Fix parallel validation, remove ngram-mod

- Rewrite validation script with temp files for true parallel curl requests
- Both prompts now complete 1024 tokens each (was cutting first short)
- Remove ngram-mod from spec pipeline (was -1.2 tok/s regression)
- True parallel throughput: 15.68 tok/s across 2 streams

scripts/validate.sh

@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
-# Validation test for llama.cpp DFlash+MTP server
-# Sends a simple prompt and measures generation speed (tokens/sec)
+# Validation test for llama.cpp DFlash server
+# Sends 2 parallel prompts and measures generation speed (tokens/sec)
 set -euo pipefail
 
 PORT="${LLAMA_PORT:-8080}"
@@ -31,7 +31,11 @@ echo ""
 echo "Sending 2 parallel validation prompts..."
 START_TIME=$(date +%s%N)
 
-RESPONSE1=$(curl -s \
+TMP1=$(mktemp)
+TMP2=$(mktemp)
+trap "rm -f $TMP1 $TMP2" EXIT
+
+curl -s \
     -X POST "http://${HOST}:${PORT}/v1/chat/completions" \
     -H "Content-Type: application/json" \
     -d "{
@@ -41,9 +45,10 @@ RESPONSE1=$(curl -s \
         \"max_tokens\": ${MAX_TOKENS},
         \"temperature\": 0.7,
         \"top_p\": 0.95
-    }" &)
+    }" > "$TMP1" &
+PID1=$!
 
-RESPONSE2=$(curl -s \
+curl -s \
     -X POST "http://${HOST}:${PORT}/v1/chat/completions" \
     -H "Content-Type: application/json" \
     -d "{
@@ -53,22 +58,23 @@ RESPONSE2=$(curl -s \
         \"max_tokens\": ${MAX_TOKENS},
         \"temperature\": 0.7,
         \"top_p\": 0.95
-    }" &)
+    }" > "$TMP2" &
+PID2=$!
 
-wait
+wait $PID1 $PID2
 
 END_TIME=$(date +%s%N)
 
 # Extract token counts and timing from responses
-COMPLETION_TOKENS1=$(echo "$RESPONSE1" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d['usage']['completion_tokens'])" 2>/dev/null || echo "0")
-COMPLETION_TOKENS2=$(echo "$RESPONSE2" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d['usage']['completion_tokens'])" 2>/dev/null || echo "0")
+COMPLETION_TOKENS1=$(python3 -c "import json; d=json.load(open('$TMP1')); print(d['usage']['completion_tokens'])" 2>/dev/null || echo "0")
+COMPLETION_TOKENS2=$(python3 -c "import json; d=json.load(open('$TMP2')); print(d['usage']['completion_tokens'])" 2>/dev/null || echo "0")
 TOTAL_TOKENS=$((COMPLETION_TOKENS1 + COMPLETION_TOKENS2))
 TOTAL_TIME_MS=$(( (END_TIME - START_TIME) / 1000000 ))
 
 if [ "$TOTAL_TOKENS" -eq 0 ]; then
     echo "FAIL: No tokens generated or response parse error"
-    echo "Response1: $RESPONSE1"
-    echo "Response2: $RESPONSE2"
+    echo "Response1: $(cat $TMP1)"
+    echo "Response2: $(cat $TMP2)"
     exit 1
 fi