Apply lintrunner

Author: neuropilot-captain

examples/mediatek/README.md

@@ -147,4 +147,3 @@ python3 eval_utils/eval_oss_result.py --eval_type piq --target_f edsr --output_f
     - `eval_type`: topk/piq/segmentation
     - `target_f`: folder contain golden data files. file name is `golden_<data_idx>_0.bin`
     - `output_f`: folder contain model output data files. file name is `output_<data_idx>_0.bin`
-

examples/mediatek/aot_utils/oss_utils/utils.py

@@ -6,11 +6,16 @@
 
 import os
 from typing import Optional
+
 import numpy as np
 
 import torch
 from executorch import exir
-from executorch.backends.mediatek import NeuropilotPartitioner, NeuropilotQuantizer, Precision
+from executorch.backends.mediatek import (
+    NeuropilotPartitioner,
+    NeuropilotQuantizer,
+    Precision,
+)
 from torch.ao.quantization.quantize_pt2e import convert_pt2e, prepare_pt2e
 
 
@@ -42,11 +47,17 @@ def build_executorch_binary(
 
     edge_compile_config = exir.EdgeCompileConfig(_check_ir_validity=False)
     # skipped op names are used for deeplabV3 model
-    neuro_partitioner = NeuropilotPartitioner([], op_names_to_skip={'aten_convolution_default_106', 'aten_convolution_default_107'})
+    neuro_partitioner = NeuropilotPartitioner(
+        [],
+        op_names_to_skip={
+            "aten_convolution_default_106",
+            "aten_convolution_default_107",
+        },
+    )
     edge_prog = to_edge_transform_and_lower(
-            aten_dialect,
-            compile_config=edge_compile_config,
-            partitioner=[neuro_partitioner],
+        aten_dialect,
+        compile_config=edge_compile_config,
+        partitioner=[neuro_partitioner],
     )
 
     exec_prog = edge_prog.to_executorch(
@@ -62,5 +73,3 @@ def make_output_dir(path: str):
             os.remove(os.path.join(path, f))
         os.removedirs(path)
     os.makedirs(path)
-
-

examples/mediatek/eval_utils/eval_oss_result.py

@@ -4,20 +4,23 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
-import sys
+import argparse
+import json
 import os
-import piq
+import sys
+
 import numpy as np
-import argparse
+import piq
 import torch
-import json
 
 
 def check_data(target_f, predict_f):
     target_files = os.listdir(target_f)
     predict_files = os.listdir(predict_f)
     if len(target_files) != len(predict_files):
-        raise RuntimeError("Data number in target folder and prediction folder must be same")
+        raise RuntimeError(
+            "Data number in target folder and prediction folder must be same"
+        )
 
     predict_set = set(predict_files)
     for f in target_files:
@@ -38,7 +41,7 @@ def eval_topk(target_f, predict_f):
     def solve(prob, target, k):
         _, indices = torch.topk(prob, k=k, sorted=True)
         golden = torch.reshape(target, [-1, 1])
-        correct = (golden == indices)
+        correct = golden == indices
         if torch.any(correct):
             return 1
         else:
@@ -62,18 +65,18 @@ def solve(prob, target, k):
 
 def eval_piq(target_f, predict_f):
     target_files = os.listdir(target_f)
-    
+
     psnr_list = []
     ssim_list = []
     for target_name in target_files:
         pred_name = target_name.replace("golden", "output")
         hr = np.fromfile(os.path.join(target_f, target_name), dtype=np.float32)
-        hr = hr.reshape((1,448,448,3))
+        hr = hr.reshape((1, 448, 448, 3))
         hr = np.moveaxis(hr, 3, 1)
         hr = torch.from_numpy(hr)
 
         sr = np.fromfile(os.path.join(predict_f, pred_name), dtype=np.float32)
-        sr = sr.reshape((1,448,448,3))
+        sr = sr.reshape((1, 448, 448, 3))
         sr = np.moveaxis(sr, 3, 1)
         sr = torch.from_numpy(sr).clamp(0, 1)
 
@@ -188,11 +191,9 @@ def histogram(golden, predict):
 
     check_data(args.target_f, args.out_f)
 
-    if args.eval_type == 'topk':
+    if args.eval_type == "topk":
         eval_topk(args.target_f, args.out_f)
-    elif args.eval_type == 'piq':
+    elif args.eval_type == "piq":
         eval_piq(args.target_f, args.out_f)
-    elif args.eval_type == 'segmentation':
+    elif args.eval_type == "segmentation":
         eval_segmentation(args.target_f, args.out_f)
-
-

examples/mediatek/executor_runner/mtk_oss_executor_runner.cpp

@@ -18,12 +18,12 @@
  * all fp32 tensors.
  */
 
+#include <cstdlib>
 #include <ctime>
-#include <iostream>
+#include <filesystem>
 #include <fstream>
+#include <iostream>
 #include <memory>
-#include <cstdlib>
-#include <filesystem>
 
 #include <gflags/gflags.h>
 
@@ -38,14 +38,23 @@
 static uint8_t method_allocator_pool[8 * 1024U * 1024U]; // 8 MB
 
 // Model Path
-DEFINE_string(model_path, "model.pte", "Model serialized in flatbuffer format. Default to 'model.pte'");
-DEFINE_string(input_list, "input_list.txt", "Model input list. Default to 'input_list.txt'");
-DEFINE_string(output_folder, "outputs", "Model output folder. Default to 'outputs'");
+DEFINE_string(
+    model_path,
+    "model.pte",
+    "Model serialized in flatbuffer format. Default to 'model.pte'");
+DEFINE_string(
+    input_list,
+    "input_list.txt",
+    "Model input list. Default to 'input_list.txt'");
+DEFINE_string(
+    output_folder,
+    "outputs",
+    "Model output folder. Default to 'outputs'");
 
 using namespace torch::executor;
 using torch::executor::MemoryAllocator;
-using torch::executor::util::FileDataLoader;
 using torch::executor::util::BufferCleanup;
+using torch::executor::util::FileDataLoader;
 using namespace std::filesystem;
 
 int main(int argc, char** argv) {
@@ -132,11 +141,12 @@ int main(int argc, char** argv) {
   // fast/small SRAM, or for memory associated with particular cores.
   std::vector<std::unique_ptr<uint8_t[]>> planned_buffers; // Owns the memory
   std::vector<Span<uint8_t>> planned_spans; // Passed to the allocator
-  size_t num_memory_planned_buffers = method_meta_result->num_memory_planned_buffers();
+  size_t num_memory_planned_buffers =
+      method_meta_result->num_memory_planned_buffers();
   for (size_t id = 0; id < num_memory_planned_buffers; ++id) {
     // .get() will always succeed because id < num_memory_planned_buffers.
-    size_t buffer_size =
-        static_cast<size_t>(method_meta_result->memory_planned_buffer_size(id).get());
+    size_t buffer_size = static_cast<size_t>(
+        method_meta_result->memory_planned_buffer_size(id).get());
     ET_LOG(Info, "Setting up planned buffer %zu, size %zu.", id, buffer_size);
     planned_buffers.push_back(std::make_unique<uint8_t[]>(buffer_size));
     planned_spans.push_back({planned_buffers.back().get(), buffer_size});
@@ -164,8 +174,8 @@ int main(int argc, char** argv) {
   std::ifstream input_list(FLAGS_input_list);
   ET_CHECK_MSG(
       input_list.is_open(),
-        "Error: cannot open input file %s",
-        FLAGS_input_list.c_str());
+      "Error: cannot open input file %s",
+      FLAGS_input_list.c_str());
 
   auto split = [](std::string s, std::string delimiter) {
     size_t pos_start = 0, pos_end, delim_len = delimiter.length();
@@ -225,7 +235,7 @@ int main(int argc, char** argv) {
           nbytes);
 
       fin.seekg(0, fin.beg);
-      fin.read(static_cast<char*> (data_ptr), file_size);
+      fin.read(static_cast<char*>(data_ptr), file_size);
       fin.close();
       inputs[num_allocated++] = data_ptr;
 
@@ -243,8 +253,7 @@ int main(int argc, char** argv) {
       Tensor tensor(&impl);
       Error ret = method->set_input(tensor, i);
       if (ret != Error::Ok) {
-        ET_LOG(
-            Error, "Failed to set input %zu: 0x%" PRIx32, i, (uint32_t)ret);
+        ET_LOG(Error, "Failed to set input %zu: 0x%" PRIx32, i, (uint32_t)ret);
         // The BufferCleanup will free the inputs when it goes out of scope.
         BufferCleanup cleanup({inputs, num_allocated});
         return 1;
@@ -258,14 +267,12 @@ int main(int argc, char** argv) {
     Error status = Error::Ok;
     status = method->execute();
     auto after_exec = std::chrono::high_resolution_clock::now();
-    double elapsed_time =
-        std::chrono::duration_cast<std::chrono::microseconds>(
-            after_exec - before_exec).count() / 1000.0;
-
-    ET_LOG(
-        Info,
-        "Inference took %f ms",
-        elapsed_time);
+    double elapsed_time = std::chrono::duration_cast<std::chrono::microseconds>(
+                              after_exec - before_exec)
+                              .count() /
+        1000.0;
+
+    ET_LOG(Info, "Inference took %f ms", elapsed_time);
     ET_CHECK_MSG(
         status == Error::Ok,
         "Execution of method %s failed with status 0x%" PRIx32,
@@ -282,11 +289,9 @@ int main(int argc, char** argv) {
     for (size_t i = 0; i < output_size; i++) {
       auto output_tensor = outputs[i].toTensor();
       auto output_file_name = FLAGS_output_folder + "/output_" +
-          std::to_string(inference_index) + "_" +
-          std::to_string(i) + ".bin";
+          std::to_string(inference_index) + "_" + std::to_string(i) + ".bin";
       std::ofstream fout(output_file_name.c_str(), std::ios::binary);
-      fout.write(
-          output_tensor.const_data_ptr<char>(), output_tensor.nbytes());
+      fout.write(output_tensor.const_data_ptr<char>(), output_tensor.nbytes());
       fout.close();
     }
 

examples/mediatek/model_export_scripts/deeplab_v3.py

@@ -4,21 +4,22 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
+import argparse
 import json
 import os
 import random
 import re
 import sys
+
 import numpy as np
-import argparse
 
 import torch
 from executorch.backends.mediatek import Precision
-from executorch.examples.models.deeplab_v3 import DeepLabV3ResNet101Model
 from executorch.examples.mediatek.aot_utils.oss_utils.utils import (
     build_executorch_binary,
     make_output_dir,
 )
+from executorch.examples.models.deeplab_v3 import DeepLabV3ResNet101Model
 
 
 class NhwcWrappedModel(torch.nn.Module):
@@ -112,7 +113,7 @@ def get_dataset(data_size, dataset_dir, download):
     for idx, data in enumerate(targets):
         file_name = f"{args.artifact}/golden_{idx}_0.bin"
         data.tofile(file_name)
-        if idx==0:
+        if idx == 0:
             print("golden shape: ", data.shape)
             print("golden type: ", data.dtype)
 
@@ -126,4 +127,3 @@ def get_dataset(data_size, dataset_dir, download):
         inputs,
         quant_dtype=Precision.A8W8,
     )
-

examples/mediatek/model_export_scripts/edsr.py

@@ -4,18 +4,19 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
+import argparse
 import json
 import os
+
 import numpy as np
-import argparse
 
 import torch
 from executorch.backends.mediatek import Precision
-from executorch.examples.models.edsr import EdsrModel
 from executorch.examples.mediatek.aot_utils.oss_utils.utils import (
     build_executorch_binary,
     make_output_dir,
 )
+from executorch.examples.models.edsr import EdsrModel
 
 from PIL import Image
 from torch.utils.data import Dataset
@@ -33,6 +34,7 @@ def forward(self, input1):
         nchw_output = self.edsr(nchw_input1)
         return nchw_output.permute(0, 2, 3, 1)
 
+
 class SrDataset(Dataset):
     def __init__(self, hr_dir: str, lr_dir: str):
         self.input_size = np.asanyarray([224, 224])
@@ -59,7 +61,11 @@ def __len__(self):
 
     def _resize_img(self, file: str, scale: int):
         with Image.open(file) as img:
-            return to_tensor(img.resize(tuple(self.input_size * scale))).unsqueeze(0).permute(0, 2, 3, 1)
+            return (
+                to_tensor(img.resize(tuple(self.input_size * scale)))
+                .unsqueeze(0)
+                .permute(0, 2, 3, 1)
+            )
 
     def get_input_list(self):
         input_list = ""
@@ -164,5 +170,3 @@ def get_dataset(hr_dir: str, lr_dir: str, default_dataset: str, dataset_dir: str
         [(input,) for input in inputs],
         quant_dtype=Precision.A8W8,
     )
-
-

examples/mediatek/model_export_scripts/inception_v3.py

@@ -4,18 +4,19 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
+import argparse
 import json
 import os
+
 import numpy as np
-import argparse
 
 import torch
 from executorch.backends.mediatek import Precision
-from executorch.examples.models.inception_v3 import InceptionV3Model
 from executorch.examples.mediatek.aot_utils.oss_utils.utils import (
     build_executorch_binary,
     make_output_dir,
 )
+from executorch.examples.models.inception_v3 import InceptionV3Model
 
 
 class NhwcWrappedModel(torch.nn.Module):
@@ -28,6 +29,7 @@ def forward(self, input1):
         output = self.inception(nchw_input1)
         return output
 
+
 def get_dataset(dataset_path, data_size):
     from torchvision import datasets, transforms
 
@@ -55,7 +57,7 @@ def get_data_loader():
         if index >= data_size:
             break
         feature, target = data
-        feature = feature.permute(0, 2, 3, 1) # NHWC
+        feature = feature.permute(0, 2, 3, 1)  # NHWC
         inputs.append((feature,))
         targets.append(target)
         input_list += f"input_{index}_0.bin\n"
@@ -120,5 +122,3 @@ def get_data_loader():
         inputs,
         quant_dtype=Precision.A8W8,
     )
-
-