Merge pull request #335 from deephealthproject/develop

Fix Windows + Update docs
deephealthproject · Mar 9, 2022 · e839f80 · e839f80
2 parents feffd06 + 19a2fba
commit e839f80
Show file tree

Hide file tree

Showing 6 changed files with 23 additions and 28 deletions.
diff --git a/docs/sphinx/source/bundle/initializers.rst b/docs/sphinx/source/bundle/initializers.rst
@@ -1,11 +1,6 @@
 Initializers
 ============
 
-.. note::
-
-    Section in progress
-
-    Read this: https://github.com/deephealthproject/eddl/blob/master/docs/markdown/eddl_progress.md#initializers
 
 GlorotNormal
 ------------

diff --git a/docs/sphinx/source/model/onnx.rst b/docs/sphinx/source/model/onnx.rst
@@ -5,7 +5,7 @@ Save and load ONNX models
 Exporting to onnx
 ---------------------
 
-.. doxygenfunction:: save_net_to_onnx_file(Net* net, string path)
+.. doxygenfunction:: save_net_to_onnx_file(Net *net, string path, int seq_len = 0)
 
 Example:
 

diff --git a/examples/nn/2_cifar10/4_cifar_vgg16_bn.cpp b/examples/nn/2_cifar10/4_cifar_vgg16_bn.cpp
@@ -82,9 +82,9 @@ int main(int argc, char **argv){
   if (use_cpu) {
       cs = CS_CPU();
   } else {
-      //cs = CS_GPU({1}); // one GPU
+      cs = CS_GPU({1}); // one GPU
       // cs = CS_GPU({1,1},100); // two GPU with weight sync every 100 batches
-       cs = CS_CPU();
+//       cs = CS_CPU();
       // cs = CS_FPGA({1});
   }
 

diff --git a/formulas/conda/eddl/build.sh b/formulas/conda/eddl/build.sh
@@ -30,7 +30,7 @@ echo "CPU_COUNT=$CPU_COUNT"
 echo "#################################################"
 
 # Build makefiles
-mkdir build && cd build/ && cmake .. -DBUILD_TARGET=CUDNN \
+mkdir build && cd build/ && cmake .. -DBUILD_TARGET=CPU \
       -DBUILD_SUPERBUILD=OFF \
       -DBUILD_EXAMPLES=OFF \
       -DBUILD_TESTS=OFF \

diff --git a/formulas/conda/eddl/meta.yaml b/formulas/conda/eddl/meta.yaml
@@ -1,6 +1,6 @@
-{% set name = "eddl-cudnn" %}  # If this is a package for GPU, use: "eddl-gpu"
-{% set version = "1.0.2a" %}
-{% set sha256 = "70d6067d44cb547e218236e5bb72faf45f602a7b8214f869626ca47f241481a0" %}
+{% set name = "eddl-cpu" %}  # eddl-cpu, eddl-gpu, eddl-cudnn
+{% set version = "1.1b" %}
+{% set sha256 = "534ca4bd5e83236c164a26a058e08df971d28c0ae5b32674bec364ae0370843a" %}
 
 package:
   name: {{ name|lower }}
@@ -19,23 +19,23 @@ requirements:
   build:
     - {{ compiler('cxx') }}
   host:
-    - cmake>=3.17.2
-    - eigen==3.3.7
+    - cmake=3.19
+    - eigen=3.3|3.4
     - protobuf==3.11.4
-    - libprotobuf==3.11.4  # We need to avoid problems with paths (idk why)
+    - libprotobuf==3.11.4
     - zlib==1.2.11
     - openssl==1.1.1i
-    - graphviz  # Build & Run  (versions can be problematic)
-    - wget
+    - wget=1.20|1.16
+    - graphviz
   run:
-    - cmake>=3.17.2
-    - eigen==3.3.7
+    - cmake=3.19
+    - eigen=3.3|3.4
     - protobuf==3.11.4
-    - libprotobuf==3.11.4  # We need to avoid problems with paths (idk why)
+    - libprotobuf==3.11.4
     - zlib==1.2.11
     - openssl==1.1.1i
-    - graphviz  # Build & Run  (versions can be problematic)
-    - wget
+    - wget=1.20|1.16
+    - graphviz
 
 test:
   commands:

diff --git a/src/layers/core/layer_dense.cpp b/src/layers/core/layer_dense.cpp
@@ -63,8 +63,8 @@ void LDense::resize(int batch) {
 }
 
 void LDense::forward() {
-    input->reshape_({input->size / input->shape.back(), input->shape.back()});
-    output->reshape_({output->size / output->shape.back(), output->shape.back()});
+    input->reshape_({static_cast<int>(input->size / input->shape.back()), input->shape.back()});
+    output->reshape_({static_cast<int>(output->size / output->shape.back()), output->shape.back()});
     Tensor::mult2D(input, 0, W, 0, output, 0);
     if (use_bias) Tensor::sum2D_rowwise(output, bias, output);
 
@@ -73,11 +73,11 @@ void LDense::forward() {
 }
 
 void LDense::backward() {
-    input->reshape_({input->size / input->shape.back(), input->shape.back()});
-    parent[0]->delta->reshape_({input->size / input->shape.back(), input->shape.back()});
+    input->reshape_({static_cast<int>(input->size / input->shape.back()), input->shape.back()});
+    parent[0]->delta->reshape_({static_cast<int>(input->size / input->shape.back()), input->shape.back()});
 
-    output->reshape_({output->size / output->shape.back(), output->shape.back()});
-    delta->reshape_({output->size / output->shape.back(), output->shape.back()});
+    output->reshape_({static_cast<int>(output->size / output->shape.back()), output->shape.back()});
+    delta->reshape_({static_cast<int>(output->size / output->shape.back()), output->shape.back()});
 
     //get gradients with provided delta
     if (trainable) {