idk man 🐍 stuff

Makefile

@@ -29,7 +29,7 @@ COMMON= -Iinclude/ -Isrc/
 CFLAGS=-Wall -Wno-unknown-pragmas -Wfatal-errors -fPIC
 
 ifeq ($(OPENMP), 1) 
-COMMON+= -fopenmp
+CFLAGS+= -fopenmp
 endif
 
 ifeq ($(DEBUG), 1) 
@@ -68,7 +68,8 @@ EXECOBJ = $(addprefix $(OBJDIR), $(EXECOBJA))
 OBJS = $(addprefix $(OBJDIR), $(OBJ))
 DEPS = $(wildcard src/*.h) Makefile include/darknet.h
 
-all: obj backup results $(SLIB) $(ALIB) $(EXEC)
+#all: obj backup results $(SLIB) $(ALIB) $(EXEC)
+all: obj  results $(SLIB) $(ALIB) $(EXEC)
 
 
 $(EXEC): $(EXECOBJ) $(ALIB)

examples/darknet.c

@@ -146,7 +146,7 @@ void oneoff(char *cfgfile, char *weightfile, char *outfile)
     int c = net.layers[net.n - 2].c;
     scal_cpu(oldn*c, .1, net.layers[net.n - 2].weights, 1);
     scal_cpu(oldn, 0, net.layers[net.n - 2].biases, 1);
-    net.layers[net.n - 2].n = 9418;
+    net.layers[net.n - 2].n = 11921;
     net.layers[net.n - 2].biases += 5;
     net.layers[net.n - 2].weights += 5*c;
     if(weightfile){

examples/rnn.c

@@ -108,22 +108,6 @@ float_pair get_rnn_data(unsigned char *text, size_t *offsets, int characters, si
     return p;
 }
 
-void reset_rnn_state(network net, int b)
-{
-    int i;
-    for (i = 0; i < net.n; ++i) {
-        #ifdef GPU
-        layer l = net.layers[i];
-        if(l.state_gpu){
-            fill_gpu(l.outputs, 0, l.state_gpu + l.outputs*b, 1);
-        }
-        if(l.h_gpu){
-            fill_gpu(l.outputs, 0, l.h_gpu + l.outputs*b, 1);
-        }
-        #endif
-    }
-}
-
 void train_char_rnn(char *cfgfile, char *weightfile, char *filename, int clear, int tokenized)
 {
     srand(time(0));
@@ -194,7 +178,7 @@ void train_char_rnn(char *cfgfile, char *weightfile, char *filename, int clear,
             if(rand()%64 == 0){
                 //fprintf(stderr, "Reset\n");
                 offsets[j] = rand_size_t()%size;
-                reset_rnn_state(net, j);
+                reset_network_state(net, j);
             }
         }
 
@@ -304,7 +288,7 @@ void test_tactic_rnn_multi(char *cfgfile, char *weightfile, int num, float temp,
     float *out = 0;
 
     while(1){
-        reset_rnn_state(net, 0);
+        reset_network_state(net, 0);
         while((c = getc(stdin)) != EOF && c != 0){
             input[c] = 1;
             out = network_predict(net, input);
@@ -482,7 +466,7 @@ void vec_char_rnn(char *cfgfile, char *weightfile, char *seed)
     int i;
     char *line;
     while((line=fgetl(stdin)) != 0){
-        reset_rnn_state(net, 0);
+        reset_network_state(net, 0);
         for(i = 0; i < seed_len; ++i){
             c = seed[i];
             input[(int)c] = 1;

include/darknet.h

@@ -647,6 +647,7 @@ void zero_objectness(layer l);
 void get_region_boxes(layer l, int w, int h, int netw, int neth, float thresh, float **probs, box *boxes, float **masks, int only_objectness, int *map, float tree_thresh, int relative);
 void free_network(network net);
 void set_batch_network(network *net, int b);
+void set_temp_network(network net, float t);
 image load_image(char *filename, int w, int h, int c);
 image load_image_color(char *filename, int w, int h);
 image make_image(int w, int h, int c);
@@ -699,6 +700,9 @@ int network_width(network *net);
 int network_height(network *net);
 float *network_predict_image(network *net, image im);
 
+void reset_network_state(network net, int b);
+void reset_network_state(network net, int b);
+
 char **get_labels(char *filename);
 void do_nms_sort(box *boxes, float **probs, int total, int classes, float thresh);
 void do_nms_obj(box *boxes, float **probs, int total, int classes, float thresh);

python/darknet.py

@@ -1,4 +1,19 @@
 from ctypes import *
+import math
+import random
+
+def sample(probs):
+    s = sum(probs)
+    probs = [a/s for a in probs]
+    r = random.uniform(0, 1)
+    for i in range(len(probs)):
+        r = r - probs[i]
+        if r <= 0:
+            return i
+    return len(probs)-1
+
+def c_array(ctype, values):
+    return (ctype * len(values))(*values)
 
 class IMAGE(Structure):
     _fields_ = [("w", c_int),
@@ -10,43 +25,42 @@ class METADATA(Structure):
     _fields_ = [("classes", c_int),
                 ("names", POINTER(c_char_p))]
 
-lib = CDLL("/home/pjreddie/documents/darknet/libdarknet.so", RTLD_GLOBAL)
+#lib = CDLL("/home/pjreddie/documents/darknet/libdarknet.so", RTLD_GLOBAL)
+lib = CDLL("libdarknet.so", RTLD_GLOBAL)
 lib.network_width.argtypes = [c_void_p]
 lib.network_width.restype = c_int
 lib.network_height.argtypes = [c_void_p]
 lib.network_height.restype = c_int
 
-def load_meta(f):
+predict = lib.network_predict_p
-    lib.get_metadata.argtypes = [c_char_p]
+predict.argtypes = [c_void_p, POINTER(c_float)]
-    lib.get_metadata.restype = METADATA
+predict.restype = POINTER(c_float)
-    return lib.get_metadata(f)
 
-def load_net(cfg, weights):
+reset_rnn = lib.reset_rnn
-    load_network = lib.load_network_p
+reset_rnn.argtypes = [c_void_p]
-    load_network.argtypes = [c_char_p, c_char_p, c_int]
-    load_network.restype = c_void_p
-    return load_network(cfg, weights, 0)
 
-def load_img(f):
+load_net = lib.load_network_p
-    load_image = lib.load_image_color
+load_net.argtypes = [c_char_p, c_char_p, c_int]
-    load_image.argtypes = [c_char_p, c_int, c_int]
+load_net.restype = c_void_p
-    load_image.restype = IMAGE
-    return load_image(f, 0, 0)
 
-def letterbox_img(im, w, h):
+letterbox_image = lib.letterbox_image
-    letterbox_image = lib.letterbox_image
+letterbox_image.argtypes = [IMAGE, c_int, c_int]
-    letterbox_image.argtypes = [IMAGE, c_int, c_int]
+letterbox_image.restype = IMAGE
-    letterbox_image.restype = IMAGE
-    return letterbox_image(im, w, h)
 
-def predict(net, im):
+load_meta = lib.get_metadata
-    pred = lib.network_predict_image
+lib.get_metadata.argtypes = [c_char_p]
-    pred.argtypes = [c_void_p, IMAGE]
+lib.get_metadata.restype = METADATA
-    pred.restype = POINTER(c_float)
+
-    return pred(net, im)
+load_image = lib.load_image_color
+load_image.argtypes = [c_char_p, c_int, c_int]
+load_image.restype = IMAGE
+
+predict_image = lib.network_predict_image
+predict_image.argtypes = [c_void_p, IMAGE]
+predict_image.restype = POINTER(c_float)
 
 def classify(net, meta, im):
-    out = predict(net, im)
+    out = predict_image(net, im)
     res = []
     for i in range(meta.classes):
         res.append((meta.names[i], out[i]))
@@ -54,17 +68,19 @@ def classify(net, meta, im):
     return res
 
 def detect(net, meta, im):
-    out = predict(net, im)
+    out = predict_image(net, im)
     res = []
     for i in range(meta.classes):
         res.append((meta.names[i], out[i]))
     res = sorted(res, key=lambda x: -x[1])
     return res
+    
 
 if __name__ == "__main__":
-    net = load_net("cfg/densenet.cfg", "/home/pjreddie/trained/densenet201.weights")
+    net = load_net("cfg/densenet201.cfg", "/home/pjreddie/trained/densenet201.weights", 0)
-    im = load_img("data/wolf.jpg")
+    im = load_image("data/wolf.jpg", 0, 0)
     meta = load_meta("cfg/imagenet1k.data")
     r = classify(net, meta, im)
     print r[:10]
+    
 

src/network.c

@@ -82,6 +82,27 @@ void reset_momentum(network net)
     #endif
 }
 
+void reset_network_state(network net, int b)
+{
+    int i;
+    for (i = 0; i < net.n; ++i) {
+        #ifdef GPU
+        layer l = net.layers[i];
+        if(l.state_gpu){
+            fill_gpu(l.outputs, 0, l.state_gpu + l.outputs*b, 1);
+        }
+        if(l.h_gpu){
+            fill_gpu(l.outputs, 0, l.h_gpu + l.outputs*b, 1);
+        }
+        #endif
+    }
+}
+
+void reset_rnn(network *net)
+{
+    reset_network_state(*net, 0);
+}
+
 float get_current_rate(network net)
 {
     size_t batch_num = get_current_batch(net);
@@ -302,6 +323,15 @@ float train_network(network net, data d)
     return (float)sum/(n*batch);
 }
 
+void set_temp_network(network net, float t)
+{
+    int i;
+    for(i = 0; i < net.n; ++i){
+        net.layers[i].temperature = t;
+    }
+}
+
+
 void set_batch_network(network *net, int b)
 {
     net->batch = b;

src/region_layer.c

@@ -203,7 +203,7 @@ void forward_region_layer(const layer l, network net)
                         int class_index = entry_index(l, b, n, l.coords + 1);
                         int obj_index = entry_index(l, b, n, l.coords);
                         float scale =  l.output[obj_index];
-                        //l.delta[obj_index] = l.noobject_scale * (0 - l.output[obj_index]);
+                        l.delta[obj_index] = l.noobject_scale * (0 - l.output[obj_index]);
                         float p = scale*get_hierarchy_probability(l.output + class_index, l.softmax_tree, class, l.w*l.h);
                         if(p > maxp){
                             maxp = p;
@@ -213,8 +213,8 @@ void forward_region_layer(const layer l, network net)
                     int class_index = entry_index(l, b, maxi, l.coords + 1);
                     int obj_index = entry_index(l, b, maxi, l.coords);
                     delta_region_class(l.output, l.delta, class_index, class, l.classes, l.softmax_tree, l.class_scale, l.w*l.h, &avg_cat);
-                    //if(l.output[obj_index] < .3) l.delta[obj_index] = l.object_scale * (.3 - l.output[obj_index]);
+                    if(l.output[obj_index] < .3) l.delta[obj_index] = l.object_scale * (.3 - l.output[obj_index]);
-                    //else  l.delta[obj_index] = 0;
+                    else  l.delta[obj_index] = 0;
                     l.delta[obj_index] = 0;
                     ++class_count;
                     onlyclass = 1;