more

Author: ishandhanani

.github/configs/nvidia-master.yaml

@@ -798,6 +798,7 @@ dsr1-fp8-gb200-dynamo-sglang:
         additional-settings:
         - "PREFILL_NODES=4"
         - "N_ADDITIONAL_FRONTENDS=9"
+        - "SCRIPT_MODE=max-tpt"
       decode:
         num-worker: 1
         tp: 1

benchmarks/dsr1_fp4_gb200_dynamo-sglang_slurm.sh

@@ -0,0 +1,37 @@
+
+#!/bin/bash
+
+set -x
+
+source "$(dirname "$0")/benchmark_lib.sh"
+
+check_env_vars CONC_LIST ISL OSL IMAGE SPEC_DECODING MODEL_PATH \
+    PREFILL_NUM_WORKERS PREFILL_TP PREFILL_EP PREFILL_DP_ATTN \
+    DECODE_NUM_WORKERS DECODE_TP DECODE_EP DECODE_DP_ATTN \
+    PREFILL_NODES DECODE_NODES N_ADDITIONAL_FRONTENDS SGL_SLURM_JOBS_PATH # SGL_SLURM_JOBS_PATH FIXME
+
+# Always clone and setup Dynamo
+echo "Cloning Dynamo repository..."
+git clone --branch ishan/sa-1.1-sgl-dsr1 https://github.com/ai-dynamo/dynamo.git
+
+cd "$SGL_SLURM_JOBS_PATH"
+
+# Set up SGL launch script-specific environment variables
+export TIME_LIMIT="04:00:00"
+export MODEL_PATH=$MODEL_PATH
+export CONFIG_DIR=$CONFIG_DIR
+export CONTAINER_IMAGE=$IMAGE
+export GPU_TYPE="gb200-fp4"
+
+# Launch jobs based on ISL/OSL
+# Replace ' ' in CONC_LIST with 'x' such that the concurrency list is represented
+# by a list of numbers delimted by 'x'. This is because of how the underlying launch script
+# expects the concurrencies.
+bash ./submit_disagg.sh $PREFILL_NODES \
+    $PREFILL_NUM_WORKERS \
+    $DECODE_NODES \
+    $DECODE_NUM_WORKERS \
+    $N_ADDITIONAL_FRONTENDS \
+    $ISL $OSL "${CONC_LIST// /x}" inf \
+    $GPU_TYPE \
+    $SCRIPT_MODE

benchmarks/dsr1_fp8_gb200_dynamo-sglang_slurm.sh

@@ -21,6 +21,7 @@ export TIME_LIMIT="04:00:00"
 export MODEL_PATH=$MODEL_PATH
 export CONFIG_DIR=$CONFIG_DIR
 export CONTAINER_IMAGE=$IMAGE
+export GPU_TYPE="gb200-fp8"
 
 # Launch jobs based on ISL/OSL
 # Replace ' ' in CONC_LIST with 'x' such that the concurrency list is represented
@@ -32,4 +33,5 @@ bash ./submit_disagg.sh $PREFILL_NODES \
     $DECODE_NUM_WORKERS \
     $N_ADDITIONAL_FRONTENDS \
     $ISL $OSL "${CONC_LIST// /x}" inf \
+    $GPU_TYPE \
     $SCRIPT_MODE