From 5d8a1313d621491a2bd6743740ef5385d9a6d89d Mon Sep 17 00:00:00 2001
From: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date: Wed, 11 Jul 2018 01:30:30 -0400
Subject: [PATCH] decodes something...

---
 dnn/lpcnet.py      |  7 ++++---
 dnn/test_lpcnet.py | 22 +++++++++++++++-------
 2 files changed, 19 insertions(+), 10 deletions(-)

diff --git a/dnn/lpcnet.py b/dnn/lpcnet.py
index 21858815..dcb7f3dd 100644
--- a/dnn/lpcnet.py
+++ b/dnn/lpcnet.py
@@ -9,7 +9,7 @@ import numpy as np
 import h5py
 import sys
 
-rnn_units=64
+rnn_units=512
 pcm_bits = 8
 pcm_levels = 2**pcm_bits
 nb_used_features = 37
@@ -20,6 +20,7 @@ def new_wavernn_model():
     pitch = Input(shape=(None, 1))
     feat = Input(shape=(None, nb_used_features))
     dec_feat = Input(shape=(None, 32))
+    dec_state = Input(shape=(rnn_units,))
 
     conv1 = Conv1D(16, 7, padding='causal')
     pconv1 = Conv1D(16, 5, padding='same')
@@ -48,8 +49,8 @@ def new_wavernn_model():
     encoder = Model(feat, cfeat)
     
     dec_rnn_in = Concatenate()([cpcm, cpitch, dec_feat])
-    dec_gru_out, state = rnn(dec_rnn_in)
+    dec_gru_out, state = rnn(dec_rnn_in, initial_state=dec_state)
     dec_ulaw_prob = md(dec_gru_out)
 
-    decoder = Model([pcm, pitch, dec_feat], [dec_ulaw_prob, state])
+    decoder = Model([pcm, pitch, dec_feat, dec_state], [dec_ulaw_prob, state])
     return model, encoder, decoder
diff --git a/dnn/test_lpcnet.py b/dnn/test_lpcnet.py
index 185893be..4c47274e 100755
--- a/dnn/test_lpcnet.py
+++ b/dnn/test_lpcnet.py
@@ -21,7 +21,7 @@ batch_size = 64
 
 model, enc, dec = lpcnet.new_wavernn_model()
 model.compile(optimizer=Adadiff(), loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy'])
-model.summary()
+#model.summary()
 
 pcmfile = sys.argv[1]
 feature_file = sys.argv[2]
@@ -47,14 +47,15 @@ in_data = np.reshape(in_data, (nb_frames*pcm_chunk_size, 1))
 out_data = np.reshape(data, (nb_frames*pcm_chunk_size, 1))
 
 
-model.load_weights('lpcnet1h_30.h5')
+model.load_weights('lpcnet1i_30.h5')
 
 order = 16
 
 pcm = 0.*out_data
-exc = 0.*out_data
+exc = out_data-0
 pitch = np.zeros((1, 1, 1), dtype='float32')
-iexc = np.zeros((1, 1, 1), dtype='float32')
+fexc = np.zeros((1, 1, 1), dtype='float32')
+iexc = np.zeros((1, 1, 1), dtype='int16')
 state = np.zeros((1, lpcnet.rnn_units), dtype='float32')
 for c in range(1, nb_frames):
     cfeat = enc.predict(features[c:c+1, :, :nb_used_features])
@@ -68,7 +69,14 @@ for c in range(1, nb_frames):
         period = period - 4
         for i in range(frame_size):
             pitch[0, 0, 0] = exc[f*frame_size + i - period, 0]
-            #p, state = dec.predict([
-            pcm[f*frame_size + i, 0] = gain*out_data[f*frame_size + i, 0] - sum(a*pcm[f*frame_size + i - 1:f*frame_size + i - order-1:-1, 0])
-            print(pcm[f*frame_size + i, 0])
+            fexc[0, 0, 0] = exc[f*frame_size + i - 1]
+            #print(cfeat.shape)
+            p, state = dec.predict([fexc, pitch, cfeat[:, fr:fr+1, :], state])
+            p = p/(1e-5 + np.sum(p))
+            #print(np.sum(p))
+            iexc[0, 0, 0] = np.argmax(np.random.multinomial(1, p[0,0,:], 1))-128
+            exc[f*frame_size + i] = iexc[0, 0, 0]/16.
+            #out_data[f*frame_size + i, 0] = iexc[0, 0, 0]
+            pcm[f*frame_size + i, 0] = gain*iexc[0, 0, 0] - sum(a*pcm[f*frame_size + i - 1:f*frame_size + i - order-1:-1, 0])
+            print(iexc[0, 0, 0], out_data[f*frame_size + i, 0], pcm[f*frame_size + i, 0])