Neofytos Dimitriou

Hi @jcupitt, I appreciate your help. Thank you! > You'd need to share your benchmark for me to comment on the speed. SSD ``` (pytorch_env) nd26@volta:~/Downloads$ time vips avg CMU-1.svs...

> That's curious, you should see user about 2x real on a two-core machine. What processor are you using? ``` description: CPU product: Intel(R) Core(TM) i3-8350K CPU @ 4.00GHz ```...

Hi @jcupitt, I just had the chance to get back to this. I've managed to run your code, and indeed, the latter (with fetch) is impressively faster. Correct me if...

> Those are big patches! Indeed! > How many slide images will you be working from? Around 650!

> 650 isn't so many. I would open them all (ie. make 650 pipelines) and then use fetch to pull out tiles. > > You could try using crop as...

Hey @bacalfa, I've successfully managed to install pysurvival (on Windows) from your repository. Thank you!

@bjkomer Is it possible to load the trials if they were saved as follows (not pickled and saved): ``` domain = base.Domain(objective, search_space) main_plot_history(trials) data = pd.DataFrame(trials.trials) data = pd.concat([data.drop(["result"],...

Indeed, I had used hp.choice for allowing TPE determine which classifier to choose next. Nevertheless, following this error, my current methodology simply runs TPE on the different classifiers independently (hence...

Can't you retrieve that parameter through the objective function and save it from there?