Bump test suite to include punet tests and fix xfail script. (iree-or…

…g#17897)

Highlights from this update:

*
nod-ai/SHARK-TestSuite@fc9fd62
Fixing the `update_config_xfails.py` script
*
nod-ai/SHARK-TestSuite@2c0a312
(and follow-ups) Adding the [Partitioned UNet
(punet)](https://github.com/nod-ai/sharktank/tree/main/sharktank/sharktank/models/punet)
model to the test suite, using data from
https://huggingface.co/amd-shark/sdxl-quant-models

This also changes the 'models' pytest command to include all output,
even for xfail'd tests. This will let us see the latest compile/run
errors for these programs using the tested flags.

ci-exactly: build_packages,regression_test

.github/workflows/pkgci_regression_test.yml

@@ -90,7 +90,7 @@ jobs:
         uses: actions/checkout@v4.1.7
         with:
           repository: nod-ai/SHARK-TestSuite
-          ref: a06e730ce325c12db40bb89b43e8e6e897052e96
+          ref: 1da89a12a8b00dc77506d702f61300803b6240b7
           path: SHARK-TestSuite
           submodules: false
           lfs: false
@@ -198,7 +198,7 @@ jobs:
         uses: actions/checkout@v4.1.7
         with:
           repository: nod-ai/SHARK-TestSuite
-          ref: a06e730ce325c12db40bb89b43e8e6e897052e96
+          ref: 1da89a12a8b00dc77506d702f61300803b6240b7
           path: SHARK-TestSuite
           submodules: false
           lfs: true
@@ -220,7 +220,7 @@ jobs:
           pytest \
             SHARK-TestSuite/iree_tests/pytorch/models \
             SHARK-TestSuite/iree_tests/sharktank \
-            -rpfE \
+            -rA \
             -k real_weights \
             --no-skip-tests-missing-files \
             --capture=no \

build_tools/pkgci/external_test_suite/models_gpu_rocm_gfx90a.json

@@ -20,9 +20,11 @@
     "pytorch/models/resnet50",
     "pytorch/models/sdxl-vae-decode-tank",
 
-    // error: 'builtin.module' op failed to run transform dialect passes
-    // (transform spec file is specific to SDXL?)
-    "sharktank/llama/open-llama-3b-v2-f16"
+    // TODO(#17874): error: a handle passed as operand #0 and consumed by this operation points to a payload entity more than once
+    "sharktank/llama/open-llama-3b-v2-f16",
+
+    "sharktank/punet/fp16",
+    "sharktank/punet/int8"
   ],
   "expected_run_failures": []
 }

build_tools/pkgci/external_test_suite/models_gpu_rocm_gfx942.json

@@ -20,9 +20,11 @@
     "pytorch/models/resnet50",
     "pytorch/models/sdxl-vae-decode-tank",
 
-    // error: 'builtin.module' op failed to run transform dialect passes
-    // (transform spec file is specific to SDXL?)
-    "sharktank/llama/open-llama-3b-v2-f16"
+      // TODO(#17874): error: a handle passed as operand #0 and consumed by this operation points to a payload entity more than once
+    "sharktank/llama/open-llama-3b-v2-f16",
+
+    "sharktank/punet/fp16",
+    "sharktank/punet/int8"
   ],
   "expected_run_failures": []
 }

build_tools/pkgci/external_test_suite/models_gpu_vulkan.json

@@ -12,13 +12,16 @@
     // TODO(#17344): need to regenerate .mlirbc
     "pytorch/models/opt-125M",
     "pytorch/models/resnet50",
+
     "pytorch/models/sdxl-prompt-encoder-tank",
     "pytorch/models/sdxl-scheduled-unet-3-tank",
     "pytorch/models/sdxl-vae-decode-tank",
 
-    // error: 'builtin.module' op failed to run transform dialect passes
-    // (transform spec file is specific to SDXL?)
-    "sharktank/llama/open-llama-3b-v2-f16"
+    // TODO(#17874): error: a handle passed as operand #0 and consumed by this operation points to a payload entity more than once
+    "sharktank/llama/open-llama-3b-v2-f16",
+
+    "sharktank/punet/fp16",
+    "sharktank/punet/int8"
   ],
   "expected_run_failures": []
 }