Scale P800 long-context matrix to concurrency=1 for stability

- P800 INT8 cannot sustain >1 concurrent long-context request without OOM;
  set all scenarios to concurrency=1.
- Keep max_running=2 because max_running=1 triggers req_to_token_pool
  allocation failures in this image.
- Reduce 64k max_context_len to 72000 (65536 + 4096 + pad).
- Make run_bench.sh resilient to container death: detect dead container and
  skip remaining scenarios in the group instead of crashing the orchestrator.
- Update README to document the concurrency limitation.

experiments/dsv4_p800_long_context_matrix/README.md

@@ -9,7 +9,7 @@ Long-context throughput/latency matrix for SGLang-XPU on Kunlun P800 (8x XPU).
 - Hardware: 8x Kunlun P800 XPU
 - Input lengths: 64k, 128k
 - Output lengths: 256, 1024, 4096
-- Concurrency: sweeps per context group (see `config.env`)
+- Concurrency: **1** — P800 INT8 can only sustain one concurrent long-context request without OOM; `max_running=2` is kept because `max_running=1` triggers `req_to_token_pool` allocation failures in this image.
 
 ## Usage
 

experiments/dsv4_p800_long_context_matrix/config.env

@@ -1,7 +1,9 @@
 # Long-context matrix for SGLang-XPU on DeepSeek-V4-Flash-INT8 (P800, TP=8).
 # Input lengths: 64k, 128k (P800 INT8 max input length is ~131072 tokens).
 # Output lengths: 256, 1k, 4k
-# Concurrency: conservative to fit XPU memory.
+# Concurrency: P800 INT8 can only sustain 1 concurrent long-context request
+# without OOM, so all scenarios use concurrency=1. max_running is kept at 2
+# because max_running=1 triggers req_to_token_pool allocation failures.
 
 EXPERIMENT="dsv4_p800_long_context_matrix"
 MODEL_NAME="DeepSeek-V4-Flash-INT8"
@@ -16,33 +18,21 @@ CONTEXT_PAD="${CONTEXT_PAD:-1024}"
 # Per-context server settings. Each group is run with its own server start.
 # Format: "input_label max_context_len max_running"
 declare -a CONTEXT_GROUPS=(
-  "64k  80000  10"
+  "64k  72000  2"
-  "128k 140000 5"
+  "128k 140000 2"
 )
 
 # Scenarios per group: "concurrency input_len output_len num_prompts"
 declare -a SCENARIOS_64K=(
-  "10 65536  256  50"
+  "1 65536  256  10"
-  "5  65536  256  25"
+  "1 65536  1024 10"
-  "2  65536  256  10"
+  "1 65536  4096 10"
-  "10 65536  1024 50"
-  "5  65536  1024 25"
-  "2  65536  1024 10"
-  "10 65536  4096 50"
-  "5  65536  4096 25"
-  "2  65536  4096 10"
 )
 
 declare -a SCENARIOS_128K=(
-  "5  131072 256  20"
+  "1 131072 256  5"
-  "2  131072 256  10"
+  "1 131072 1024 5"
-  "1  131072 256  5"
+  "1 131072 4096 5"
-  "5  131072 1024 20"
-  "2  131072 1024 10"
-  "1  131072 1024 5"
-  "5  131072 4096 20"
-  "2  131072 4096 10"
-  "1  131072 4096 5"
 )
 
 CONTAINER_NAME="${CONTAINER_NAME:-sglang-dsv4-flash}"

experiments/dsv4_p800_long_context_matrix/run_bench.sh

@@ -38,6 +38,10 @@ is_server_healthy() {
   curl --fail --silent --show-error --max-time 5 "http://127.0.0.1:${PORT}/health" >/dev/null 2>&1
 }
 
+container_running() {
+  docker inspect "$CONTAINER_NAME" >/dev/null 2>&1
+}
+
 stop_server() {
   log "stopping container ${CONTAINER_NAME}"
   docker rm -f "$CONTAINER_NAME" 2>/dev/null || true
@@ -59,6 +63,10 @@ run_warmup() {
   local detail_log="${LOG_DIR}/warmup_${input_len}_${output_len}.log"
 
   log "warming up (input=${input_len}, output=${output_len}, num=1)"
+  if ! container_running; then
+    log "WARNING: container not running, skipping warmup"
+    return 1
+  fi
   docker exec "$CONTAINER_NAME" mkdir -p "$(dirname "$container_output")"
 
   docker exec "$CONTAINER_NAME" \
@@ -142,6 +150,11 @@ run_group() {
       continue
     fi
 
+    if ! container_running; then
+      log "WARNING: container died, skipping remaining scenarios in group=${group_label}"
+      break
+    fi
+
     log "running scenario: group=${group_label} c=${concurrency} i=${input_len} o=${output_len} n=${num_prompts}"
 
     docker exec "$CONTAINER_NAME" mkdir -p "$(dirname "$container_output")"
@@ -168,6 +181,10 @@ run_group() {
 
     if [[ "$bench_rc" -ne 0 ]]; then
       log "ERROR: scenario group=${group_label} c=${concurrency} i=${input_len} o=${output_len} failed; see ${detail_log}"
+      if ! container_running; then
+        log "WARNING: container died, skipping remaining scenarios in group=${group_label}"
+        break
+      fi
       continue
     fi
 
@@ -241,8 +258,8 @@ data["config"] = {
     "container_name": "sglang-dsv4-flash",
     "context_pad": 1024,
     "groups": [
-        {"label": "64k", "max_context_len": 80000, "max_running": 10},
+        {"label": "64k", "max_context_len": 72000, "max_running": 2},
-        {"label": "128k", "max_context_len": 140000, "max_running": 5}
+        {"label": "128k", "max_context_len": 140000, "max_running": 2}
     ]
 }
 with open(path, "w", encoding="utf-8") as f: