use fixed torch and torchvision

build.sh

@@ -0,0 +1,24 @@
+#!/bin/bash
+
+set -x
+set -e
+
+cmake -GNinja -Bbuild \
+    -DCMAKE_BUILD_TYPE=Release \
+    -DCMAKE_C_COMPILER=clang \
+    -DCMAKE_CXX_COMPILER=clang++ \
+    -DCMAKE_LINKER=lld \
+    -DLLVM_ENABLE_ASSERTIONS=ON \
+    -DLLVM_ENABLE_PROJECTS=mlir \
+    -DLLVM_EXTERNAL_PROJECTS="torch-mlir" \
+    -DLLVM_EXTERNAL_TORCH_MLIR_SOURCE_DIR="$(pwd)" \
+    -DLLVM_TARGETS_TO_BUILD=host \
+    -DMLIR_ENABLE_BINDINGS_PYTHON=ON \
+    -DTORCH_MLIR_USE_INSTALLED_PYTORCH=ON \
+    -DTORCH_MLIR_ENABLE_PYTORCH_EXTENSIONS=ON \
+    -DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=ON \
+    -DTORCH_MLIR_ENABLE_LTC=OFF \
+    $(pwd)/externals/llvm-project/llvm
+
+cmake --build build --target TorchMLIRPythonModules TorchMLIRJITIRImporterPybind check-torch-mlir-pt1 check-torch-mlir
+

build_tools/update_abstract_interp_lib.sh

@@ -41,7 +41,7 @@ if [ ! -z ${TORCH_MLIR_EXT_MODULES} ]; then
   ext_module="${TORCH_MLIR_EXT_MODULES} "
 fi
 
-PYTHONPATH="${pypath}" python \
+PYTHONPATH="${pypath}" python3 \
   -m torch_mlir.jit_ir_importer.build_tools.abstract_interp_lib_gen \
   --pytorch_op_extensions=${ext_module:-""} \
   --torch_transforms_cpp_dir="${torch_transforms_cpp_dir}"

build_tools/update_torch_ods.sh

@@ -42,7 +42,7 @@ if [ ! -z ${TORCH_MLIR_EXT_MODULES} ]; then
 fi
 
 set +u
-PYTHONPATH="${PYTHONPATH}:${pypath}" python \
+PYTHONPATH="${pypath}" python3 \
   -m torch_mlir.jit_ir_importer.build_tools.torch_ods_gen \
   --torch_ir_include_dir="${torch_ir_include_dir}" \
   --pytorch_op_extensions="${ext_module}" \

lib/Dialect/Torch/Transforms/AbstractInterpLibrary.cpp

@@ -11480,7 +11480,7 @@ StringRef mlir::torch::Torch::getAbstractInterpLibrary() {
 "      } else {\n"
 "        %11 = torch.aten.eq.int %2#1, %int15 : !torch.int, !torch.int -> !torch.bool\n"
 "        %12:2 = torch.prim.If %11 -> (!torch.bool, !torch.tuple<int, int>) {\n"
-"          %13 = torch.prim.TupleConstruct %1#1, %int6 : !torch.int, !torch.int -> !torch.tuple<int, int>\n"
+"          %13 = torch.prim.TupleConstruct %1#1, %int15 : !torch.int, !torch.int -> !torch.tuple<int, int>\n"
 "          torch.prim.If.yield %true, %13 : !torch.bool, !torch.tuple<int, int>\n"
 "        } else {\n"
 "          torch.prim.If.yield %false, %0 : !torch.bool, !torch.tuple<int, int>\n"

linux-requirements.txt

@@ -0,0 +1,4 @@
+--extra-index-url https://download.pytorch.org/whl/cpu
+--pre
+torch==2.4.0+cpu
+torchvision==0.19.0+cpu

macos-requirements.txt

@@ -0,0 +1,4 @@
+-f https://download.pytorch.org/whl/cpu
+--pre
+torch==2.4.0
+torchvision==0.19.0

projects/pt1/python/torch_mlir/jit_ir_importer/build_tools/abstract_interp_lib_gen.py

@@ -2710,7 +2710,7 @@ def aten〇_weight_norm_interface〡dtype(v_rank_dtype: Tuple[int, int], g_rank_
     elif g_dtype == torch.complex64:
         return v_dtype, torch.float32
     elif g_dtype == torch.bfloat16:
-        return v_dtype, torch.float32
+        return v_dtype, torch.bfloat16
     return v_dtype, g_dtype
 
 @check_dtype_function(_check_tensors_with_the_same_dtype(num_of_tensors=1))

test.sh

@@ -0,0 +1,3 @@
+set -x
+
+PYTHONPATH="$(pwd)/build/tools/torch-mlir/python_packages/torch_mlir:$(pwd)/projects/pt1" python3 -m e2e_testing.main --config="$1" $2 $3 $4