File size: 4,186 Bytes
c5e4524
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import json
import re

import openai
import pandas as pd
import requests
import spacy
import spacy_transformers
import streamlit_scrollable_textbox as stx
import torch
from sentence_transformers import SentenceTransformer
from tqdm import tqdm
from transformers import (
    AutoModelForMaskedLM,
    AutoModelForSeq2SeqLM,
    AutoTokenizer,
    T5ForConditionalGeneration,
    T5Tokenizer,
    pipeline,
)

import pinecone
import streamlit as st


@st.experimental_singleton
def get_data():
    data = pd.read_csv("earnings_calls_cleaned_metadata.csv")
    return data


# Initialize Spacy Model


@st.experimental_singleton
def get_spacy_model():
    return spacy.load("en_core_web_trf")


@st.experimental_singleton
def get_flan_alpaca_xl_model():
    model = AutoModelForSeq2SeqLM.from_pretrained(
        "/home/user/app/models/flan-alpaca-xl/"
    )
    tokenizer = AutoTokenizer.from_pretrained(
        "/home/user/app/models/flan-alpaca-xl/"
    )
    return model, tokenizer


# Initialize models from HuggingFace


@st.experimental_singleton
def get_t5_model():
    return pipeline("summarization", model="t5-small", tokenizer="t5-small")


@st.experimental_singleton
def get_flan_t5_model():
    tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-large")
    model = T5ForConditionalGeneration.from_pretrained("google/flan-t5-large")
    return model, tokenizer


@st.experimental_singleton
def get_mpnet_embedding_model():
    device = "cuda" if torch.cuda.is_available() else "cpu"
    model = SentenceTransformer(
        "sentence-transformers/all-mpnet-base-v2", device=device
    )
    model.max_seq_length = 512
    return model


@st.experimental_singleton
def get_splade_sparse_embedding_model():
    model_sparse = "naver/splade-cocondenser-ensembledistil"
    # check device
    device = "cuda" if torch.cuda.is_available() else "cpu"
    tokenizer = AutoTokenizer.from_pretrained(model_sparse)
    model_sparse = AutoModelForMaskedLM.from_pretrained(model_sparse)
    # move to gpu if available
    model_sparse.to(device)
    return model_sparse, tokenizer


@st.experimental_singleton
def get_sgpt_embedding_model():
    device = "cuda" if torch.cuda.is_available() else "cpu"
    model = SentenceTransformer(
        "Muennighoff/SGPT-125M-weightedmean-nli-bitfit", device=device
    )
    model.max_seq_length = 512
    return model


@st.experimental_memo
def save_key(api_key):
    return api_key


# Text Generation


def gpt_model(prompt):
    response = openai.Completion.create(
        model="text-davinci-003",
        prompt=prompt,
        temperature=0,
        max_tokens=1024,
    )
    return response.choices[0].text


def generate_text_flan_t5(model, tokenizer, input_text):
    input_ids = tokenizer(input_text, return_tensors="pt").input_ids
    outputs = model.generate(input_ids, temperature=0.5, max_length=512)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)


# Entity Extraction


def generate_entities_flan_alpaca_inference_api(prompt):
    API_URL = "https://api-inference.huggingface.co/models/declare-lab/flan-alpaca-xl"
    API_TOKEN = st.secrets["hg_key"]
    headers = {"Authorization": f"Bearer {API_TOKEN}"}
    payload = {
        "inputs": prompt,
        "parameters": {
            "do_sample": True,
            "temperature": 0.1,
            "max_length": 80,
        },
        "options": {"use_cache": False, "wait_for_model": True},
    }
    try:
        data = json.dumps(payload)
        # Key not used as headers=headers not passed
        response = requests.request("POST", API_URL, data=data)
        output = json.loads(response.content.decode("utf-8"))[0][
            "generated_text"
        ]
    except:
        output = ""
    print(output)
    return output


def generate_entities_flan_alpaca_checkpoint(model, tokenizer, prompt):
    model_inputs = tokenizer(prompt, return_tensors="pt")
    input_ids = model_inputs["input_ids"]
    generation_output = model.generate(
        input_ids=input_ids,
        temperature=0.1,
        top_p=0.5,
        max_new_tokens=1024,
    )
    output = tokenizer.decode(generation_output[0], skip_special_tokens=True)
    return output