Use n_gen from context, fix tps mismatch in gemma3 #3954
  Add this suggestion to a batch that can be applied as a single commit.
  This suggestion is invalid because no changes were made to the code.
  Suggestions cannot be applied while the pull request is closed.
  Suggestions cannot be applied while viewing a subset of changes.
  Only one suggestion per line can be applied in a batch.
  Add this suggestion to a batch that can be applied as a single commit.
  Applying suggestions on deleted lines is not supported.
  You must change the existing code in this line in order to create a valid suggestion.
  Outdated suggestions cannot be applied.
  This suggestion has been applied or marked resolved.
  Suggestions cannot be applied from pending reviews.
  Suggestions cannot be applied on multi-line comments.
  Suggestions cannot be applied while the pull request is queued to merge.
  Suggestion cannot be applied right now. Please check back later.
  
    
  
    
This PR is trying to address #3947 . When benchmarking gemma3, this model generates eos token very quick, i.e. 3~4 tokens, but we still calculate the decoding tps by 128. So this model displays a very high decoding speed.
before fix:
decode=128
316.10 ± 2.58
decode=128
311.16 ± 1.67
decode=128
11.91 ± 0.14
after fix:
decode=128
12.28 ± 0.22
decode=128
12.22 ± 0.04
decode=128
12.04 ± 0.02