[None][perf] Validate DeepSeek R1 transceiver v2 performance#16152
[None][perf] Validate DeepSeek R1 transceiver v2 performance#16152asfiyab-nvidia wants to merge 2 commits into
Conversation
📝 WalkthroughWalkthroughThis PR adds ChangesPerf-sanity NIXL cache transceiver runtime configuration
Estimated code review effort: 1 (Trivial) | ~5 minutes 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In
`@tests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con1024_ctx1_dep4_gen1_dep32_eplb0_mtp3_ccb-NIXL.yaml`:
- Around line 63-66: The NIXL perf-sanity YAMLs are inconsistent because some
`cache_transceiver_config` blocks include `transceiver_runtime: PYTHON` while
many other `backend: NIXL` configs do not. Update the remaining disaggregated
perf-sanity YAMLs that use NIXL so their `cache_transceiver_config` matches the
same runtime setting as the existing GB200 config, using the
`cache_transceiver_config` and `backend: NIXL` entries as the target locations.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 3cd88616-8c8a-4e0c-8835-83a1a3a23aae
📒 Files selected for processing (28)
tests/scripts/perf-sanity/disaggregated/b200_deepseek-r1-fp4_1k1k_con1_ctx1_dep4_gen1_tep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/b200_deepseek-r1-fp4_1k1k_con2048_ctx1_dep4_gen1_dep8_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/b200_deepseek-r1-fp4_1k1k_con256_ctx1_dep4_gen1_dep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/b200_deepseek-r1-fp4_8k1k_con1536_ctx1_dep4_gen1_dep8_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/b200_deepseek-r1-fp4_8k1k_con1_ctx1_dep4_gen1_tep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/b200_deepseek-r1-fp4_8k1k_con256_ctx1_dep4_gen1_dep8_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_128k8k_con128_ctx1_pp8_gen1_dep16_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_128k8k_con1_ctx1_pp8_gen1_tep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_128k8k_con64_ctx1_pp8_gen1_dep32_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con1024_ctx1_dep4_gen1_dep32_eplb0_mtp0_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con1024_ctx1_dep4_gen1_dep32_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con1024_ctx1_dep4_gen1_dep8_eplb0_mtp0_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con1_ctx1_dep4_gen1_tep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con2048_ctx2_dep4_gen1_dep16_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con2048_ctx2_dep4_gen1_dep16_eplb288_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con3072_ctx1_dep4_gen1_dep4_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_8k1k_con1024_ctx1_dep4_gen1_dep32_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_8k1k_con1_ctx1_dep4_gen1_tep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_128k8k_con1_ctx1_pp4_gen1_tep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_128k8k_con256_ctx1_pp4_gen1_dep8_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_128k8k_con64_ctx1_pp4_gen1_dep16_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_1k1k_con1024_ctx1_dep4_gen1_dep32_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_1k1k_con1_ctx1_dep4_gen1_tep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_1k1k_con3072_ctx1_dep4_gen1_dep4_eplb0_mtp1_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_8k1k_con1024_ctx1_dep4_gen1_dep32_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_8k1k_con1_ctx1_dep4_gen1_tep8_eplb0_mtp3_ccb-NIXL.yamltests/scripts/perf-sanity/disaggregated/gb300_deepseek-r1-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp1_ccb-NIXL.yaml
| cache_transceiver_config: | ||
| max_tokens_in_buffer: 16384 | ||
| backend: NIXL | ||
| transceiver_runtime: PYTHON |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Find all disaggregated perf-sanity NIXL configs and check whether they all have transceiver_runtime: PYTHON alongside backend: NIXL
for f in $(fd -e yaml . tests/scripts/perf-sanity/disaggregated | rg -l 'NIXL'); do
if rg -q 'backend: NIXL' "$f" && ! rg -q 'transceiver_runtime: PYTHON' "$f"; then
echo "MISSING: $f"
fi
doneRepository: NVIDIA/TensorRT-LLM
Length of output: 205
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Locate all disaggregated perf-sanity YAMLs mentioning NIXL and inspect the
# backend/runtime pairings.
files=$(fd -a -e yaml . tests/scripts/perf-sanity/disaggregated 2>/dev/null | sort)
echo "== NIXL-related files =="
echo "$files" | rg 'NIXL' || true
echo
echo "== Files with backend: NIXL but missing transceiver_runtime: PYTHON =="
missing=0
while IFS= read -r f; do
[[ -z "$f" ]] && continue
if rg -q 'backend:\s*NIXL' "$f"; then
if ! rg -q 'transceiver_runtime:\s*PYTHON' "$f"; then
echo "$f"
missing=1
fi
fi
done <<< "$files"
exit "$missing"Repository: NVIDIA/TensorRT-LLM
Length of output: 18185
🏁 Script executed:
#!/bin/bash
set -euo pipefail
root="tests/scripts/perf-sanity/disaggregated"
python3 - <<'PY'
from pathlib import Path
import re
root = Path("tests/scripts/perf-sanity/disaggregated")
files = sorted(root.glob("*NIXL.yaml"))
backend_nixl = []
with_python = []
missing = []
for f in files:
text = f.read_text()
if re.search(r'backend:\s*NIXL\b', text):
backend_nixl.append(f)
if re.search(r'transceiver_runtime:\s*PYTHON\b', text):
with_python.append(f)
else:
missing.append(f)
print("backend:NIXL files:", len(backend_nixl))
print("with transceiver_runtime:PYTHON:", len(with_python))
print("missing transceiver_runtime:PYTHON:", len(missing))
print("\nMissing files:")
for f in missing:
print(f"- {f}")
PYRepository: NVIDIA/TensorRT-LLM
Length of output: 6020
Extend transceiver_runtime: PYTHON to the rest of the NIXL perf-sanity configs. 47 backend: NIXL YAMLs in tests/scripts/perf-sanity/disaggregated/ still omit it, so this suite will mix runtimes and produce inconsistent perf results.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In
`@tests/scripts/perf-sanity/disaggregated/gb200_deepseek-r1-fp4_1k1k_con1024_ctx1_dep4_gen1_dep32_eplb0_mtp3_ccb-NIXL.yaml`
around lines 63 - 66, The NIXL perf-sanity YAMLs are inconsistent because some
`cache_transceiver_config` blocks include `transceiver_runtime: PYTHON` while
many other `backend: NIXL` configs do not. Update the remaining disaggregated
perf-sanity YAMLs that use NIXL so their `cache_transceiver_config` matches the
same runtime setting as the existing GB200 config, using the
`cache_transceiver_config` and `backend: NIXL` entries as the target locations.
|
/bot run --stage-list "DGX_B200-16_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU8-Post-Merge-1, GB200-8_GPUs-2_Nodes-PyTorch-PerfSanity-Node2-GPU8-Post-Merge-1, GB200-12_GPUs-3_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE2-GPU8-Post-Merge-1, GB200-16_GPUs-4_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE2-GPU8-Post-Merge-1, GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-1, GB200-24_GPUs-6_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE4-GPU16-Post-Merge-1, GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-1, GB200-40_GPUs-10_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE8-GPU32-Post-Merge-1, GB300-8_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU4-Post-Merge-1, GB300-12_GPUs-3_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE2-GPU8-Post-Merge-1, GB300-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-1" |
|
PR_Github #58519 [ run ] triggered by Bot. Commit: |
|
PR_Github #58519 [ run ] completed with state
|
|
/bot run --disable-fail-fast --stage-list "DGX_B200-8_GPUs-PyTorch-PerfSanity-Post-Merge-1, |
|
PR_Github #58688 [ run ] triggered by Bot. Commit: |
|
PR_Github #58688 [ run ] completed with state
|
|
/bot run --disable-reuse-test --disable-fail-fast --stage-list "DGX_B200-8_GPUs-PyTorch-PerfSanity-Post-Merge-1, |
|
PR_Github #59321 [ run ] triggered by Bot. Commit: |
|
PR_Github #59321 [ run ] completed with state
|
5171328 to
4b4cdcb
Compare
|
/bot run --disable-reuse-test --disable-fail-fast --stage-list "DGX_B200-16_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU8-Post-Merge-1, DGX_B200-8_GPUs-PyTorch-PerfSanity-Post-Merge-1, DGX_B200-8_GPUs-PyTorch-PerfSanity-Post-Merge-2, DGX_B200-8_GPUs-PyTorch-PerfSanity-Post-Merge-4, DGX_B200-8_GPUs-PyTorch-PerfSanity-Post-Merge-5, DGX_B200-8_GPUs-PyTorch-PerfSanity-Post-Merge-6, GB200-12_GPUs-3_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE2-GPU8-Post-Merge-1, GB200-16_GPUs-4_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE2-GPU8-Post-Merge-1, GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-1, GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-1, GB200-4_GPUs-PyTorch-PerfSanity-Post-Merge-3, GB200-4_GPUs-PyTorch-PerfSanity-Post-Merge-5, GB200-4_GPUs-PyTorch-PerfSanity-Post-Merge-6, GB200-4_GPUs-PyTorch-PerfSanity-Post-Merge-8, GB200-4_GPUs-PyTorch-PerfSanity-Post-Merge-9, GB200-8_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU4-Post-Merge-1, GB300-12_GPUs-3_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE2-GPU8-Post-Merge-1, GB300-4_GPUs-PyTorch-PerfSanity-Post-Merge-2" |
|
PR_Github #59690 [ run ] triggered by Bot. Commit: |
|
PR_Github #59690 [ run ] completed with state
|
Signed-off-by: Asfiya Baig <asfiyab@nvidia.com>
Signed-off-by: Asfiya Baig <asfiyab@nvidia.com>
4b4cdcb to
2bfc673
Compare
|
/bot run --disable-reuse-test --disable-fail-fast --stage-list "DGX_B200-16_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU8-Post-Merge-1, GB200-8_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU4-Post-Merge-1, GB200-12_GPUs-3_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE2-GPU8-Post-Merge-1, GB200-16_GPUs-4_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE2-GPU8-Post-Merge-1, GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-1, GB200-24_GPUs-6_Nodes-PyTorch-Disagg-PerfSanity-CTX2-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-1, GB200-24_GPUs-6_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE4-GPU16-Post-Merge-1, GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-1, GB200-40_GPUs-10_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE8-GPU32-Post-Merge-1, GB300-8_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU4-Post-Merge-1, GB300-12_GPUs-3_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE2-GPU8-Post-Merge-1, GB300-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-1" |
|
PR_Github #59867 [ run ] triggered by Bot. Commit: |
|
PR_Github #59868 [ run ] triggered by Bot. Commit: |
|
PR_Github #59867 [ run ] completed with state |
|
PR_Github #59868 [ run ] completed with state
|
|
See #16536 |
4 participants
Summary by CodeRabbit
Description
Test Coverage
PR Checklist
Please review the following before submitting your PR:
PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.
PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.
Test cases are provided for new code paths (see test instructions)
If PR introduces API changes, an appropriate PR label is added - either
api-compatibleorapi-breaking. Forapi-breaking, includeBREAKINGin the PR title.Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
Update tava architecture diagram if there is a significant design change in PR.
The reviewers assigned automatically/manually are appropriate for the PR.
Please check this after reviewing the above items as appropriate for this PR.
GitHub Bot Help
To see a list of available CI bot commands, please comment
/bot help.