You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am benchmarking our HPC platform on a config with no GPU's (48 virtual cores and 128 GB memory). How long does data processing usually take and how long does training usually take?
I took a loot at the tail of the stdout (it has been running since Wednesday) just now and see that it says the following:
Does this mean that day_0 hasnt even finished processing yet or is it just on a recently started epoch? I cant look at the whole stdout right now since the file is about 57 GB big.
Thank you,
Rumit
The text was updated successfully, but these errors were encountered:
I apolgize for the delay in response. The pre-processing takes multiple days, please refer to the discussion in #119 and #58 for more details.
Also, I wanted to highlight the discussion in #199 (comment) "If you are interested in dealing with a large dataset, you will need to have uninterrupted access to a machine/node with at least 256GB RAM and 8TB of disk space for a long period of time (weeks). My advice is to try working with smaller dataset before trying the larger one."
Good Afternoon,
I am benchmarking our HPC platform on a config with no GPU's (48 virtual cores and 128 GB memory). How long does data processing usually take and how long does training usually take?
I took a loot at the tail of the stdout (it has been running since Wednesday) just now and see that it says the following:
Processing ./input/day_0.npz
Load 10508943/199563535 Split: 1 Label True: 0 Stored: 0
Does this mean that day_0 hasnt even finished processing yet or is it just on a recently started epoch? I cant look at the whole stdout right now since the file is about 57 GB big.
Thank you,
Rumit
The text was updated successfully, but these errors were encountered: