jeremiedb commented on issue #10928: Optimizers memory usage URL: https://github.com/apache/incubator-mxnet/issues/10928#issuecomment-396429822 As an additional pointer, the issue seems tied to R's NDArray being set to read only after an initial mutation. https://github.com/apache/incubator-mxnet/blob/master/R-package/src/ndarray.h#L86 @tqchen ? Basic comparative example between R and Python: R: ``` > state <- mx.nd.array(c(1,2,3,3)) > grad <- mx.nd.array(1:4) > weight <- mx.nd.array(1:4) > mx.nd.sgd.mom.update(weight, grad, state, lr = 0.1, momentum = 0.5, out = weight) [1] 1.4 2.8 4.2 5.1 > mx.nd.sgd.mom.update(weight, grad, state, lr = 0.1, momentum = 0.5, out = weight) Error in mx.nd.sgd.mom.update(weight, grad, state, lr = 0.1, momentum = 0.5, : ./ndarray.h:87: RCheck failed: ptr_->writable && !ptr_->moved Passing a read only NDArray to mutate function ``` Python: ``` state = mx.nd.array([1, 2, 3, 3]) grad = mx.nd.array([1,2,3,4]) weight = mx.nd.array([1,2,3,4]) print(nd.sgd_mom_update(weight, grad, state, momentum = 0.5, lr=0.1, out=weight)) print(nd.sgd_mom_update(weight, grad, state, momentum = 0.5, lr=0.1, out=weight)) [1.4 2.8 4.2 5.1] <NDArray 4 @cpu(0)> [1.5 3. 4.5 5.25] <NDArray 4 @cpu(0)> ```
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services