity'] >= min_threshold]

        # If max_num_passages is specified, limit the number of passages returned
        if max_num_passages is not None:
            df_filtered = df_filtered.nlargest(max_num_passages, 'Similarity')

        # Sort by similarity (or z-score if preferred)
        df_filtered = df_filtered.sort_values(by='Similarity', ascending=False)

        return df_filtered

    except Exception as e:
        # Log the exception message or handle it as needed
        print(f"An error occurred: {e}")
        return pd.DataFrame()  # Return an empty DataFrame in case of error




def RAG_retrieval_Percentile(queryText, passages, percentile=90,max_num_passages=None, min_threshold=0.5):
    try:
        # Encoding and similarity computation remains the same

        similarities = retrievePassageSimilarities(queryText, passages)

        # Determine threshold based on percentile
        threshold = np.percentile(similarities.flatten(), percentile)

        # Create a DataFrame
        df = pd.DataFrame({
            'Passage': passages,
            'Similarity': similarities.flatten()
        })

        # Filter using percentile threshold
        df_filtered = df[df['Similarity'] >= threshold]

        if min_threshold:
            # Filter the DataFrame also on min similarity threshold
            df_filtered = df[df['Similarity'] >= min_threshold]

        # If max_num_passages is specified, limit the number of passages returned
        if max_num_passages is not None:
            df_filtered = df_filtered.nlargest(max_num_passages, 'Similarity')

        # Sort by similarity
        df_filtered = df_filtered.sort_values(by='Similarity', ascending=False)

        return df_filtered

    except Exception as e:
        # Log the exception message or handle it as needed
        print(f"An error occurred: {e}")
        return pd.DataFrame()  # Return an empty DataFrame in case of error



def RAG_retrieval_TopK(queryText, passages, top_fraction=0.1, max_num_passages=None, min_threshold=0.5):
    try:
        # Encoding and similarity computation (assuming retrievePassageSimilarities is defined elsewhere)
        similarities = retrievePassageSimilarities(queryText, passages)

        # Calculate the number of passages to select based on top fraction
        num_passages_TopFraction = max(1, int(top_fraction * len(passages)))

        # Create a DataFrame
        df = pd.DataFrame({
            'Passage': passages,
            'Similarity': similarities.flatten()
        })

        # Select the top passages dynamically
        df_filtered = df.nlargest(num_passages_TopFraction, 'Similarity')

        if min_threshold:
            # Filter the DataFrame also on min similarity threshold
            df_filtered = df_filtered[df_filtered['Similarity'] >= min_threshold]

        # If max_num_passages is specified, limit the number of passages returned
        if max_num_passages is not None:
            df_filtered = df_filtered.nlargest(max_num_passages, 'Similarity')

        # Sort by similarity
        df_filtered = df_filtered.sort_values(by='Similarity', ascending=False)

        return df_filtered

    except Exception as e:
        # Log the exception message or handle it as needed
        print(f"An error occurred: {e}")
        return pd.DataFrame()  # Return an empty DataFrame in case of error



if __name__ == '__main__':

    queryText = 'A man is eating a piece of bread'

    # Define the passages list
    passages = [
        "A man is eating food.",
        "A man is eating pasta.",
        "The girl is carrying a baby.",
        "A man is riding a horse.",
    ]

    #df_retrieved = RAG_retrieval_Base(queryText, passages)
    #df_retrieved = RAG_retrieval_Base(queryText, passages, min_threshold=0.5)
    #df_retrieved = RAG_retrieval_Base(queryText, passages, max_num_passages=3)
    df_retrieved = RAG_retrieval_Base(queryText, passages, min_threshold=0.5, max_num_passages=3)

    #df_retrieved = RAG_retrieval_Z_scores(queryText, passages, z_threshold=1.0)
    #df_retrieved = RAG_retrieval_Z_scores(queryText, passages, z_threshold=1.0,max_num_passages=3)

    #df_retrieved = RAG_retrieval_Percentile(queryText, passages, percentile=80)
    # df_retrieved = RAG_retrieval_Percentile(queryText, passages, percentile=80, max_num_passages=3)

    ##df_retrieved = RAG_retrieval_TopK(queryText, passages, top_fraction=0.2)
    #df_retrieved = RAG_retrieval_TopK(queryText, passages, top_fraction=0.2, max_num_passages=3)


    print(df_retrieved)

    #labelTriplesLIST_RAGGED = df_retrieved['Passage'].apply(lambda x: (x,)).tolist()


    print("end of computations")

# VERSION WITHOUT ACCELERATE
#
# #https://www.mixedbread.ai/blog/mxbai-embed-large-v1
# #https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1
#
# import os
#
# os.environ["HF_HUB_CACHE"] = "/eos/jeodpp/home/users/consose/cache/huggingface/hub"
# os.environ["HUGGINGFACE_HUB_CACHE"] = "/eos/jeodpp/home/users/consose/cache/huggingface/hub"
# os.environ["HF_HOME"] = "/eos/jeodpp/home/users/consose/cache/huggingface/hub"
#
# import time
# import pandas as pd
# import numpy as np
#
#
#
# from typing import Dict
#
# import torch
# import numpy as np
# from transformers import AutoModel, AutoTokenizer
# from sentence_transformers.util import cos_sim
#
# # For retrieval you need to pass this prompt. Please find our more in our blog post.
# def transform_queryText(queryText: str) -> str:
#     """ For retrieval, add the prompt for queryText (not for documents).
#     """
#     return f'Represent this sentence for searching relevant passages: {queryText}'
#
# # The model works really well with cls pooling (default) but also with mean pooling.
# def pooling(outputs: torch.Tensor, inputs: Dict,  strategy: str = 'cls') -> np.ndarray:
#     if strategy == 'cls':
#         outputs = outputs[:, 0]
#     elif strategy == 'mean':
#         outputs = torch.sum(
#             outputs * inputs["attention_mask"][:, :, None], dim=1) / torch.sum(inputs["attention_mask"], dim=1, keepdim=True)
#     else:
#         raise NotImplementedError
#     return outputs.detach().cpu().numpy()
#
# # 1. load model
# model_id = 'mixedbread-ai/mxbai-embed-large-v1'
# tokenizer = AutoTokenizer.from_pretrained(model_id)
# model = AutoModel.from_pretrained(model_id).cuda()
#
# queryText = 'A man is eating a piece of bread'
#
# # Define the passages list
# passages = [
#     "A man is eating food.",
#     "A man is eating pasta.",
#     "The girl is carrying a baby.",
#     "A man is riding a horse.",
# ]
#
# # Create the docs list by adding the transformed queryText and then the passages
# docs = [transform_queryText(queryText)] + passages
#
# # 2. encode
# inputs = tokenizer(docs, padding=True, return_tensors='pt')
# for k, v in inputs.items():
#     inputs[k] = v.cuda()
# outputs = model(**inputs).last_hidden_state
# embeddings = pooling(outputs, inputs, 'cls')
#
# similarities = cos_sim(embeddings[0], embeddings[1:])
#
# print('similarities:', similarities)
#
#
# # Create a DataFrame
# df = pd.DataFrame({
#     'Passage': passages,
#     'Similarity': similarities.flatten()  # Flatten the similarity tensor/array to ensure compatibility
# })
#
# # Display the DataFrame
# print(df)
#
#
# print("end of computations")