Spaces:
Sleeping
Sleeping
File size: 2,685 Bytes
f3f3f48 | 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 | from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import load_model
import pickle
import numpy as np
# model yang ingin dimuat
try:
model = load_model("models/best_model.h5", compile=True)
with open("models/tokenizer_input.pkl", 'rb') as f:
tokenizer_inputs = pickle.load(f)
with open("models/tokenizer_target.pkl", 'rb') as f:
tokenizer_outputs = pickle.load(f)
except Exception as e:
print(f"Error loading model: {e}")
raise
# kalkulasi sample temperature agar lebih hangat generatif nya
def sample_with_temperature(probs, temperature=1.0, top_k=None):
if temperature != 1.0:
probs = np.log(probs) / temperature
probs = np.exp(probs)
probs = probs / np.sum(probs)
if top_k is not None:
top_k_indices = np.argpartition(probs, -top_k)[-top_k:]
top_k_probs = probs[top_k_indices]
top_k_probs = top_k_probs / np.sum(top_k_probs) # Renormalize
sampled_index = np.random.choice(top_k_indices, p=top_k_probs)
else:
sampled_index = np.random.choice(len(probs), p=probs)
return sampled_index
# fungsi prediksi teks
def predict_with_main_model(user_text, tokenizer_input, tokenizer_target, model,
max_len=15, temperature=1.0, top_k=None, max_encoder_len=9, max_decoder_len=15):
if max_len is None:
max_len = max_decoder_len
input_seq = tokenizer_input.texts_to_sequences([user_text])
encoder_input = pad_sequences(input_seq, maxlen=max_encoder_len, padding='post')
start_token = tokenizer_target.word_index.get('<sos>', 1)
end_token = tokenizer_target.word_index.get('<eos>', 2)
decoder_input = np.zeros((1, max_len - 1), dtype='int32')
decoder_input[0, 0] = start_token
decoded_tokens = []
for i in range(1, max_len - 1):
predictions = model.predict([encoder_input, decoder_input], verbose=0)
token_probs = predictions[0, i - 1]
if top_k:
token_id = sample_with_temperature(token_probs, temperature, top_k)
else:
token_id = np.argmax(token_probs)
if token_id == end_token:
break
word = tokenizer_target.index_word.get(token_id, '')
if word and word != '<sos>':
decoded_tokens.append(word)
decoder_input[0, i] = token_id
return ' '.join(decoded_tokens)
# ada penambahan riwayat pesan (memori)
def chatbot(user_message):
response = predict_with_main_model(
user_message,
tokenizer_inputs,
tokenizer_outputs,
model,
temperature=1.0,
top_k=10
)
return response |