File size: 3,607 Bytes
f49f7d2
7dbb743
 
 
2c7a30f
7dbb743
 
f8828b6
dd5526c
 
a33168c
f49f7d2
7dbb743
 
 
f8828b6
 
 
 
 
 
f49f7d2
f8828b6
f49f7d2
a4ac5c1
f49f7d2
 
 
 
 
 
 
 
7dbb743
f49f7d2
 
 
 
 
 
 
 
7dbb743
 
 
 
f49f7d2
 
 
 
 
 
 
7dbb743
7438d14
7dbb743
 
 
 
 
 
f8828b6
 
 
7dbb743
f8828b6
 
 
 
 
 
 
 
f49f7d2
e42e6ee
f8828b6
e42e6ee
7dbb743
e007ca3
7dbb743
 
f49f7d2
7dbb743
 
 
 
 
 
f49f7d2
7dbb743
 
7438d14
7dbb743
f49f7d2
 
7dbb743
 
f8828b6
7dbb743
f49f7d2
7dbb743
e007ca3
7dbb743
 
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
from flask import Flask, render_template, request
from joblib import load
import pandas as pd
import re
from customFunctions import *
import json
import datetime
import numpy as np
from huggingface_hub import hf_hub_download
import torch
import os
import logging

pd.set_option('display.max_colwidth', 1000)

# Patch torch.load to always load on CPU
original_torch_load = torch.load
def cpu_load(*args, **kwargs):
    return original_torch_load(*args, map_location=torch.device('cpu'), **kwargs)
torch.load = cpu_load

app = Flask(__name__)

# Logging setup
LOG_DIR = "/tmp/logs"  # Use a universally writable directory
LOG_FILE = os.path.join(LOG_DIR, "usage_log.jsonl")
os.makedirs(LOG_DIR, exist_ok=True)
logging.basicConfig(
    filename=LOG_FILE,
    level=logging.INFO,
    format='%(asctime)s [%(levelname)s] %(message)s'
)

PIPELINES = [
    {'id': 8, 'name': 'Embedded using BioWordVec', 'filename': "pipeline_ex3_s4.joblib"},
    {'id': 1, 'name': 'Baseline', 'filename': "pipeline_ex1_s1.joblib"},
    {'id': 2, 'name': 'Trained on a FeedForward NN', 'filename': "pipeline_ex1_s2.joblib"},
    {'id': 3, 'name': 'Trained on a CRF', 'filename': "pipeline_ex1_s3.joblib"},
    {'id': 4, 'name': 'Trained on a small dataset', 'filename': "pipeline_ex2_s3.joblib"},
    {'id': 5, 'name': 'Trained on a large dataset', 'filename': "pipeline_ex2_s2.joblib"},
    {'id': 6, 'name': 'Embedded using TFIDF', 'filename': "pipeline_ex3_s2.joblib"},
    {'id': 7, 'name': 'Embedded using GloVe', 'filename': "pipeline_ex3_s3.joblib"},
]

pipeline_metadata = [{'id': p['id'], 'name': p['name']} for p in PIPELINES]

def load_pipeline_from_hub(filename):
    cache_dir = "/tmp/hf_cache"
    os.environ["HF_HUB_CACHE"] = cache_dir
    repo_id = 'hw01558/nlp-coursework-pipelines'
    local_path = hf_hub_download(repo_id=repo_id, filename=filename, cache_dir=cache_dir)
    return load(local_path)

def get_pipeline_by_id(pipelines, pipeline_id):
    return next((p['filename'] for p in pipelines if p['id'] == pipeline_id), None)

def get_name_by_id(pipelines, pipeline_id):
    return next((p['name'] for p in pipelines if p['id'] == pipeline_id), None)

def requestResults(text, pipeline):
    labels = pipeline.predict(text)
    if isinstance(labels, np.ndarray):
        labels = labels.tolist()
    return labels[0]

def log_interaction(user_input, model_name, predictions):
    log_entry = {
        "timestamp": datetime.datetime.utcnow().isoformat(),
        "model": model_name,
        "user_input": user_input,
        "predictions": predictions
    }
    try:
        logging.info(json.dumps(log_entry))
        print("[INFO] Logged interaction successfully.")
    except Exception as e:
        print(f"[ERROR] Could not write log entry: {e}")


@app.route('/')
def index():
    return render_template('index.html', pipelines=pipeline_metadata)

@app.route('/', methods=['POST'])
def get_data():
    if request.method == 'POST':
        text = request.form['search']
        tokens = re.findall(r"\w+|[^\w\s]", text)
        tokens_formatted = pd.Series([pd.Series(tokens)])

        pipeline_id = int(request.form['pipeline_select'])
        pipeline = load_pipeline_from_hub(get_pipeline_by_id(PIPELINES, pipeline_id))
        name = get_name_by_id(PIPELINES, pipeline_id)

        labels = requestResults(tokens_formatted, pipeline)
        results = dict(zip(tokens, labels))

        log_interaction(text, name, results)

        return render_template('index.html', results=results, name=name, pipelines=pipeline_metadata)


if __name__ == '__main__':
    app.run(host="0.0.0.0", port=7860)