Upload 6 files

sentiment_analyzer/api.py

@@ -0,0 +1,30 @@
+from typing import Dict
+
+from fastapi import Depends, FastAPI
+from pydantic import BaseModel
+
+from .classifier.model import Model, get_model
+
+app = FastAPI()
+
+
+class SentimentRequest(BaseModel):
+    text: str
+
+class SentimentResponse(BaseModel):
+    probabilities: Dict[str, float]
+    sentiment: str
+    confidence: float
+
+
+@app.post("/predict", response_model=SentimentResponse)
+def predict(request: SentimentRequest, model: Model = Depends(get_model)):
+    sentiment, confidence, probabilities = model.predict(request.text)
+    return SentimentResponse(
+        sentiment=sentiment, confidence=confidence, probabilities=probabilities
+    )
+
+
+@app.get("/")
+def read_root():
+    return {"TrueFoundry": "Project"}

sentiment_analyzer/classifier/model.py

@@ -0,0 +1,56 @@
+import json
+
+import torch
+import torch.nn.functional as F
+from transformers import BertTokenizer
+
+from .sentiment_classifier import SentimentClassifier
+
+with open("config.json") as json_file:
+    config = json.load(json_file)
+
+
+class Model:
+    def __init__(self):
+
+        self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+
+        self.tokenizer = BertTokenizer.from_pretrained(config["BERT_MODEL"])
+
+        classifier = SentimentClassifier(len(config["CLASS_NAMES"]))
+        classifier.load_state_dict(
+            torch.load(config["PRE_TRAINED_MODEL"], map_location=self.device)
+        )
+        classifier = classifier.eval()
+        self.classifier = classifier.to(self.device)
+
+    def predict(self, text):
+        encoded_text = self.tokenizer.encode_plus(
+            text,
+            max_length=config["MAX_SEQUENCE_LEN"],
+            add_special_tokens=True,
+            return_token_type_ids=False,
+            pad_to_max_length=True,
+            return_attention_mask=True,
+            return_tensors="pt",
+        )
+        input_ids = encoded_text["input_ids"].to(self.device)
+        attention_mask = encoded_text["attention_mask"].to(self.device)
+
+        with torch.no_grad():
+            probabilities = F.softmax(self.classifier(input_ids, attention_mask), dim=1)
+        confidence, predicted_class = torch.max(probabilities, dim=1)
+        predicted_class = predicted_class.cpu().item()
+        probabilities = probabilities.flatten().cpu().numpy().tolist()
+        return (
+            config["CLASS_NAMES"][predicted_class],
+            confidence,
+            dict(zip(config["CLASS_NAMES"], probabilities)),
+        )
+
+
+model = Model()
+
+
+def get_model():
+    return model

sentiment_analyzer/classifier/sentiment_classifier.py

@@ -0,0 +1,20 @@
+import json
+
+from torch import nn
+from transformers import BertModel
+
+with open("config.json") as json_file:
+    config = json.load(json_file)
+
+
+class SentimentClassifier(nn.Module):
+    def __init__(self, n_classes):
+        super(SentimentClassifier, self).__init__()
+        self.bert = BertModel.from_pretrained(config["BERT_MODEL"])
+        self.drop = nn.Dropout(p=0.3)
+        self.out = nn.Linear(self.bert.config.hidden_size, n_classes)
+
+    def forward(self, input_ids, attention_mask):
+        _, pooled_output = self.bert(input_ids=input_ids, attention_mask=attention_mask, return_dict=False)
+        output = self.drop(pooled_output)
+        return self.out(output)