Bot for telegram

src/bot.py

@@ -0,0 +1,62 @@
+#!/usr/bin/env python3
+
+from telegram.ext import Updater
+from telegram.ext import CommandHandler
+from telegram.ext import MessageHandler, Filters
+import logging
+
+exec(open("dialrequirements.py").read())
+
+
+def start(update, context):
+    context.bot.send_message(
+        chat_id=update.effective_chat.id, text="Ongi etorri elkarrizketara."
+    )
+
+
+def echo(update, context):
+    context.bot.send_message(
+        chat_id=update.effective_chat.id, text=update.message.text
+    )
+
+
+def reply(update, context):
+    user = update.message.text
+    if user == "Agur":
+        context.bot.send_message(
+            chat_id=update.effective_chat.id, text="Agur"
+        )
+        updater.stop()
+    else:
+        sentence = evaluate(" ".join(tokenizer(user)))
+        context.bot.send_message(
+            chat_id=update.effective_chat.id, text=sentence.strip().capitalize()
+        )
+
+updater = Updater(
+    token="1213404233:AAHYV6hxaQVnypuhJytZPPtETufMZKkKS3Y", use_context=True
+)
+
+dispatcher = updater.dispatcher
+
+logging.basicConfig(
+    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+    level=logging.INFO,
+)
+
+def main():
+
+
+
+    start_handler = CommandHandler("start", start)
+    dispatcher.add_handler(start_handler)
+
+    reply_handler = MessageHandler(Filters.text & (~Filters.command), reply)
+    dispatcher.add_handler(reply_handler)
+
+    print("Start")
+    updater.start_polling()
+
+
+if __name__ == "__main__":
+    main()

src/dialrequirements.py

@@ -0,0 +1,75 @@
+#!/usr/bin/env python3
+
+import torch
+import random
+from argparse import ArgumentParser
+from model import *
+from spacy.lang.eu import Basque
+
+nlp = Basque()
+
+
+def tokenizer(s):
+    return list(map(lambda x: x.text, nlp(s)))
+
+
+parser = ArgumentParser(
+    description="Azpitituluetan oinarritutako elkarrizketa \
+sistemaren proba"
+)
+
+parser.add_argument(
+    "-decoding_strategy",
+    type=str,
+    default="top1",
+    choices=["top1", "topk", "multinomial"],
+)
+
+args = parser.parse_args()
+
+
+def decode(logits, decoding_strategy="max", k=3, temp=0.4):
+    if decoding_strategy == "top1":
+        target = logits.max(1)[1]
+    elif decoding_strategy == "topk":
+        target = logits.topk(k)[1][0][random.randint(0, k - 1)].unsqueeze(-1)
+    else:
+        target = torch.multinomial(logits.squeeze().div(temp).exp().cpu(), 1)
+    return target
+
+
+def evaluate(sentence):
+    with torch.no_grad():
+        sentence = "<sos> " + sentence + " <eos>"
+        sent_len = len(sentence.split())
+        sentence = (
+            torch.Tensor(
+                [text_field.vocab.stoi[i] for i in sentence.lower().split()]
+            )
+            .long()
+            .view(sent_len, 1)
+        )
+        target = torch.Tensor([text_field.vocab.stoi["<sos>"]]).long()
+        output_sentence = ""
+        encoder_outputs, hidden = model.encoder(sentence)
+        for t in range(MAX_LENGTH):
+            # first input to the decoder is the <sos> token
+            output, hidden = model.decoder(target, hidden, encoder_outputs)
+            target = decode(output, decoding_strategy)
+            word = text_field.vocab.itos[target.numpy()[0]]
+            if word == "<eos>":
+                return output_sentence
+            else:
+                output_sentence = output_sentence + " " + word
+    return output_sentence
+
+
+# Load model and fields
+text_field = torch.load("../model/text_field.Field")
+model = torch.load("../model/model.pt", map_location=torch.device("cpu"))
+torch.nn.Module.dump_patches = True
+MAX_LENGTH = 10
+
+
+model.eval()
+decoding_strategy = args.decoding_strategy