Unlike the import function, which is a parallelized reader, the upload function is a push from the client to the server. The specified path must be a client-side path. This is not scalable and is only intended for smaller data sizes. The client pushes the data from a local filesystem (for example, on your machine where R or Python is running) to H2O. For big-data operations, you don’t want the data stored on or flowing through the client.
Run the following command to load data that resides on the same machine that is running H2O.
> library(h2o) > h2o.init(nthreads=-1) > irisPath = "../../../smalldata/iris/iris_wheader.csv" > iris.hex = h2o.uploadFile(path = irisPath, destination_frame = "iris.hex")
>>> import h2o >>> h2o.init() >>> df = h2o.upload_file("../smalldata/iris/iris_wheader.csv")