[tensorexpr][nnc] Support quantization #66676
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
[ghstack-poisoned]
CI Flow Status⚛️ CI FlowRuleset - Version:
You can add a comment to the PR and tag @pytorchbot with the following commands: # ciflow rerun, "ciflow/default" will always be added automatically
@pytorchbot ciflow rerun
# ciflow rerun with additional labels "-l <ciflow/label_name>", which is equivalent to adding these labels manually and trigger the rerun
@pytorchbot ciflow rerun -l ciflow/scheduled -l ciflow/slowFor more information, please take a look at the CI Flow Wiki. |
This was referenced Oct 15, 2021
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit ff4c203 (more details on the Dr. CI page):
ci.pytorch.org: 1 failedThis comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
facebook-github-bot
added
the
oncall: jit
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Closed
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Closed
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Differential Revision: [D31676329](https://our.internmc.facebook.com/intern/diff/D31676329) 1. Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8 2. Introducing quantization lowerings for: ``` aten::quantize_per_tensor aten::dequantize _quantized::conv2d quantized::conv2d_relu quantized::add aten::upsample_nearest2d ``` `quantize_per_tensor` and `dequantize` can be lowered as tensorexpr or external calls. all other ops are lowered only as external calls. Testing: ``` #!/bin/bash PYTORCH_JIT_LOG_LEVEL=">>kernel:>>eval" \ PYTORCH_TENSOREXPR_DONT_USE_LLVM=1 \ USE_PYTORCH_QNNPACK=1 \ USE_QNNPACK=1 \ USE_FBGEMM=0 \ DEBUG=1 \ CXX_FLAGS="-g" \ USE_LLVM=/home/ivankobzarev/llvm90install \ USE_XNNPACK=1 \ ./scripts/build_local.sh \ -DBUILD_BINARY=ON \ -DCMAKE_BUILD_TYPE=Debug \ -DBUILD_TEST=ON \ -DUSE_LLVM=/home/ivankobzarev/llvm90install \ && ./build/bin/test_tensorexpr --gtest_filter="Quantization.*" ``` [ghstack-poisoned]
|
@IvanKobzarev has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
|
@IvanKobzarev merged this pull request in 7fbcf79. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack from ghstack:
Differential Revision: D31676329
Introuducing quantized dtypes to tensorexpr: QUInt8, QInt8
Introducing quantization lowerings for:
quantize_per_tensoranddequantizecan be lowered as tensorexpr or external calls.all other ops are lowered only as external calls.
Testing: