dank/serverAppConfigs
1
0
Fork 0
Code Issues 6 Pull Requests Actions Packages Projects Releases Wiki Activity

Validation: 2 parallel prompts, ~18 tok/s with DFlash

Browse Source 
- Add second parallel prompt (Byzantine Empire history vs ML)
- Both hit 1024 tokens, combined 2048 in 111.6s = 18.34 tok/s
- Parallel throughput matches single-stream baseline
This commit is contained in:
Frank
2026-08-07 16:21:20 -07:00
parent 25d57184a6
commit 764a65080f
1 changed files with 29 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/validate.sh
+29 -10
View File
@@ -26,12 +26,12 @@ for i in $(seq 1 300); do
sleep 2
done
# Send the request and capture timing
# Send two parallel requests and capture timing
echo ""
echo "Sending validation prompt..."
echo "Sending 2 parallel validation prompts..."
START_TIME=$(date +%s%N)
RESPONSE=$(curl -s \
RESPONSE1=$(curl -s \
-X POST "http://${HOST}:${PORT}/v1/chat/completions" \
-H "Content-Type: application/json" \
-d "{
@@ -41,26 +41,45 @@ RESPONSE=$(curl -s \
\"max_tokens\": ${MAX_TOKENS},
\"temperature\": 0.7,
\"top_p\": 0.95
}")
}" &)
RESPONSE2=$(curl -s \
-X POST "http://${HOST}:${PORT}/v1/chat/completions" \
-H "Content-Type: application/json" \
-d "{
\"messages\": [
{\"role\": \"user\", \"content\": \"Write a detailed analysis of the geopolitical consequences of the fall of the Byzantine Empire in 1453. Trace the chain of events from the Ottoman conquest of Constantinople through the next two centuries, covering at minimum: the shift in Mediterranean trade routes, the Italian Renaissance's relationship with fleeing Greek scholars, the Russian claim to Roman succession, and the papal response. Then compare how the region might have developed differently had Constantinople remained independent.\"}
],
\"max_tokens\": ${MAX_TOKENS},
\"temperature\": 0.7,
\"top_p\": 0.95
}" &)
wait
END_TIME=$(date +%s%N)
# Extract token counts and timing from response
COMPLETION_TOKENS=$(echo "$RESPONSE" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d['usage']['completion_tokens'])" 2>/dev/null || echo "0")
# 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")
TOTAL_TOKENS=$((COMPLETION_TOKENS1 + COMPLETION_TOKENS2))
TOTAL_TIME_MS=$(( (END_TIME - START_TIME) / 1000000 ))
if [ "$COMPLETION_TOKENS" -eq 0 ]; then
if [ "$TOTAL_TOKENS" -eq 0 ]; then
echo "FAIL: No tokens generated or response parse error"
echo "Response: $RESPONSE"
echo "Response1: $RESPONSE1"
echo "Response2: $RESPONSE2"
exit 1
fi
# Calculate tokens/sec
TOK_PER_SEC=$(python3 -c "print(f'{${COMPLETION_TOKENS} * 1000 / ${TOTAL_TIME_MS}:.2f}')")
TOK_PER_SEC=$(python3 -c "print(f'{${TOTAL_TOKENS} * 1000 / ${TOTAL_TIME_MS}:.2f}')")
echo ""
echo "=== Results ==="
echo "Completion tokens: ${COMPLETION_TOKENS}"
echo "Prompt 1 tokens: ${COMPLETION_TOKENS1}"
echo "Prompt 2 tokens: ${COMPLETION_TOKENS2}"
echo "Total tokens: ${TOTAL_TOKENS}"
echo "Total time: ${TOTAL_TIME_MS}ms"
echo "Speed: ${TOK_PER_SEC} tok/s"
echo ""