File size: 5,671 Bytes
57fbf67
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "7aaceacb",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "from app.config import DATASET_PATH, MODEL_SOURCE\n",
    "from app.utils import load_dataset, load_model_and_tokenizer, preprocess\n",
    "from scipy.special import softmax\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "7defab3e",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at cardiffnlp/twitter-roberta-base-sentiment-latest were not used when initializing RobertaForSequenceClassification: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "- This IS expected if you are initializing RobertaForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing RobertaForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
     ]
    }
   ],
   "source": [
    "tokenizer, model = load_model_and_tokenizer(MODEL_SOURCE)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "0a1dcfdd",
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = load_dataset(DATASET_PATH).shuffle()[\"test\"][:N_SAMPLES]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "501e6728",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "82b25de1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "N_BEVAL_BATCH_SIZE = 64\n",
    "N_SAMPLES = 500\n",
    "N_BATCHES = len(dataset[\"text\"])//EVAL_BATCH_SIZE\n",
    "N_BATCHES"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "7dd5371b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 64\n",
      "64 128\n",
      "128 192\n",
      "192 256\n",
      "256 320\n",
      "320 384\n",
      "384 448\n",
      "448 500\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "np.float64(0.71)"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "EVAL_BATCH_SIZE = 64\n",
    "N_SAMPLES = 500\n",
    "def evaluate_accuracy():\n",
    "\n",
    "    dataset = load_dataset(DATASET_PATH).shuffle()[\"test\"][:N_SAMPLES]\n",
    "    N_BATCHES = len(dataset[\"text\"])//EVAL_BATCH_SIZE\n",
    "\n",
    "    accuracy = 0\n",
    "    for i in range(N_BATCHES+1):\n",
    "\n",
    "        start = i*EVAL_BATCH_SIZE\n",
    "        end = min(N_SAMPLES, (i+1)*EVAL_BATCH_SIZE)\n",
    "        print(start, end)\n",
    "        samples, labels = dataset[\"text\"][start:end], dataset[\"label\"][start:end]\n",
    "        \n",
    "        model.eval()\n",
    "        encoded_batch = tokenizer(\n",
    "            [preprocess(t) for t in samples],\n",
    "            padding=True,          # pad to same length\n",
    "            truncation=True,       # truncate long texts\n",
    "            return_tensors=\"pt\",\n",
    "        )\n",
    "\n",
    "        with torch.no_grad():\n",
    "            output = model(**encoded_batch)\n",
    "    \n",
    "        logits = output[0].detach().cpu().numpy()\n",
    "        scores = softmax(logits, axis=-1)\n",
    "        pred_labels = scores.argmax(axis=-1)\n",
    "        accuracy += sum(pred_labels==labels)\n",
    "    accuracy/=N_SAMPLES\n",
    "    return accuracy\n",
    "evaluate_accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "dbd3bb8c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def _load_test_data():\n",
    "    \"\"\"\n",
    "    Expects CSV with columns: text,label\n",
    "    label values must be one of labels (negative, neutral, positive) or their indices (0,1,2).\n",
    "    \"\"\"\n",
    "    df = pd.read_csv(TEST_DATA_PATH)\n",
    "    # normalize label column to strings matching our 'labels' list\n",
    "    if np.issubdtype(df[\"label\"].dtype, np.number):\n",
    "        df[\"label\"] = df[\"label\"].astype(int).map(lambda i: labels[i])\n",
    "    else:\n",
    "        df[\"label\"] = df[\"label\"].str.lower().str.strip()\n",
    "    # keep only supported labels\n",
    "    df = df[df[\"label\"].isin(labels)].dropna(subset=[\"text\", \"label\"])\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ec0b086e",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "800c8018",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "ProjectEnv",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}