fix coverity issues (#1263)

Author: blzheng

csrc/cpu/aten/Matmul.cpp

@@ -373,11 +373,11 @@ at::Tensor matmul_impl(
     // We are multiplying b1 x n x m1 by x2 x m2 x p (where b1 can be a list);
     // we track m1 vs m2 separately even though they must match for nicer error
     // messages
-    int64_t n = dim_tensor1 > 1 ? tensor1.size(-2) : 1;
+    int64_t n = tensor1.size(-2);
     int64_t m1 = tensor1.size(-1);
     c10::IntArrayRef batch_tensor1(
         tensor1.sizes().data(), std::max<int64_t>(dim_tensor1 - 2, 0));
-    int64_t m2 = dim_tensor2 > 1 ? tensor2.size(-2) : 1;
+    int64_t m2 = tensor2.size(-2);
     int64_t p = tensor2.size(-1);
     c10::IntArrayRef batch_tensor2(
         tensor2.sizes().data(), std::max<int64_t>(dim_tensor2 - 2, 0));

csrc/cpu/utils/fpmath_mode.cpp

@@ -21,7 +21,7 @@ dnnl_fpmath_mode_t fpmath_mode = []() {
 }();
 
 void setFP32MathModeCpu(FP32MathMode m) {
-  dnnl_fpmath_mode_t mode;
+  dnnl_fpmath_mode_t mode = dnnl_fpmath_mode_strict;
   if (m == FP32MathMode::FP32) {
     mode = dnnl_fpmath_mode_strict;
   } else if (m == FP32MathMode::BF32) {

csrc/jit/cpu/kernels/Einsum.cpp

@@ -293,8 +293,7 @@ static Tensor sumproduct_pair(
   auto min_size = std::min(lo_count, ro_count);
   auto max_size = std::max(lo_count, ro_count);
   for (int i = dim - 1; i >= dim - max_size; i--) {
-    if (left_shape[i] == 1 && left_shape[i] == right_shape[i] &&
-        min_size >= 0) {
+    if (left_shape[i] == 1 && left_shape[i] == right_shape[i]) {
       right_shape.pop_back();
       left_shape.pop_back();
       min_size--;

intel_extension_for_pytorch/quantization/_quantization_state_utils.py

@@ -1,7 +1,6 @@
 import dataclasses
 from typing import Callable, Tuple, Any, List, Optional, Dict
 import torch
-import torch.nn as nn
 import torch.nn.functional as F
 import torch.nn.quantized.dynamic as nnqd
 from intel_extension_for_pytorch.nn.functional import interaction
@@ -344,7 +343,7 @@ def iterate_and_apply_convert(
                     args = args.to(torch.float32)
                     args = torch.quantize_per_channel(args, scale, zp, ch_axis, dtype)
                     args = args.dequantize()
-                    args = arg.to(torch.bfloat16)
+                    args = args.to(torch.bfloat16)
                 else:
                     args = torch.quantize_per_channel(args, scale, zp, ch_axis, dtype)
                     args = args.dequantize()

setup.py

@@ -66,8 +66,6 @@
 import subprocess
 import sys
 from pathlib import Path
-import warnings
-import urllib.request
 import re
 
 try:
@@ -132,7 +130,6 @@
         mkl_header = glob.glob(f'{mkl_install_dir}/include/**/mkl_version.h', recursive = True)
         if len(mkl_header) == 0:
             raise RuntimeError(f'{mkl_install_dir} doesn\'t seem to be a valid MKL library directory.\n{" ":14}mkl_version.h not found.')
-            mkl_install_dir = ''
         else:
             mkl_header = mkl_header[0]
             mkl_major = 0
@@ -152,11 +149,9 @@
             mkl_version = f'{mkl_major}.{mkl_minor}.{mkl_patch}'
             if pkg_ver.parse(mkl_version) < pkg_ver.parse('2021.0.0'):
                 raise RuntimeError(f'MKL version({mkl_version}) is not supported. Please use MKL later than 2021.0.0.')
-                mkl_install_dir = ''
             mkl_library = glob.glob(f'{mkl_install_dir}/lib/**/libmkl_core.a', recursive = True)
             if len(mkl_library) == 0:
                 raise RuntimeError(f'libmkl_core.a not found in {mkl_install_dir}/lib/intel64.')
-                mkl_install_dir = ''
     if not mkl_install_dir:
         subprocess.check_call([sys.executable, '-m', 'pip', 'install', 'mkl-include>=2021.0.0'])
         subprocess.check_call([sys.executable, '-m', 'pip', 'install', '--no-deps', 'mkl-static>=2021.0.0'])

tests/cpu/test_autocast.py

@@ -557,7 +557,7 @@ def _test_lstm_pack_padded_sequence(self):
         batch_size = 24
         num_layers = 1
         bidirectional = True
-        num_direc = 2 if bidirectional else 1
+        num_direc = 2
         max_lens = 96
 
         sent = torch.randn(batch_size, max_lens, embedding_dim)

tests/cpu/test_softmax.py

@@ -1,7 +1,6 @@
 import torch
 import torch.nn as nn
 import intel_extension_for_pytorch as ipex
-from common_utils import TestCase
 from torch.testing._internal.jit_utils import JitTestCase
 import intel_extension_for_pytorch as ipex
 from intel_extension_for_pytorch.quantization import prepare, convert
@@ -40,10 +39,7 @@ def __init__(self):
         super().__init__()
     def forward(self, x, flag):
         if flag:
-          if flag:
             x1 = x + 1
-          else:
-            x1 = x + 2
         else:
             x1 = x + 3
         x2 = torch.softmax(x1, dim=-1)

tests/cpu/test_weight_prepack.py

@@ -1110,10 +1110,7 @@ def forward(self, x, h=None):
 
         num_directions = 2 if bidirectional else 1
 
-        if batch_first:
-            input = torch.randn(batch_size, seq_len, input_size)
-        else:
-            input = torch.randn(seq_len, batch_size, input_size)
+        input = torch.randn(batch_size, seq_len, input_size)
         h = torch.randn(num_layers * num_directions, batch_size, hidden_size)
         c = torch.randn(num_layers * num_directions, batch_size, hidden_size)