GPU number inconsistency between topology file and workload

I followed the SimAI-Simulation Tutorial and it works:

# Create network topo
$ python3 ./astra-sim-alibabacloud/inputs/topo/gen_HPN_7.0_topo_mulgpus_one_link.py -g 128 -gt A100 -bw 100Gbps -nvbw 2400Gbps

# Running
$ AS_SEND_LAT=3 AS_NVLS_ENABLE=1 ./bin/SimAI_simulator -t 16 -w ./example/microAllReduce.txt -n ./HPN_7_0_128_gpus_8_in_one_server_with_single_plane_100Gbps_A100 -c astra-sim-alibabacloud/inputs/config/SimAI.conf

It seems that the total gpu number is 128 in the topology file. However, the gpu number in ./example/microAllReduce.txt is only 8. What is the reason of this inconsistency?

Besides, when I tried to generate a workload of 128 gpus:

./bin/SimAI_analytical \
    -w /workspace/aicb/results/workload/-gpt_7B-world_size128-tp4-pp1-ep1-gbs16-mbs1-seq16-MOE-False-GEMM-False-flash_attn-True.txt \
    -g 128 \
    -g_p_s 8 \
    -r test-analytical \
    -busbw example/busbw.yaml \
    -v

and run the Sim-AI-Simulation

python3 ./astra-sim-alibabacloud/inputs/topo/gen_HPN_7.0_topo_mulgpus_one_link.py \
    -g 128 \
    -gt A100 \
    -gps 8 \
    -bw 100Gbps \
    -nvbw 2400Gbps

AS_NVLS_ENABLE=1 AS_SEND_LAT=3 ./bin/SimAI_simulator \
    -t 32 \
    -w /workspace/aicb/results/workload/-gpt_7B-world_size128-tp4-pp1-ep1-gbs16-mbs1-seq16-MOE-False-GEMM-False-flash_attn-True.txt \
    -n /workspace/SimAI/HPN_7_0_128_gpus_8_in_one_server_with_single_plane_100Gbps_A100 \
    -c astra-sim-alibabacloud/inputs/config/SimAI.conf

I could only get empty output csv files.

Start by comparing the GPU-count arguments and generated topology name in astra-sim-alibabacloud/inputs/topo/gen_HPN_7.0_topo_mulgpus_one_link.py with the workload files passed to SimAI_simulator. Then inspect astra-sim-alibabacloud/inputs/config/SimAI.conf and the simulator run output to determine why the 128-GPU workload produces empty CSV files. Done means the expected GPU-count relationship is documented or corrected and the reproduction produces meaningful output.