| import json # TODO: use the responses file after it's done instead of loading the dataset again | |
| import numpy as np | |
| from keras.saving import load_model | |
| from keras_self_attention import SeqSelfAttention | |
| from vecs import * | |
| from model_settings import * | |
| with open("dataset.json", "r") as f: | |
| dset = json.load(f) | |
| tokenizer = Tokenizer() # a tokenizer is a thing to split text into words, it might have some other stuff like making all the letters lowercase, etc. | |
| tokenizer.fit_on_texts(list(dset.keys())) | |
| model = load_model("chatbot.keras", custom_objects={"SeqSelfAttention": SeqSelfAttention}) | |
| def find_line_number(array): | |
| return sorted(zip(list(array), [x for x in range(len(array))]), key=lambda x:x[0], reverse=True)[0][1] # yeah, one big line, find the biggest value and return the number of the line | |
| def generate(text): | |
| tokens = list(tokenizer.texts_to_sequences([text,])[0]) # text into tokens (almost words) | |
| tokens = (tokens+[0,]*inp_len)[:inp_len] # cutting off the sentence after inp_len words | |
| prediction = model.predict(np.array([tokens,]))[0] | |
| line = find_line_number(prediction) | |
| return list(dset.values())[line] | |
| if __name__ == "__main__": # if this code is not being imported, open the chat | |
| while True: | |
| inp = input("User: ") | |
| print(f"Bot: {generate(inp)}") | |