visualization complete

Author: culurciello

seq-learning-char/generate.py

@@ -1,12 +1,18 @@
 # https://github.com/spro/practical-pytorch
 
 import torch
-import colorama
-from colorama import Fore, Style
+import os
 
 from helpers import *
 from model import *
 
+# Color Palette
+CP_R = '\033[31m'
+CP_G = '\033[32m'
+CP_B = '\033[34m'
+CP_Y = '\033[33m'
+CP_C = '\033[0m'
+
 def generate_GRU(decoder, prime_str='A', predict_len=100, temperature=0.8):
     hidden = decoder.init_hidden()
     prime_input = char_tensor(prime_str)
@@ -29,9 +35,8 @@ def generate_GRU(decoder, prime_str='A', predict_len=100, temperature=0.8):
         predicted_char = all_characters[top_i]
         predicted += predicted_char
         inp = char_tensor(predicted_char)
-    
-    print(Fore.BLUE + predicted[:len(prime_str)] + Fore.GREEN + predicted[len(prime_str):])
-    print(Style.RESET_ALL)
+
+    print(CP_B + predicted[:len(prime_str)] + CP_G + predicted[len(prime_str):] + CP_C)
     # return predicted
 
 
@@ -63,6 +68,5 @@ def generate_CNN(decoder, prime_str, predict_len=100, temperature=0.8):
         inp[:-1] = inp[1:]
         inp[-1] = top_i
 
-    print(Fore.BLUE + predicted[:len(prime_str)] + Fore.GREEN + predicted[len(prime_str):])
-    print(Style.RESET_ALL)
+    print(CP_B + predicted[:len(prime_str)] + CP_G + predicted[len(prime_str):] + CP_C)
     # return predicted

seq-learning-char/train.py

@@ -161,14 +161,15 @@ def save():
         loss_avg += loss
 
         # generate some text to see performance:
+        test_len = 100
         if epoch % args.print_every == 0:
             print('[elapsed time: %s, epoch: %d,  percent complete: %d%%, loss: %.4f]' % (time_since(start), epoch, epoch / args.epochs * 100, loss))
             if args.sequencer == 'GRU':
-                init_str = random_piece_text(args.pint)
-                generate_GRU(model, init_str, 100)
+                init_str = random_piece_text(args.pint+test_len)
+                generate_GRU(model, init_str[:args.pint], test_len)
             elif args.sequencer == 'CNN' or 'Att':
-                init_str,_ = random_training_set(args.pint)
-                generate_CNN(model, init_str, 100)
+                init_str,_ = random_training_set(args.pint+test_len)
+                generate_CNN(model, init_str[:args.pint], test_len)
 
     print("Saving...")
     save()