Question about the training_reuse_numb_steps #1600
              
                Unanswered
              
          
                  
                    
                      xiongluojie
                    
                  
                
                  asked this question in
                Q&A
              
            Replies: 1 comment 1 reply
-
| may not always be true. the new data (data labeled by the previous iteration) should be properly trained with  | 
Beta Was this translation helpful? Give feedback.
                  
                    1 reply
                  
                
            
  
    Sign up for free
    to join this conversation on GitHub.
    Already have an account?
    Sign in to comment
  
        
    
Uh oh!
There was an error while loading. Please reload this page.
-
When I continue training a model from the last iteration, what is the appropriate value for training_reuse_numb_steps? Can it be significantly fewer than the number of steps used in the initial training of the model?
Thanks!
Beta Was this translation helpful? Give feedback.
All reactions