diff --git a/config.pbtxt b/config.pbtxt index 5d90dc8..185add8 100644 --- a/config.pbtxt +++ b/config.pbtxt @@ -1,7 +1,7 @@ # Triton backend to use name: "base-gemma-3-1b-it" backend: "python" -max_batch_size: 4 +max_batch_size: 5 # Triton should expect as input a single string # input of variable length named 'text_input' @@ -125,8 +125,8 @@ parameters: [ instance_group [ { - kind: KIND_CPU - count: 8 + kind: KIND_GPU + count: 7 } ]