github/AMD-SHARK-Studio
1
1
Fork 0
mirror of https://github.com/nod-ai/AMD-SHARK-Studio.git synced 2026-04-03 03:00:17 -04:00
Code Issues Packages Projects Releases Wiki Activity

Remove flags that are default in upstream IREE (#1785)

Browse Source 
* Remove index bits flags now set by default

* Update shark_studio_imports.py
This commit is contained in:
Ean Garvey
2023-08-24 11:57:54 -05:00
committed by GitHub
parent e37d6720eb
commit 8e3860c9e6
6 changed files with 9 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
apps/stable_diffusion/profiling_with_iree.md
View File

@@ -7,16 +7,16 @@ Compile Commands FP32/FP16:
```shell
Vulkan AMD:
iree-compile --iree-input-type=none --iree-hal-target-backends=vulkan --iree-vulkan-target-triple=rdna2-unknown-linux --iree-stream-resource-index-bits=64 --iree-vm-target-index-bits=64 /path/to/input/mlir -o /path/to/output/vmfb
iree-compile --iree-input-type=none --iree-hal-target-backends=vulkan --iree-vulkan-target-triple=rdna2-unknown-linux /path/to/input/mlir -o /path/to/output/vmfb
# add --mlir-print-debuginfo --mlir-print-op-on-diagnostic=true for debug
# use –iree-input-type=auto or "mhlo_legacy" or "stablehlo" for TF models
CUDA NVIDIA:
iree-compile --iree-input-type=none --iree-hal-target-backends=cuda --iree-stream-resource-index-bits=64 --iree-vm-target-index-bits=64 /path/to/input/mlir -o /path/to/output/vmfb
iree-compile --iree-input-type=none --iree-hal-target-backends=cuda /path/to/input/mlir -o /path/to/output/vmfb
CPU:
iree-compile --iree-input-type=none --iree-hal-target-backends=llvm-cpu --iree-stream-resource-index-bits=64 --iree-vm-target-index-bits=64 /path/to/input/mlir -o /path/to/output/vmfb
iree-compile --iree-input-type=none --iree-hal-target-backends=llvm-cpu /path/to/input/mlir -o /path/to/output/vmfb
```

1
apps/stable_diffusion/shark_studio_imports.py
View File

@@ -45,6 +45,7 @@ datas += collect_data_files("iree")
datas += collect_data_files("google_cloud_storage")
datas += collect_data_files("shark", include_py_files=True)
datas += collect_data_files("timm", include_py_files=True)
datas += collect_data_files("tqdm")
datas += collect_data_files("tkinter")
datas += collect_data_files("webview")
datas += collect_data_files("sentencepiece")

8
cpp/README.md
View File

@@ -40,7 +40,7 @@ cmake --build build/
*Prepare the model*
```bash
wget https://storage.googleapis.com/shark_tank/latest/resnet50_tf/resnet50_tf.mlir
iree-compile --iree-input-type=auto --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --iree-llvmcpu-embedded-linker-path=`python3 -c 'import sysconfig; print(sysconfig.get_paths()["purelib"])'`/iree/compiler/tools/../_mlir_libs/iree-lld --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --mlir-pass-pipeline-crash-reproducer=ist/core-reproducer.mlir --iree-llvmcpu-target-cpu-features=host -iree-vulkan-target-triple=rdna2-unknown-linux --iree-stream-resource-index-bits=64 --iree-vm-target-index-bits=64 resnet50_tf.mlir -o resnet50_tf.vmfb
iree-compile --iree-input-type=auto --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --iree-llvmcpu-embedded-linker-path=`python3 -c 'import sysconfig; print(sysconfig.get_paths()["purelib"])'`/iree/compiler/tools/../_mlir_libs/iree-lld --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --mlir-pass-pipeline-crash-reproducer=ist/core-reproducer.mlir --iree-llvmcpu-target-cpu-features=host -iree-vulkan-target-triple=rdna2-unknown-linux resnet50_tf.mlir -o resnet50_tf.vmfb
```
*Prepare the input*
@@ -65,18 +65,18 @@ A tool for benchmarking other models is built and can be invoked with a command
see `./build/vulkan_gui/iree-vulkan-gui --help` for an explanation on the function input. For example, stable diffusion unet can be tested with the following commands:
```bash
wget https://storage.googleapis.com/shark_tank/quinn/stable_diff_tf/stable_diff_tf.mlir
iree-compile --iree-input-type=auto --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --iree-llvmcpu-target-cpu-features=host -iree-vulkan-target-triple=rdna2-unknown-linux --iree-stream-resource-index-bits=64 --iree-vm-target-index-bits=64 stable_diff_tf.mlir -o stable_diff_tf.vmfb
iree-compile --iree-input-type=auto --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --iree-llvmcpu-target-cpu-features=host -iree-vulkan-target-triple=rdna2-unknown-linux stable_diff_tf.mlir -o stable_diff_tf.vmfb
./build/vulkan_gui/iree-vulkan-gui --module-file=stable_diff_tf.vmfb --function_input=2x4x64x64xf32 --function_input=1xf32 --function_input=2x77x768xf32
```
VAE and Autoencoder are also available
```bash
# VAE
wget https://storage.googleapis.com/shark_tank/quinn/stable_diff_tf/vae_tf/vae.mlir
iree-compile --iree-input-type=auto --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --iree-llvmcpu-target-cpu-features=host -iree-vulkan-target-triple=rdna2-unknown-linux --iree-stream-resource-index-bits=64 --iree-vm-target-index-bits=64 vae.mlir -o vae.vmfb
iree-compile --iree-input-type=auto --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --iree-llvmcpu-target-cpu-features=host -iree-vulkan-target-triple=rdna2-unknown-linux vae.mlir -o vae.vmfb
./build/vulkan_gui/iree-vulkan-gui --module-file=stable_diff_tf.vmfb --function_input=1x4x64x64xf32
# CLIP Autoencoder
wget https://storage.googleapis.com/shark_tank/quinn/stable_diff_tf/clip_tf/clip_autoencoder.mlir
iree-compile --iree-input-type=auto --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --iree-llvmcpu-target-cpu-features=host -iree-vulkan-target-triple=rdna2-unknown-linux --iree-stream-resource-index-bits=64 --iree-vm-target-index-bits=64 clip_autoencoder.mlir -o clip_autoencoder.vmfb
iree-compile --iree-input-type=auto --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --iree-llvmcpu-target-cpu-features=host -iree-vulkan-target-triple=rdna2-unknown-linux clip_autoencoder.mlir -o clip_autoencoder.vmfb
./build/vulkan_gui/iree-vulkan-gui --module-file=stable_diff_tf.vmfb --function_input=1x77xi32 --function_input=1x77xi32
```

2
docs/shark_iree_profiling.md
View File

@@ -55,7 +55,7 @@ The command line for compilation will start something like this, where the `-` n
The `-o output_filename.vmfb` flag can be used to specify the location to save the compiled vmfb. Note that a dump of the
dispatches that can be compiled + run in isolation can be generated by adding `--iree-hal-dump-executable-benchmarks-to=/some/directory`. Say, if they are in the `benchmarks` directory, the following compile/run commands would work for Vulkan on RDNA3.
```
iree-compile --iree-input-type=none --iree-hal-target-backends=vulkan --iree-vulkan-target-triple=rdna3-unknown-linux --iree-stream-resource-index-bits=64 --iree-vm-target-index-bits=64 benchmarks/module_forward_dispatch_${NUM}_vulkan_spirv_fb.mlir -o benchmarks/module_forward_dispatch_${NUM}_vulkan_spirv_fb.vmfb
iree-compile --iree-input-type=none --iree-hal-target-backends=vulkan --iree-vulkan-target-triple=rdna3-unknown-linux benchmarks/module_forward_dispatch_${NUM}_vulkan_spirv_fb.mlir -o benchmarks/module_forward_dispatch_${NUM}_vulkan_spirv_fb.vmfb
iree-benchmark-module --module=benchmarks/module_forward_dispatch_${NUM}_vulkan_spirv_fb.vmfb --function=forward --device=vulkan
```

2
shark/iree_utils/compile_utils.py
View File

@@ -94,9 +94,7 @@ def get_iree_frontend_args(frontend):
# Common args to be used given any frontend or device.
def get_iree_common_args():
return [
"--iree-stream-resource-index-bits=64",
"--iree-stream-resource-max-allocation-size=4294967295",
"--iree-vm-target-index-bits=64",
"--iree-vm-bytecode-module-strip-source-map=true",
"--iree-util-zero-fill-elided-attrs",
]

2
tank/examples/bert_tf/bert_large_run.py
View File

@@ -85,8 +85,6 @@ if __name__ == "__main__":
args = [
"--iree-llvmcpu-target-cpu-features=host",
"--iree-mhlo-demote-i64-to-i32=false",
"--iree-stream-resource-index-bits=64",
"--iree-vm-target-index-bits=64",
]
backend_config = "dylib"
# backend = "cuda"