MLJ Interface is not working anymore about betaml.jl HOT 6 CLOSED

going to check, thanks to report....

from betaml.jl.

Hello, I can't reproduce... this script, where I used your code, works as expected...

using Pkg
Pkg.activate(temp=true)
Pkg.add("MLJ")
using DelimitedFiles,MLJ


# Iris dataset
X, y        = @load_iris

# Your code...
modelType = @load RandomForestClassifier pkg = "BetaML" verbosity=1
mod = modelType(
n_trees = 2,
max_depth = 10
)

mach        = machine(mod, X, y)
fit!(mach)
cat_est    = predict(mach, X)
ŷ =   MLJ.mode.(cat_est)
accuracy = sum(y .== ŷ) / length(y) # 0.95

Be sure you have the latest versions of both MLJ and BetaML (respectively v0.20.2 and v0.11.3)

from betaml.jl.

Hello, do you still have the issue after updating to latest versions?

from betaml.jl.

I found this issue because I was having a similar problem. Attempting to load the RandomForestClassifier model directly was failing to resolve, but doing so with @load worked. On further analysis, it appears that RandomForestClassifier is in the sub-module Bmlj, which needs to be explicitly included. For example:

# This Works
model = (MLJ.@load RandomForestClassifier pkg = "BetaML")()

# Also Works
model = BetaML.Bmlj.RandomForestClassifier()

# Doesn't Work
model = BetaML.RandomForestClassifier()

Moreover, the links to the MLJ models appear to be broken in the README.

I'm assuming this is a recent change? It would be nice to be able to access the MLJ models directly from the BetaML module. Alternatively, the documentation should be clearer to avoid confusion.

from betaml.jl.

Hi, model = BetaML.RandomForestClassifier() is NOT supposed to work.
You can access RandomForestClassifier and RandomForestRegressor trought the MLJ @load method, or the RandomForestEstimator trought the BetaML interface (the first two are just wrappers to the third one).
While names are different in this case, sometimes the names of the models are the same, this is why BetaML does NOT export the MLJ-interface models.
If you want the MLJ interface models without using @load, you can do:

using BetaML # or import
m = BetaML.Bmlj.RandomForestClassifier()

Thanks for the report on the links and let me know if I can close this issue...

from betaml.jl.

..closing... feel free to reopen if needed...

from betaml.jl.