Skip to content

Benchmarks subset #2614

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kwasd merged 16 commits into from
Nov 12, 2024
Merged

Benchmarks subset #2614

Changes from 15 commits
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
16479fc
Benchmarks core subset
kwasd Nov 3, 2024
488f57c
Test skiplist approach
kwasd Nov 12, 2024
284d898
Test skiplist approach
kwasd Nov 12, 2024
6ccc243
Test skiplist approach
kwasd Nov 12, 2024
c67ca36
Test skiplist approach
kwasd Nov 12, 2024
0fbe25b
Test skiplist approach
kwasd Nov 12, 2024
12aa9ef
Test skiplist approach
kwasd Nov 12, 2024
d10efe0
Test skiplist approach
kwasd Nov 12, 2024
bed2f70
Merge branch 'main' into feature/benchmarks-subset
kwasd Nov 12, 2024
df852e2
Parse input as JSON
kwasd Nov 12, 2024
aecd310
Merge branch 'feature/benchmarks-subset' of https://github.com/intel/…
kwasd Nov 12, 2024
8bc9c48
Update .github/workflows/triton-benchmarks.yml
kwasd Nov 12, 2024
d895bc4
Update .github/workflows/triton-benchmarks.yml
kwasd Nov 12, 2024
fcdfae4
Update .github/workflows/triton-benchmarks.yml
kwasd Nov 12, 2024
a2854ff
Parse input as JSON
kwasd Nov 12, 2024
e6d02f1
fix typo
kwasd Nov 12, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 20 additions & 16 deletions .github/workflows/triton-benchmarks.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,10 @@ on:
description: Run name
type: string
default: "Triton benchmarks"
skip_benchmarks:
description: JSON list of benchmarks to skip
type: string
default: "[]"
schedule:
- cron: "5 23 * * *"
pull_request:
Expand Down Expand Up @@ -112,7 +116,7 @@ jobs:
python setup.py install

- name: Run Triton Softmax kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'fused_softmax.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python fused_softmax.py --reports $REPORTS
Expand All @@ -121,7 +125,7 @@ jobs:
python ../../scripts/build_report.py $REPORTS/softmax-performance.csv $REPORTS/softmax-xetla-report.csv --benchmark softmax --compiler xetla --param_cols "N" --tflops_col XeTLA-TFlops --hbm_col "XeTLA-GB/s" --tag $TAG

- name: Run Triton GEMM kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_benchmark.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python gemm_benchmark.py --reports $REPORTS
Expand All @@ -132,7 +136,7 @@ jobs:
python ../../scripts/build_report.py $REPORTS/matmul-performance-base.csv $REPORTS/gemm-xetla-report.csv --benchmark gemm --compiler xetla --param_cols "B,M,K,N" --tflops_col XeTLA-TFlops --hbm_col "XeTLA-GB/s" --tag $TAG

- name: Run Triton GEMM kernel benchmark - default path
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'f), 'gemm_benchmark.py_default') }}
run: |
cd benchmarks/triton_kernels_benchmark
# Default path:
Expand All @@ -148,7 +152,7 @@ jobs:
python ../../scripts/build_report.py $REPORTS/matmul-performance-default-path.csv $REPORTS/gemm-triton-default-report.csv --benchmark gemm --compiler triton --param_cols "B,M,K,N" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Triton GEMM kernel benchmark - advanced path
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_benchmark.py_advanced') }}
run: |
cd benchmarks/triton_kernels_benchmark
# Advanced path:
Expand All @@ -164,7 +168,7 @@ jobs:
python ../../scripts/build_report.py $REPORTS/matmul-performance-adv-path.csv $REPORTS/gemm-triton-advanced-report.csv --benchmark gemm --compiler triton --param_cols "B,M,K,N" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Triton GEMM (A@B^t) kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_benchmark.py_abt') }}
run: |
cd benchmarks/triton_kernels_benchmark
TRANSPOSE_B=1 python gemm_benchmark.py --reports $REPORTS
Expand All @@ -175,7 +179,7 @@ jobs:
python ../../scripts/build_report.py $REPORTS/matmul-performance-bt.csv $REPORTS/gemm-bt-onednn-report.csv --benchmark gemm-bt --compiler onednn --param_cols "B,M,K,N" --tflops_col onednn-TFlops --hbm_col "onednn-GB/s" --tag $TAG

- name: Run Triton GEMM (A^t@B) kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_benchmark.py_atb') }}
run: |
cd benchmarks/triton_kernels_benchmark
TRANSPOSE_A=1 python gemm_benchmark.py --reports $REPORTS
Expand All @@ -186,47 +190,47 @@ jobs:
python ../../scripts/build_report.py $REPORTS/matmul-performance-at.csv $REPORTS/gemm-at-onednn-report.csv --benchmark gemm-at --compiler onednn --param_cols "B,M,K,N" --tflops_col onednn-TFlops --hbm_col "onednn-GB/s" --tag $TAG

- name: Run Triton GEMM (stream-k) kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_streamk_benchmark.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python gemm_streamk_benchmark.py --reports $REPORTS
source ../../scripts/capture-hw-details.sh
python ../../scripts/build_report.py $REPORTS/matmul-streamk-performance.csv $REPORTS/gemm-streamk-triton-report.csv --benchmark gemm-streamk --compiler triton --param_cols "M,K,N" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Triton GEMM (split-k) kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_splitk_benchmark.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python gemm_splitk_benchmark.py --reports $REPORTS
source ../../scripts/capture-hw-details.sh
python ../../scripts/build_report.py $REPORTS/matmul-splitk-performance.csv $REPORTS/gemm-splitk-triton-report.csv --benchmark gemm-splitk --compiler triton --param_cols "M,K,N" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Triton GEMM + PreOp (exp) kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_preop_exp_benchmark.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python gemm_preop_exp_benchmark.py --reports $REPORTS
source ../../scripts/capture-hw-details.sh
python ../../scripts/build_report.py $REPORTS/matmul-performance-preop-exp.csv $REPORTS/gemm-preop-exp-triton-report.csv --benchmark gemm-preop-exp --compiler triton --param_cols "B,M,K,N" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Triton GEMM + PostOp (Gelu) kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_postop_gelu_benchmark.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python gemm_postop_gelu_benchmark.py --reports $REPORTS
source ../../scripts/capture-hw-details.sh
python ../../scripts/build_report.py $REPORTS/matmul-performance-postop-gelu.csv $REPORTS/gemm-postop-gelu-triton-report.csv --benchmark gemm-postop-gelu --compiler triton --param_cols "B,M,K,N" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Triton GEMM + PostOp (add matrix) kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_postop_addmatrix_benchmark.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python gemm_postop_addmatrix_benchmark.py --reports $REPORTS
source ../../scripts/capture-hw-details.sh
python ../../scripts/build_report.py $REPORTS/matmul-performance-postop-addmatrix.csv $REPORTS/gemm-postop-addmatrix-triton-report.csv --benchmark gemm-postop-addmatrix --compiler triton --param_cols "B,M,K,N" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Triton FA kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flash_attention_fwd_benchmark.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python flash_attention_fwd_benchmark.py --reports $REPORTS
Expand All @@ -236,7 +240,7 @@ jobs:
python ../../scripts/build_report.py $REPORTS/attn-performance.csv $REPORTS/attn-xetla-report.csv --benchmark attn --compiler xetla --param_cols "Z,H,N_CTX,D_HEAD,CAUSAL" --tflops_col XeTLA-TFlops --hbm_col "XeTLA-GB/s" --tag $TAG

- name: Run Triton FA kernel benchmark - default path
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmark || '[]'), 'flash_attention_fwd_benchmark.py_default') }}
run: |
cd benchmarks/triton_kernels_benchmark
TRITON_INTEL_ADVANCED_PATH=0 \
Expand All @@ -249,7 +253,7 @@ jobs:
python ../../scripts/build_report.py $REPORTS/attn-performance.csv $REPORTS/attn-triton-default-report.csv --benchmark attn --compiler triton --param_cols "Z,H,N_CTX,D_HEAD,CAUSAL" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Triton FA kernel benchmark - advanced path
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flash_attention_fwd_benchmark.py_advanced') }}
run: |
cd benchmarks/triton_kernels_benchmark
TRITON_INTEL_ADVANCED_PATH=1 \
Expand All @@ -262,15 +266,15 @@ jobs:
python ../../scripts/build_report.py $REPORTS/attn-performance.csv $REPORTS/attn-triton-advanced-report.csv --benchmark attn --compiler triton --param_cols "Z,H,N_CTX,D_HEAD,CAUSAL" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run Prefix Sums kernel benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'prefix_sums.py') }}
run: |
cd benchmarks/triton_kernels_benchmark
python prefix_sums.py --reports $REPORTS
source ../../scripts/capture-hw-details.sh
python ../../scripts/build_report.py $REPORTS/prefix-sums.csv $REPORTS/prefix_sums-triton-report.csv --benchmark prefix_sums --compiler triton --param_cols "N" --tflops_col Triton-TFlops --hbm_col "Triton-GB/s" --tag $TAG

- name: Run micro benchmark
if: ${{ steps.install.outcome == 'success' && !cancelled() }}
if: ${{ steps.install.outcome == 'success' && !cancelled() && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'micro_benchmarks') }}
run: |
cd benchmarks/micro_benchmarks
python run_benchmarks.py --reports $REPORTS
Expand Down