samskalicky commented on a change in pull request #15886: Graph Partition API
URL: https://github.com/apache/incubator-mxnet/pull/15886#discussion_r317702345
##########
File path: src/c_api/c_api_symbolic.cc
##########
@@ -1199,3 +1200,73 @@ int MXShallowCopySymbol(SymbolHandle src, SymbolHandle*
out) {
*out = out_sym;
API_END_HANDLE_ERROR(delete out_sym);
}
+
+int MXOptimizeForBackend(SymbolHandle sym_handle,
+ const char* backend_name,
+ SymbolHandle* ret_sym_handle,
+ const mx_uint len,
+ NDArrayHandle* in_args_handle,
+ const mx_uint num_options,
+ const char** keys,
+ const char** vals) {
+ nnvm::Symbol *s = new nnvm::Symbol();
+ API_BEGIN();
+ nnvm::Symbol *sym = static_cast<nnvm::Symbol *>(sym_handle);
+ *s = sym->Copy();
+ nnvm::Graph g = Symbol2Graph(*s);
+ if (len) {
+ NDArray **in_args_ptr = reinterpret_cast<NDArray**>(in_args_handle);
+ Context default_ctx = in_args_ptr[0]->ctx();
Review comment:
this context is only used to inferStorageTypes. In this pass we're not
creating any lasting tensors/arrays so context doesnt matter. In these passes
we just want to know shape/type info which is context agnostic.
It doenst matter if the user uses a diff context when they bind since there
are no arrays created in the previous context to convert over.
I think the only reason that context is required for InferStorageTypes is so
that MKLDNN can use it, correct me if im wrong.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
