examples: rnn: fix incorrect data type in int8 example

saininav · Feb 7, 2019 · 4661796 · 4661796
1 parent b5f1f92
commit 4661796
Showing 1 changed file with 6 additions and 3 deletions.
diff --git a/examples/simple_rnn_int8.cpp b/examples/simple_rnn_int8.cpp
@@ -79,7 +79,7 @@ void compute_attention(float *context_vectors, int src_seq_length_max,
         int batch, int feature_size, int8_t *weights_src_layer,
         float weights_src_layer_scale, int32_t *compensation,
         uint8_t *dec_src_layer, float dec_src_layer_scale,
-        float dec_src_layer_shift, float *annotations,
+        float dec_src_layer_shift, uint8_t *annotations,
         float *weighted_annotations, float *weights_alignments) {
     // dst_iter : (n, c) matrix
     // src_layer: (n, c) matrix
@@ -168,7 +168,10 @@ void compute_attention(float *context_vectors, int src_seq_length_max,
                 context_vectors[i * (feature_size + feature_size) + feature_size
                         + j]
                         += alignments[k * batch + i]
-                        * annotations[j + feature_size * (i + batch * k)];
+                        * (((float)annotations[j
+                                   + feature_size * (i + batch * k)]
+                                   - dec_src_layer_shift)
+                        / dec_src_layer_scale);
 }
 
 void copy_context(float *src_iter, int n_layers, int n_states, int batch,
@@ -670,7 +673,7 @@ void simple_net() {
                     feature_size, user_weights_attention_src_layer.data(),
                     weights_attention_scale, weights_attention_sum_rows.data(),
                     src_att_layer_handle, data_scale, data_shift,
-                    (float *)enc_bidir_dst_layer_memory.get_data_handle(),
+                    (uint8_t *)enc_bidir_dst_layer_memory.get_data_handle(),
                     weighted_annotations.data(),
                     user_weights_alignments.data());