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
This PR shows that large scale tunes can be effective. We have additional ideas we would like to test through SPSA now, such as artifically making L2 31 for example. However, it is not possible to try this when the param limit is around 1k, as this would require something like 50k params per layerstack.
A possible solution that has been discussed is file parsing for SPSA. This would consist of passing a list of values in csv or similar format to the worker. This would also require outputting as csv to the user on the site instead of the table and graph. The input SPSA table should also be parsed in as a csv file.
The text was updated successfully, but these errors were encountered:
As a result of official-stockfish/Stockfish#5149 , im opening this issue to bring attention back to:
#1792 and #1781
This PR shows that large scale tunes can be effective. We have additional ideas we would like to test through SPSA now, such as artifically making L2 31 for example. However, it is not possible to try this when the param limit is around 1k, as this would require something like 50k params per layerstack.
A possible solution that has been discussed is file parsing for SPSA. This would consist of passing a list of values in csv or similar format to the worker. This would also require outputting as csv to the user on the site instead of the table and graph. The input SPSA table should also be parsed in as a csv file.
The text was updated successfully, but these errors were encountered: