Skip to content

Commit 979146e

Browse files
anmyachevanmyachev
authored
Remove intel-pti==0.12.4 pin since we switched to DLE 2025.2 (#5015)
Current PTI version from DLE 2025.2.1 is 0.13.1 Signed-off-by: Anatoly Myachev <[email protected]>
1 parent 36749ec commit 979146e

File tree

1 file changed

+0
-26
lines changed

1 file changed

+0
-26
lines changed

‎.github/workflows/triton-benchmarks.yml‎

Lines changed: 0 additions & 26 deletions
Original file line numberDiff line numberDiff line change
@@ -126,16 +126,10 @@ jobs:
126126
run: |
127127
cd benchmarks
128128
pip install .
129-
pip install intel-pti==0.12.4
130-
PTI_LIBS_DIR=$(python -c "import sysconfig; print(sysconfig.get_paths()['stdlib']+'/..')")
131-
# the output should contain: `libpti.so`, `libpti_metrics.so.0.12.4` and `libpti_view.so.0.12.4`
132-
ls $PTI_LIBS_DIR
133-
echo "PTI_LIBS_DIR=$PTI_LIBS_DIR" >> $GITHUB_ENV
134129
135130
- name: Run Triton Softmax kernel benchmark
136131
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'fused_softmax.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'fused_softmax.py') }}
137132
run: |
138-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
139133
cd benchmarks/triton_kernels_benchmark
140134
python fused_softmax.py --reports $REPORTS --n_runs $N_RUNS
141135
source ../../scripts/capture-hw-details.sh
@@ -146,7 +140,6 @@ jobs:
146140
- name: Run Triton GEMM kernel benchmark
147141
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_benchmark.py') }}
148142
run: |
149-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
150143
cd benchmarks/triton_kernels_benchmark
151144
python gemm_benchmark.py --reports $REPORTS --n_runs $N_RUNS
152145
mv $REPORTS/matmul-performance.csv $REPORTS/matmul-performance-base.csv
@@ -160,7 +153,6 @@ jobs:
160153
- name: Run Triton GEMM kernel benchmark - with tensor of pointer
161154
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_tensor_of_ptr_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_tensor_of_ptr_benchmark.py') }}
162155
run: |
163-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
164156
cd benchmarks/triton_kernels_benchmark
165157
python gemm_tensor_of_ptr_benchmark.py --reports $REPORTS --n_runs $N_RUNS
166158
source ../../scripts/capture-hw-details.sh
@@ -173,7 +165,6 @@ jobs:
173165
- name: Run Triton GEMM kernel benchmark - with tensor descriptor
174166
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_tensor_desc_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_tensor_desc_benchmark.py') }}
175167
run: |
176-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
177168
cd benchmarks/triton_kernels_benchmark
178169
python gemm_tensor_desc_benchmark.py --reports $REPORTS --n_runs $N_RUNS
179170
source ../../scripts/capture-hw-details.sh
@@ -186,7 +177,6 @@ jobs:
186177
- name: Run Triton GEMM (A@B^t) kernel benchmark
187178
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_benchmark.py_abt')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_benchmark.py_abt') }}
188179
run: |
189-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
190180
cd benchmarks/triton_kernels_benchmark
191181
TRANSPOSE_B=1 python gemm_benchmark.py --reports $REPORTS --n_runs $N_RUNS
192182
mv $REPORTS/matmul-performance.csv $REPORTS/matmul-performance-bt.csv
@@ -198,7 +188,6 @@ jobs:
198188
- name: Run Triton GEMM (A^t@B) kernel benchmark
199189
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_benchmark.py_atb')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_benchmark.py_atb') }}
200190
run: |
201-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
202191
cd benchmarks/triton_kernels_benchmark
203192
TRANSPOSE_A=1 python gemm_benchmark.py --reports $REPORTS --n_runs $N_RUNS
204193
mv $REPORTS/matmul-performance.csv $REPORTS/matmul-performance-at.csv
@@ -210,7 +199,6 @@ jobs:
210199
- name: Run Triton GEMM (stream-k) kernel benchmark
211200
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_streamk_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_streamk_benchmark.py') }}
212201
run: |
213-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
214202
cd benchmarks/triton_kernels_benchmark
215203
python gemm_streamk_benchmark.py --reports $REPORTS --n_runs $N_RUNS
216204
source ../../scripts/capture-hw-details.sh
@@ -220,7 +208,6 @@ jobs:
220208
- name: Run Triton GEMM (split-k) kernel benchmark
221209
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_splitk_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_splitk_benchmark.py') }}
222210
run: |
223-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
224211
cd benchmarks/triton_kernels_benchmark
225212
python gemm_splitk_benchmark.py --reports $REPORTS --n_runs $N_RUNS
226213
source ../../scripts/capture-hw-details.sh
@@ -230,7 +217,6 @@ jobs:
230217
- name: Run Triton GEMM + PreOp (exp) kernel benchmark
231218
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_preop_exp_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_preop_exp_benchmark.py') }}
232219
run: |
233-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
234220
cd benchmarks/triton_kernels_benchmark
235221
python gemm_preop_exp_benchmark.py --reports $REPORTS --n_runs $N_RUNS
236222
source ../../scripts/capture-hw-details.sh
@@ -239,7 +225,6 @@ jobs:
239225
- name: Run Triton GEMM + PostOp (Gelu) kernel benchmark
240226
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_postop_gelu_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_postop_gelu_benchmark.py') }}
241227
run: |
242-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
243228
cd benchmarks/triton_kernels_benchmark
244229
python gemm_postop_gelu_benchmark.py --reports $REPORTS --n_runs $N_RUNS
245230
source ../../scripts/capture-hw-details.sh
@@ -248,7 +233,6 @@ jobs:
248233
- name: Run Triton GEMM + PostOp (add matrix) kernel benchmark bfloat16
249234
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_postop_addmatrix_benchmark_bfloat16.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_postop_addmatrix_benchmark_bfloat16.py') }}
250235
run: |
251-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
252236
cd benchmarks/triton_kernels_benchmark
253237
python gemm_postop_addmatrix_benchmark.py --reports $REPORTS --n_runs $N_RUNS
254238
source ../../scripts/capture-hw-details.sh
@@ -258,7 +242,6 @@ jobs:
258242
- name: Run Triton GEMM + PostOp (add matrix) kernel benchmark int8
259243
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'gemm_postop_addmatrix_benchmark_int8.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'gemm_postop_addmatrix_benchmark_int8.py') }}
260244
run: |
261-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
262245
cd benchmarks/triton_kernels_benchmark
263246
INT8_ONLY=1 python gemm_postop_addmatrix_benchmark.py --reports $REPORTS --n_runs $N_RUNS
264247
source ../../scripts/capture-hw-details.sh
@@ -268,7 +251,6 @@ jobs:
268251
- name: Run Triton FA fwd kernel benchmark
269252
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'flash_attention_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flash_attention_benchmark.py') }}
270253
run: |
271-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
272254
cd benchmarks/triton_kernels_benchmark
273255
python flash_attention_benchmark.py --reports $REPORTS --n_runs $N_RUNS
274256
@@ -279,7 +261,6 @@ jobs:
279261
- name: Run Triton FA bwd kernel benchmark
280262
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'flash_attention_bwd_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flash_attention_bwd_benchmark.py') }}
281263
run: |
282-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
283264
cd benchmarks/triton_kernels_benchmark
284265
FA_KERNEL_MODE="bwd" \
285266
python flash_attention_benchmark.py --reports $REPORTS --n_runs $N_RUNS
@@ -292,7 +273,6 @@ jobs:
292273
- name: Run Triton FA fwd kernel benchmark - with tensor descriptors
293274
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'flash_attention_tensor_desc_benchmark.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flash_attention_tensor_desc_benchmark.py') }}
294275
run: |
295-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
296276
cd benchmarks/triton_kernels_benchmark
297277
python flash_attention_tensor_desc_benchmark.py --reports $REPORTS --n_runs $N_RUNS
298278
mv $REPORTS/attn-performance.csv $REPORTS/attn-tensor-desc-performance.csv
@@ -304,7 +284,6 @@ jobs:
304284
- name: Run Triton FlexAttention Causal Mask fwd kernel benchmark
305285
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'flex_attention_benchmark_causal_mask.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flex_attention_benchmark_causal_mask.py') }}
306286
run: |
307-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
308287
cd benchmarks/triton_kernels_benchmark
309288
python flex_attention_benchmark_causal_mask.py --reports $REPORTS --n_runs $N_RUNS
310289
@@ -315,7 +294,6 @@ jobs:
315294
- name: Run Triton FlexAttention (batch_size=4) Causal Mask fwd kernel benchmark
316295
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'flex_attention_benchmark_batch4-causal_mask.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flex_attention_benchmark_batch4-causal_mask.py') }}
317296
run: |
318-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
319297
cd benchmarks/triton_kernels_benchmark
320298
BATCH_SIZE=4 python flex_attention_benchmark_causal_mask.py --reports $REPORTS --n_runs $N_RUNS
321299
@@ -326,7 +304,6 @@ jobs:
326304
- name: Run Triton FlexAttention (batch_size=16) Causal Mask fwd kernel benchmark
327305
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'flex_attention_benchmark_batch16-causal_mask.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flex_attention_benchmark_batch16-causal_mask.py') }}
328306
run: |
329-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
330307
cd benchmarks/triton_kernels_benchmark
331308
BATCH_SIZE=16 python flex_attention_benchmark_causal_mask.py --reports $REPORTS --n_runs $N_RUNS
332309
@@ -337,7 +314,6 @@ jobs:
337314
- name: Run Triton FlexAttention Custom Masks fwd kernel benchmark
338315
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'flex_attention_benchmark_custom_masks.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'flex_attention_benchmark_custom_masks.py') }}
339316
run: |
340-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
341317
cd benchmarks/triton_kernels_benchmark
342318
python flex_attention_benchmark_custom_masks.py --reports $REPORTS --n_runs $N_RUNS
343319
@@ -351,7 +327,6 @@ jobs:
351327
- name: Run Prefix Sums kernel benchmark
352328
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'prefix_sums.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'prefix_sums.py') }}
353329
run: |
354-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
355330
cd benchmarks/triton_kernels_benchmark
356331
python prefix_sums.py --reports $REPORTS --n_runs $N_RUNS
357332
source ../../scripts/capture-hw-details.sh
@@ -360,7 +335,6 @@ jobs:
360335
- name: Run micro benchmark
361336
if: ${{ steps.install.outcome == 'success' && !cancelled() && (inputs.benchmarks == '' || contains(fromJson(inputs.benchmarks || '[]'), 'micro_benchmarks.py')) && !contains(fromJson(inputs.skip_benchmarks || '[]'), 'micro_benchmarks') }}
362337
run: |
363-
export LD_LIBRARY_PATH=$PTI_LIBS_DIR:$LD_LIBRARY_PATH
364338
cd benchmarks/micro_benchmarks
365339
python run_benchmarks.py --reports $REPORTS
366340

0 commit comments

Comments
 (0)