[AMD] Switch Qwen3.5 FP8 MI355X benchmarks to aiter attention backend

Author: yichiche

benchmarks/single_node/fixed_seq_len/qwen3.5_fp8_mi355x.sh

@@ -18,21 +18,21 @@ fi
 
 if [[ "$MODEL" != /* ]]; then hf download "$MODEL"; fi
 
+export SGLANG_USE_AITER_UNIFIED_ATTN=1
+export SGLANG_USE_AITER=1
+
 SERVER_LOG=/workspace/server.log
-CONTEXT_LENGTH=$((ISL + OSL + 20))
-MAX_PREFILL_TOKENS=32768
 
 EVAL_CONTEXT_ARGS=""
 if [ "${EVAL_ONLY}" = "true" ]; then
     setup_eval_context
     EVAL_CONTEXT_ARGS="--context-length $EVAL_MAX_MODEL_LEN"
-else EVAL_CONTEXT_ARGS="--context-length $CONTEXT_LENGTH"
 fi
 # Start GPU monitoring (power, temperature, clocks every second)
 start_gpu_monitor
 
 python3 -m sglang.launch_server \
-    --attention-backend triton \
+    --attention-backend aiter \
     --model-path $MODEL \
     --host=0.0.0.0 \
     --port $PORT \
@@ -41,11 +41,13 @@ python3 -m sglang.launch_server \
     --trust-remote-code \
     --tokenizer-worker-num 6 \
     --enable-aiter-allreduce-fusion \
-    --cuda-graph-max-bs $CONC \
+    --max-running-requests 512 \
     --disable-radix-cache \
-    --max-prefill-tokens $MAX_PREFILL_TOKENS \
+    --chunked-prefill-size 32768 \
     --scheduler-recv-interval 30 \
-    --mem-fraction-static 0.8 $EVAL_CONTEXT_ARGS > $SERVER_LOG 2>&1 &
+    --mem-fraction-static 0.9 \
+    --model-loader-extra-config '{"enable_multithread_load": true}' \
+    --page-size 16 $EVAL_CONTEXT_ARGS > $SERVER_LOG 2>&1 &
 
 SERVER_PID=$!
 

benchmarks/single_node/fixed_seq_len/qwen3.5_fp8_mi355x_mtp.sh

@@ -18,21 +18,21 @@ fi
 
 if [[ "$MODEL" != /* ]]; then hf download "$MODEL"; fi
 
+export SGLANG_USE_AITER_UNIFIED_ATTN=1
+export SGLANG_USE_AITER=1
+
 SERVER_LOG=/workspace/server.log
-CONTEXT_LENGTH=$((ISL + OSL + 20))
-MAX_PREFILL_TOKENS=32768
 
 EVAL_CONTEXT_ARGS=""
 if [ "${EVAL_ONLY}" = "true" ]; then
     setup_eval_context
     EVAL_CONTEXT_ARGS="--context-length $EVAL_MAX_MODEL_LEN"
-else EVAL_CONTEXT_ARGS="--context-length $CONTEXT_LENGTH"
 fi
 # Start GPU monitoring (power, temperature, clocks every second)
 start_gpu_monitor
 
 python3 -m sglang.launch_server \
-    --attention-backend triton \
+    --attention-backend aiter \
     --model-path $MODEL \
     --host=0.0.0.0 \
     --port $PORT \
@@ -41,11 +41,13 @@ python3 -m sglang.launch_server \
     --trust-remote-code \
     --tokenizer-worker-num 6 \
     --enable-aiter-allreduce-fusion \
-    --cuda-graph-max-bs $CONC \
+    --max-running-requests 512 \
     --disable-radix-cache \
-    --max-prefill-tokens $MAX_PREFILL_TOKENS \
+    --chunked-prefill-size 32768 \
     --scheduler-recv-interval 30 \
-    --mem-fraction-static 0.8 \
+    --mem-fraction-static 0.9 \
+    --model-loader-extra-config '{"enable_multithread_load": true}' \
+    --page-size 16 \
     --speculative-algorithm EAGLE \
     --speculative-num-steps 3 \
     --speculative-eagle-topk 1 \