move ntasks per node to framework level instead of runner level

Author: cquil11

benchmarks/dsr1_fp4_gb200_dynamo-trt_slurm.sh

@@ -39,10 +39,11 @@ fi
 
 kind=dynamo_disagg
 additional_slurm_args="--time=04:00:00"
+ntasks_per_node=4
 
 gen_nodes=$(((DECODE_TP + 3)/4 * DECODE_NUM_WORKERS))
 total_nodes=$((PREFILL_NUM_WORKERS + gen_nodes))
-total_tasks=$((total_nodes * NTASKS_PER_NODE))
+total_tasks=$((total_nodes * ntasks_per_node))
 
 decode_eplb_num_slots=0
 

runners/launch_gb200-nv.sh

@@ -8,8 +8,6 @@ set -x
 export SLURM_PARTITION="batch"
 export SLURM_ACCOUNT="benchmark"
 export SLURM_JOB_NAME="benchmark-dynamo.job"
-# For GB200 we have 4 GPUs per node
-export NTASKS_PER_NODE=4
 
 # For SGLang - we are working on updating the 8k1k configs 
 # For now we add conditionals to this script to use newer code for the 1k1k configs