diff --git a/Makefile b/Makefile
index c9b6ecac..de515d32 100644
--- a/Makefile
+++ b/Makefile
@@ -1,5 +1,5 @@
-GPU=0
-OPENCV=0
+GPU=1
+OPENCV=1
 DEBUG=0
 
 ARCH= --gpu-architecture=compute_20 --gpu-code=compute_20
diff --git a/cfg/rnn.cfg b/cfg/rnn.cfg
index f57ec67c..a67e1fa1 100644
--- a/cfg/rnn.cfg
+++ b/cfg/rnn.cfg
@@ -1,11 +1,11 @@
 [net]
 subdivisions=1
 inputs=256
-batch = 1
-time_steps = 1
+batch = 128
 momentum=0.9
-decay=0.0005
-max_batches = 50000000
+decay=0.001
+max_batches = 50000
+time_steps=900
 learning_rate=0.1
 
 [rnn]
diff --git a/src/activation_kernels.cu b/src/activation_kernels.cu
index d5607daa..99933c86 100644
--- a/src/activation_kernels.cu
+++ b/src/activation_kernels.cu
@@ -9,6 +9,7 @@ extern "C" {
 
 __device__ float linear_activate_kernel(float x){return x;}
 __device__ float logistic_activate_kernel(float x){return 1./(1. + exp(-x));}
+__device__ float loggy_activate_kernel(float x){return 2./(1. + exp(-x)) - 1;}
 __device__ float relu_activate_kernel(float x){return x*(x>0);}
 __device__ float elu_activate_kernel(float x){return (x >= 0)*x + (x < 0)*(exp(x)-1);}
 __device__ float relie_activate_kernel(float x){return x*(x>0);}
@@ -24,6 +25,11 @@ __device__ float plse_activate_kernel(float x)
  
 __device__ float linear_gradient_kernel(float x){return 1;}
 __device__ float logistic_gradient_kernel(float x){return (1-x)*x;}
+__device__ float loggy_gradient_kernel(float x)
+{
+    float y = (x+1.)/2.;
+    return 2*(1-y)*y;
+}
 __device__ float relu_gradient_kernel(float x){return (x>0);}
 __device__ float elu_gradient_kernel(float x){return (x >= 0) + (x < 0)*(x + 1);}
 __device__ float relie_gradient_kernel(float x){return (x>0) ? 1 : .01;}
@@ -39,6 +45,8 @@ __device__ float activate_kernel(float x, ACTIVATION a)
             return linear_activate_kernel(x);
         case LOGISTIC:
             return logistic_activate_kernel(x);
+        case LOGGY:
+            return loggy_activate_kernel(x);
         case RELU:
             return relu_activate_kernel(x);
         case ELU:
@@ -64,6 +72,8 @@ __device__ float gradient_kernel(float x, ACTIVATION a)
             return linear_gradient_kernel(x);
         case LOGISTIC:
             return logistic_gradient_kernel(x);
+        case LOGGY:
+            return loggy_gradient_kernel(x);
         case RELU:
             return relu_gradient_kernel(x);
         case ELU:
diff --git a/src/activations.c b/src/activations.c
index 5a62ef51..07e3a459 100644
--- a/src/activations.c
+++ b/src/activations.c
@@ -10,6 +10,8 @@ char *get_activation_string(ACTIVATION a)
     switch(a){
         case LOGISTIC:
             return "logistic";
+        case LOGGY:
+            return "loggy";
         case RELU:
             return "relu";
         case ELU:
@@ -35,6 +37,7 @@ char *get_activation_string(ACTIVATION a)
 ACTIVATION get_activation(char *s)
 {
     if (strcmp(s, "logistic")==0) return LOGISTIC;
+    if (strcmp(s, "loggy")==0) return LOGGY;
     if (strcmp(s, "relu")==0) return RELU;
     if (strcmp(s, "elu")==0) return ELU;
     if (strcmp(s, "relie")==0) return RELIE;
@@ -54,6 +57,8 @@ float activate(float x, ACTIVATION a)
             return linear_activate(x);
         case LOGISTIC:
             return logistic_activate(x);
+        case LOGGY:
+            return loggy_activate(x);
         case RELU:
             return relu_activate(x);
         case ELU:
@@ -87,6 +92,8 @@ float gradient(float x, ACTIVATION a)
             return linear_gradient(x);
         case LOGISTIC:
             return logistic_gradient(x);
+        case LOGGY:
+            return loggy_gradient(x);
         case RELU:
             return relu_gradient(x);
         case ELU:
diff --git a/src/activations.h b/src/activations.h
index d824d1e7..78060258 100644
--- a/src/activations.h
+++ b/src/activations.h
@@ -4,7 +4,7 @@
 #include "math.h"
 
 typedef enum{
-    LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU
+    LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY
 }ACTIVATION;
 
 ACTIVATION get_activation(char *s);
@@ -21,6 +21,7 @@ void gradient_array_ongpu(float *x, int n, ACTIVATION a, float *delta);
 
 static inline float linear_activate(float x){return x;}
 static inline float logistic_activate(float x){return 1./(1. + exp(-x));}
+static inline float loggy_activate(float x){return 2./(1. + exp(-x)) - 1;}
 static inline float relu_activate(float x){return x*(x>0);}
 static inline float elu_activate(float x){return (x >= 0)*x + (x < 0)*(exp(x)-1);}
 static inline float relie_activate(float x){return x*(x>0);}
@@ -36,6 +37,11 @@ static inline float plse_activate(float x)
 
 static inline float linear_gradient(float x){return 1;}
 static inline float logistic_gradient(float x){return (1-x)*x;}
+static inline float loggy_gradient(float x)
+{
+    float y = (x+1.)/2.;
+    return 2*(1-y)*y;
+}
 static inline float relu_gradient(float x){return (x>0);}
 static inline float elu_gradient(float x){return (x >= 0) + (x < 0)*(x + 1);}
 static inline float relie_gradient(float x){return (x>0) ? 1 : .01;}
diff --git a/src/darknet.c b/src/darknet.c
index 938609ea..c4006cee 100644
--- a/src/darknet.c
+++ b/src/darknet.c
@@ -206,7 +206,6 @@ int main(int argc, char **argv)
     gpu_index = find_int_arg(argc, argv, "-i", 0);
     if(find_arg(argc, argv, "-nogpu")) {
         gpu_index = -1;
-        printf("nogpu\n");
     }
 
 #ifndef GPU
diff --git a/src/layer.h b/src/layer.h
index fc76234f..91042a21 100644
--- a/src/layer.h
+++ b/src/layer.h
@@ -34,6 +34,7 @@ struct layer{
     ACTIVATION activation;
     COST_TYPE cost_type;
     int batch_normalize;
+    int shortcut;
     int batch;
     int forced;
     int flipped;
diff --git a/src/nightmare.c b/src/nightmare.c
index bc2060f2..2b1c76cd 100644
--- a/src/nightmare.c
+++ b/src/nightmare.c
@@ -13,7 +13,7 @@ float abs_mean(float *x, int n)
     int i;
     float sum = 0;
     for (i = 0; i < n; ++i){
-        sum += abs(x[i]);
+        sum += fabs(x[i]);
     }
     return sum/n;
 }
diff --git a/src/parser.c b/src/parser.c
index a48f207c..68a9f049 100644
--- a/src/parser.c
+++ b/src/parser.c
@@ -176,8 +176,11 @@ layer parse_rnn(list *options, size_params params)
     char *activation_s = option_find_str(options, "activation", "logistic");
     ACTIVATION activation = get_activation(activation_s);
     int batch_normalize = option_find_int_quiet(options, "batch_normalize", 0);
+    int logistic = option_find_int_quiet(options, "logistic", 0);
 
-    layer l = make_rnn_layer(params.batch, params.inputs, hidden, output, params.time_steps, activation, batch_normalize);
+    layer l = make_rnn_layer(params.batch, params.inputs, hidden, output, params.time_steps, activation, batch_normalize, logistic);
+
+    l.shortcut = option_find_int_quiet(options, "shortcut", 0);
 
     return l;
 }
diff --git a/src/rnn.c b/src/rnn.c
index d3e7e51f..aee53ffb 100644
--- a/src/rnn.c
+++ b/src/rnn.c
@@ -19,6 +19,12 @@ float_pair get_rnn_data(char *text, int len, int batch, int steps)
     int i,j;
     for(i = 0; i < batch; ++i){
         int index = rand() %(len - steps - 1);
+        int done = 1;
+        while(!done){
+            index = rand() %(len - steps - 1);
+            while(index < len-steps-1 && text[index++] != '\n');
+            if (index < len-steps-1) done = 1;
+        }
         for(j = 0; j < steps; ++j){
             x[(j*batch + i)*256 + text[index + j]] = 1;
             y[(j*batch + i)*256 + text[index + j + 1]] = 1;
@@ -48,13 +54,13 @@ void train_char_rnn(char *cfgfile, char *weightfile, char *filename)
     srand(time(0));
     data_seed = time(0);
     char *base = basecfg(cfgfile);
-    printf("%s\n", base);
+    fprintf(stderr, "%s\n", base);
     float avg_loss = -1;
     network net = parse_network_cfg(cfgfile);
     if(weightfile){
         load_weights(&net, weightfile);
     }
-    printf("Learning Rate: %g, Momentum: %g, Decay: %g\n", net.learning_rate, net.momentum, net.decay);
+    fprintf(stderr, "Learning Rate: %g, Momentum: %g, Decay: %g\n", net.learning_rate, net.momentum, net.decay);
     int batch = net.batch;
     int steps = net.time_steps;
     int i = (*net.seen)/net.batch;
@@ -71,7 +77,7 @@ void train_char_rnn(char *cfgfile, char *weightfile, char *filename)
         if (avg_loss < 0) avg_loss = loss;
         avg_loss = avg_loss*.9 + loss*.1;
 
-        printf("%d: %f, %f avg, %f rate, %lf seconds\n", i, loss, avg_loss, get_current_rate(net), sec(clock()-time));
+        fprintf(stderr, "%d: %f, %f avg, %f rate, %lf seconds\n", i, loss, avg_loss, get_current_rate(net), sec(clock()-time));
         if(i%100==0){
             char buff[256];
             sprintf(buff, "%s/%s_%d.weights", backup_directory, base, i);
@@ -92,7 +98,7 @@ void test_char_rnn(char *cfgfile, char *weightfile, int num, char *seed, float t
 {
     srand(rseed);
     char *base = basecfg(cfgfile);
-    printf("%s\n", base);
+    fprintf(stderr, "%s\n", base);
 
     network net = parse_network_cfg(cfgfile);
     if(weightfile){
@@ -128,6 +134,43 @@ void test_char_rnn(char *cfgfile, char *weightfile, int num, char *seed, float t
     printf("\n");
 }
 
+void valid_char_rnn(char *cfgfile, char *weightfile, char *filename)
+{
+    FILE *fp = fopen(filename, "r");
+    //FILE *fp = fopen("data/ab.txt", "r");
+    //FILE *fp = fopen("data/grrm/asoiaf.txt", "r");
+
+    fseek(fp, 0, SEEK_END); 
+    size_t size = ftell(fp);
+    fseek(fp, 0, SEEK_SET); 
+
+    char *text = calloc(size, sizeof(char));
+    fread(text, 1, size, fp);
+    fclose(fp);
+
+    char *base = basecfg(cfgfile);
+    fprintf(stderr, "%s\n", base);
+
+    network net = parse_network_cfg(cfgfile);
+    if(weightfile){
+        load_weights(&net, weightfile);
+    }
+    
+    int i;
+    char c;
+    float *input = calloc(256, sizeof(float));
+    float sum = 0;
+    for(i = 0; i < size-1; ++i){
+        c = text[i];
+        input[(int)c] = 1;
+        float *out = network_predict(net, input);
+        input[(int)c] = 0;
+        sum += log(out[(int)text[i+1]]);
+    }
+    printf("Log Probability: %f\n", sum);
+}
+
+
 void run_char_rnn(int argc, char **argv)
 {
     if(argc < 4){
@@ -143,5 +186,6 @@ void run_char_rnn(int argc, char **argv)
     char *cfg = argv[3];
     char *weights = (argc > 4) ? argv[4] : 0;
     if(0==strcmp(argv[2], "train")) train_char_rnn(cfg, weights, filename);
+    else if(0==strcmp(argv[2], "valid")) valid_char_rnn(cfg, weights, filename);
     else if(0==strcmp(argv[2], "test")) test_char_rnn(cfg, weights, len, seed, temp, rseed);
 }
diff --git a/src/rnn_layer.c b/src/rnn_layer.c
index 63582858..e58e0a4b 100644
--- a/src/rnn_layer.c
+++ b/src/rnn_layer.c
@@ -11,9 +11,9 @@
 #include <string.h>
 
 
-layer make_rnn_layer(int batch, int inputs, int hidden, int outputs, int steps, ACTIVATION activation, int batch_normalize)
+layer make_rnn_layer(int batch, int inputs, int hidden, int outputs, int steps, ACTIVATION activation, int batch_normalize, int log)
 {
-    printf("%d %d\n", batch, steps);
+    fprintf(stderr, "RNN Layer: %d inputs, %d outputs\n", inputs, outputs);
     batch = batch / steps;
     layer l = {0};
     l.batch = batch;
@@ -25,14 +25,17 @@ layer make_rnn_layer(int batch, int inputs, int hidden, int outputs, int steps,
     l.state = calloc(batch*hidden, sizeof(float));
 
     l.input_layer = malloc(sizeof(layer));
+    fprintf(stderr, "\t\t");
     *(l.input_layer) = make_connected_layer(batch*steps, inputs, hidden, activation, batch_normalize);
     l.input_layer->batch = batch;
 
     l.self_layer = malloc(sizeof(layer));
-    *(l.self_layer) = make_connected_layer(batch*steps, hidden, hidden, activation, batch_normalize);
+    fprintf(stderr, "\t\t");
+    *(l.self_layer) = make_connected_layer(batch*steps, hidden, hidden, (log==2)?LOGGY:(log==1?LOGISTIC:activation), batch_normalize);
     l.self_layer->batch = batch;
 
     l.output_layer = malloc(sizeof(layer));
+    fprintf(stderr, "\t\t");
     *(l.output_layer) = make_connected_layer(batch*steps, hidden, outputs, activation, batch_normalize);
     l.output_layer->batch = batch;
 
@@ -46,7 +49,6 @@ layer make_rnn_layer(int batch, int inputs, int hidden, int outputs, int steps,
     l.delta_gpu = l.output_layer->delta_gpu;
     #endif
 
-    fprintf(stderr, "RNN Layer: %d inputs, %d outputs\n", inputs, outputs);
     return l;
 }
 
diff --git a/src/rnn_layer.h b/src/rnn_layer.h
index 8d4f5854..00dc1be9 100644
--- a/src/rnn_layer.h
+++ b/src/rnn_layer.h
@@ -6,7 +6,7 @@
 #include "layer.h"
 #include "network.h"
 
-layer make_rnn_layer(int batch, int inputs, int hidden, int outputs, int steps, ACTIVATION activation, int batch_normalize);
+layer make_rnn_layer(int batch, int inputs, int hidden, int outputs, int steps, ACTIVATION activation, int batch_normalize, int log);
 
 void forward_rnn_layer(layer l, network_state state);
 void backward_rnn_layer(layer l, network_state state);