merge

python/jittor/__init__.py

@@ -273,6 +273,7 @@ Var.start_grad = Var.detach_inplace = detach_inplace
 
 def unsqueeze(x, dim):
     shape = list(x.shape)
+    if dim < 0: dim += len(shape) + 1
     assert dim <= len(shape)
     return x.reshape(shape[:dim] + [1] + shape[dim:])
 Var.unsqueeze = unsqueeze

setup.py

@@ -21,7 +21,7 @@ with open(os.path.join(path, "README.md"), "r", encoding='utf8') as fh:
 
 setuptools.setup(
     name='jittor',  
-    version='1.1.5.2',
+    version='1.1.5.4',
     # scripts=[],
     author="Jittor Group",
     author_email="ran.donglang@gmail.com",

src/grad.cc

@@ -92,7 +92,10 @@ vector<VarPtr> grad(Var* loss, vector<Var*> targets) {
             Op* op = it.op;
             auto index = it.index;
             if (op->tflag != nt) continue;
+<<<<<<< HEAD
             // TODO: support two outputs backprop.
+=======
+>>>>>>> b27082f9444a4e627f7dfc574d0114302ba27b5e
             for (Var* out : op->outputs()) {
                 if (out->tflag != nt) continue;
                 Var* dout = grads[out->custom_data];

src/pyjt/py_converter.h

@@ -303,10 +303,10 @@ DEF_IS(ArrayArgs, T) from_py_object(PyObject* obj) {
         auto arr = (PyArray_Proxy*)holder.obj;
         int64 size = PyArray_Size(arr);
         T args;
-        args.ptr = arr->data;
         args.shape = vector<int64>(arr->dimensions, arr->dimensions+arr->nd);
         args.dtype = get_type_str(arr);
         args.buffer.reset(new char[size]);
+        args.ptr = (void*)args.buffer.get();
         memcpy((void*)args.buffer.get(), (void*)arr->data, size);
         return args;
     }

src/var_holder.h

@@ -106,7 +106,7 @@ struct VarHolder {
     /* detach the grad */
     // @pyjt(detach)
     inline VarHolder* detach() {
-        return new VarHolder(move(jittor::detach(var)));
+        return new VarHolder(jittor::detach(var));
     }