Support leaky_relu in mlir #4434

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

Jump to bottom

Open

pfultz2 wants to merge 6 commits into develop from mlir-leaky-relu

Collaborator

pfultz2 commented Nov 7, 2025 •

edited

Loading

Motivation

This adds support to fuse leaky_relu with mlir.

Technical Details

Rather than try and update mlir to support the operator, this rewrites leak_relu to the operators it already supports. This also refactors the other rewrites we do for mlir into a single pass prepare_mlir.

Changelog Category

- Added: New functionality.
- Changed: Changes to existing functionality.
- Removed: Functionality or support that has been removed. (Compared to a previous release)
- Optimized: Component performance that has been optimized or improved.
- Resolved Issues: Known issues from a previous version that have been resolved.
- Not Applicable: This PR is not to be included in the changelog.


          Support leaky_relu in mlir

f0293f0

pfultz2 requested a review from causten as a code owner

November 7, 2025 21:57

pfultz2 added 5 commits

November 7, 2025 15:58


          Format

72f8ebe


          Add verify test

c5f0670


          Format

7d11e5e


          Add comment

cdf22ef


          Format

4d8cb33

Collaborator

migraphx-bot commented Nov 8, 2025

Test	Batch	Rate new 4d8cb3	Rate old d80627	Diff	Compare
torchvision-resnet50	64	3,115.69	3,114.58	0.04%	✅
torchvision-resnet50_fp16	64	6,521.11	6,508.48	0.19%	✅
torchvision-densenet121	32	2,391.29	2,391.06	0.01%	✅
torchvision-densenet121_fp16	32	4,026.63	4,020.15	0.16%	✅
torchvision-inceptionv3	32	1,645.80	1,646.14	-0.02%	✅
torchvision-inceptionv3_fp16	32	2,473.74	2,472.42	0.05%	✅
cadene-inceptionv4	16	788.30	788.37	-0.01%	✅
cadene-resnext64x4	16	763.80	763.72	0.01%	✅
slim-mobilenet	64	8,023.61	8,018.97	0.06%	✅
slim-nasnetalarge	64	217.25	217.18	0.03%	✅
slim-resnet50v2	64	3,230.19	3,230.09	0.00%	✅
bert-mrpc-onnx	8	1,134.15	1,128.82	0.47%	✅
bert-mrpc-tf	1	447.73	452.21	-0.99%	✅
pytorch-examples-wlang-gru	1	331.12	294.48	12.44%	🔆
pytorch-examples-wlang-lstm	1	462.86	397.02	16.58%	🔆
torchvision-resnet50_1	1	751.77	753.31	-0.21%	✅
cadene-dpn92_1	1	416.61	415.79	0.20%	✅
cadene-resnext101_1	1	364.73	364.29	0.12%	✅
onnx-taau-downsample	1	325.70	313.27	3.97%	🔆
dlrm-criteoterabyte	1	31.73	31.46	0.86%	✅
dlrm-criteoterabyte_fp16	1	51.51	51.42	0.18%	✅
agentmodel	1	10,851.54	10,240.84	5.96%	🔆
unet_fp16	2	59.32	59.07	0.42%	✅
resnet50v1_fp16	1	964.71	969.60	-0.50%	✅
resnet50v1_int8	1	928.85	934.48	-0.60%	✅
bert_base_cased_fp16	64	1,105.14	1,104.98	0.01%	✅
bert_large_uncased_fp16	32	344.08	344.12	-0.01%	✅
bert_large_fp16	1	199.55	198.05	0.76%	✅
distilgpt2_fp16	16	2,081.05	2,086.62	-0.27%	✅
yolov5s	1	582.90	583.83	-0.16%	✅
tinyllama	1	43.80	43.78	0.05%	✅
vicuna-fastchat	1	45.00	45.15	-0.35%	✅
whisper-tiny-encoder	1	402.82	402.56	0.07%	✅
whisper-tiny-decoder	1	401.28	400.60	0.17%	✅
llama2_7b	1	19.09	19.08	0.06%	✅
qwen1.5-7b	1	23.46	23.45	0.04%	✅
phi3-3.8b	1	26.59	26.58	0.05%	✅
mask-rcnn	1	12.30	12.37	-0.63%	✅
llama3-8b	1	21.69	21.69	0.01%	✅
whisper-large-encoder	1	10.10	10.10	-0.02%	✅
whisper-large-decoder	1	101.75	101.43	0.32%	✅
mistral-7b	1	23.69	23.68	0.02%	✅
FLUX.1-schnell	1	719.25	719.02	0.03%	✅
nan	nan	nan	nan	nan%	❌

This build is not recommended to merge 🔴

Collaborator

migraphx-bot commented Nov 8, 2025

❌bert-mrpc-onnx: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌bert-mrpc-tf: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌pytorch-examples-wlang-gru: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌pytorch-examples-wlang-lstm: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌dlrm-criteoterabyte: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌agentmodel: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌unet: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌resnet50v1: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌bert_base_cased_fp16: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌bert_large_uncased_fp16: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌bert_large: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌yolov5s: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌tinyllama: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌vicuna-fastchat: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌whisper-tiny-encoder: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌whisper-tiny-decoder: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌distilgpt2_fp16: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌llama2_7b: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌qwen1.5-7b: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌phi3-3.8b: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌mask-rcnn: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌llama3-8b: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌whisper-large-decoder: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌mistral-7b: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

❌FLUX.1-schnell: ERROR - check error output

Traceback (most recent call last):
File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 26, in
import migraphx
ImportError: /src/AMDMIGraphX/build/lib/libmigraphx_gpu.so.2015000: undefined symbol: _ZN8migraphx14version_2_15_03gpu9is_reduceERKNS0_11instructionE

causten approved these changes

View reviewed changes

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet