/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/tests/python/contrib/test_arm_compute_lib/ |
H A D | test_dense.py | 71 weight_shape, argument 192 for dtype, (shape, weight_shape, units), composite in trials: 203 "weight_shape": weight_shape, 222 for dtype, (shape, weight_shape, units), composite in trials: 225 args = (shape, weight_shape, units, dtype) 246 for dtype, (shape, weight_shape, units), composite in trials: 254 input_zp, input_sc, kernel_zp, kernel_sc, weight_shape[0], weight_shape[1] 259 weight_shape, 277 "weight_shape": weight_shape, 304 args = (shape, weight_shape, units, dtype) [all …]
|
H A D | test_conv2d.py | 60 weight_shape = (kernel_h, kernel_w, shape[3] // groups, channels) 61 w = tvm.nd.array(np.random.uniform(-128, 127, weight_shape).astype(dtype)) 78 b = tvm.nd.array(np.random.uniform(-128, 127, weight_shape[3]).astype(dtype)) 137 weight_shape = (kernel_h, kernel_w, shape[3] // groups, channels) 138 w = tvm.nd.array(np.random.uniform(0, 255, weight_shape).astype(dtype)) 159 b = tvm.nd.array(np.random.uniform(0, 255, weight_shape[3]).astype("int32")) 191 weight_shape = (channels, kernel_h, kernel_w, shape[3] // groups) 226 "attrs": {"shape": [[list(weight_shape)]], "dtype": [[str(dtype)]]}, 249 "attrs": {"shape": [[[weight_shape[0]]]], "dtype": [[bias_dtype]]},
|
/dports/lang/halide/Halide-release_2019_08_27-2654-g664dc4993/apps/resnet_50/ |
H A D | Resnet50Generator.cpp | 254 int p = weight_shape.pad; in conv2D() 262 RDom r(0, input.shape[0], 0, weight_shape.w, 0, weight_shape.h); in conv2D() 264 …(c, i, j) += weights(c, r.y, r.z, r.x) * padded(r.x, weight_shape.stride * i + r.y - p, weight_sha… in conv2D() 269 output.shape = compute_shape(input, weight_shape); in conv2D() 299 int p = weight_shape.pad; in max_pool_layer() 306 RDom r(0, weight_shape.w, 0, weight_shape.h); in max_pool_layer() 308 …pool(c, i, j) = maximum(padded(c, weight_shape.stride * i + r.x - p, weight_shape.stride * j + r.y… in max_pool_layer() 318 int p = weight_shape.pad; in avg_pool_layer() 325 RDom r(0, weight_shape.w, 0, weight_shape.h); in avg_pool_layer() 326 float scale = weight_shape.w * weight_shape.h; in avg_pool_layer() [all …]
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/src/relay/transforms/ |
H A D | convert_sparse_dense.cc | 76 const Array<Array<PrimExpr> >& weight_shape) in DenseToSparseDenseMutator() 78 CHECK_EQ(weight_name.size(), weight_shape.size()); in DenseToSparseDenseMutator() 82 const auto& ws = weight_shape[i]; in DenseToSparseDenseMutator() 122 const Array<Array<PrimExpr> >& weight_shape) { in DenseToSparse() 123 auto rewriter = DenseToSparseDenseMutator(weight_name, weight_shape); in DenseToSparse() 130 const Array<Array<PrimExpr> >& weight_shape) { in DenseToSparse() 134 auto f0 = Downcast<Function>(DenseToSparse(f, weight_name, weight_shape)); in DenseToSparse()
|
/dports/misc/glow/glow-f24d960e3cc80db95ac0bc17b1900dbf60ca044a/thirdparty/onnx/onnx/optimizer/passes/ |
H A D | fuse_add_bias_into_conv.h | 55 auto weight_shape = orig_conv->node()->inputs()[1]->sizes(); in runTransform() local 64 if (weight_shape.size() > 0 && weight_shape[0].is_int) { in runTransform() 65 ONNX_ASSERT(M == -1 || M == weight_shape[0].dim); in runTransform() 66 M = weight_shape[0].dim; in runTransform() 68 rank == -1 || rank == static_cast<int64_t>(weight_shape.size())); in runTransform() 69 rank = weight_shape.size(); in runTransform()
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/python/tvm/relay/analysis/ |
H A D | sparse_dense.py | 76 memo = SparseAnalysisResult(weight_name=[], weight_shape=[]) 87 memo.weight_shape.append( 97 weight_shape=tvm.runtime.convert(memo.weight_shape),
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/tests/python/contrib/test_ethosn/ |
H A D | test_fullyconnected.py | 28 shape, weight_shape, input_zp, input_sc, kernel_zp, kernel_sc, output_zp, output_sc, dtype argument 32 w = tvm.nd.array(np.ones(weight_shape, dtype)) 41 units=weight_shape[0], 129 weight_shape, 144 weight_shape,
|
H A D | test_conv2d.py | 75 weight_shape = (kernel_h, kernel_w, shape[3] // groups, out_channels) 77 weight_shape = (kernel_h, kernel_w, out_channels, 1) 80 np.iinfo(dtype).min, high=np.iinfo(dtype).max, size=weight_shape, dtype=dtype
|
/dports/misc/py-mxnet/incubator-mxnet-1.9.0/src/operator/nn/cudnn/ |
H A D | cudnn_algoreg-inl.h | 129 mxnet::TShape data_shape, weight_shape, out_shape; member 139 this->weight_shape == other.weight_shape && 154 ret = dmlc::HashCombine(ret, key.weight_shape); in operator()
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/src/operator/nn/cudnn/ |
H A D | cudnn_algoreg-inl.h | 129 mxnet::TShape data_shape, weight_shape, out_shape; member 139 this->weight_shape == other.weight_shape && 154 ret = dmlc::HashCombine(ret, key.weight_shape); in operator()
|
/dports/www/chromium-legacy/chromium-88.0.4324.182/chrome/browser/resource_coordinator/tab_ranker/ |
H A D | native_inference.cc | 32 const int32_t* __restrict weight_shape, in FullyConnected() argument 38 const int32_t num_inputs = weight_shape[0]; in FullyConnected() 39 const int32_t num_outputs = weight_shape[1]; in FullyConnected()
|
H A D | pairwise_inference.cc | 32 const int32_t* __restrict weight_shape, in FullyConnected() argument 38 const int32_t num_inputs = weight_shape[0]; in FullyConnected() 39 const int32_t num_outputs = weight_shape[1]; in FullyConnected()
|
/dports/misc/tvm/incubator-tvm-0.6.1/src/relay/pass/ |
H A D | mac_count.cc | 139 Array<IndexExpr> weight_shape = weight_type->shape; in DenseMacCount() local 140 CHECK(data_shape.size() == 2 && weight_shape.size() == 2) in DenseMacCount() 144 int64_t d3 = static_cast<int64_t>(weight_shape[0].as<IntImm>()->value); in DenseMacCount() 145 int64_t d4 = static_cast<int64_t>(weight_shape[1].as<IntImm>()->value); in DenseMacCount()
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/src/relay/analysis/ |
H A D | mac_count.cc | 132 Array<IndexExpr> weight_shape = weight_type->shape; in DenseMacCount() local 133 CHECK(data_shape.size() == 2 && weight_shape.size() == 2) in DenseMacCount() 137 int64_t d3 = static_cast<int64_t>(weight_shape[0].as<IntImmNode>()->value); in DenseMacCount() 138 int64_t d4 = static_cast<int64_t>(weight_shape[1].as<IntImmNode>()->value); in DenseMacCount()
|
/dports/misc/py-tvm/incubator-tvm-0.6.1/src/relay/pass/ |
H A D | mac_count.cc | 139 Array<IndexExpr> weight_shape = weight_type->shape; in DenseMacCount() local 140 CHECK(data_shape.size() == 2 && weight_shape.size() == 2) in DenseMacCount() 144 int64_t d3 = static_cast<int64_t>(weight_shape[0].as<IntImm>()->value); in DenseMacCount() 145 int64_t d4 = static_cast<int64_t>(weight_shape[1].as<IntImm>()->value); in DenseMacCount()
|
/dports/www/qt5-webengine/qtwebengine-everywhere-src-5.15.2/src/3rdparty/chromium/ui/events/ozone/evdev/touch_filter/palm_model/ |
H A D | onedevice_train_palm_detection_filter_inference.cc | 178 const int32_t* __restrict weight_shape, in MatMul() argument 186 ConstMatrixMap<T>(weight_values, weight_shape[1], weight_shape[0]); in MatMul() 191 const int32_t num_inputs = weight_shape[0]; in MatMul() 192 const int32_t num_outputs = weight_shape[1]; in MatMul() 340 ConstMatrixMap<T>(weight_values, weight_shape[1], weight_shape[0]); in FullyConnected() 346 const int32_t num_inputs = weight_shape[0]; in FullyConnected() 347 const int32_t num_outputs = weight_shape[1]; in FullyConnected() 374 const int32_t num_rows = weight_shape[1]; in SparseDenseMatmulCSR() 375 const int32_t num_cols = weight_shape[0]; in SparseDenseMatmulCSR() 396 const int32_t num_rows = weight_shape[1]; in SparseFullyConnectedCSR() [all …]
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/python/tvm/relay/frontend/ |
H A D | qnn_torch.py | 480 weight_shape = infer_shape(weight) 481 kernel_size = (weight_shape[2], weight_shape[3]) 482 out_channels = weight_shape[0] 538 weight_shape = infer_shape(weight) 546 units=weight_shape[0],
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/src/runtime/contrib/dnnl/ |
H A D | dnnl_json_runtime.cc | 158 dnnl::memory::dims weight_shape = nodes_[weight_entry.id_].GetOpShape()[weight_entry.index_]; in Conv2d() local 167 OC = weight_shape[0], // output channels in Conv2d() 168 KH = weight_shape[2], // weight height in Conv2d() 169 KW = weight_shape[3], // weight width in Conv2d() 253 dnnl::memory::dims weight_shape = nodes_[weight_entry.id_].GetOpShape()[weight_entry.index_]; in Dense() local 257 OC = weight_shape[0]; // output channels in Dense()
|
/dports/misc/tvm/incubator-tvm-0.6.1/vta/tutorials/optimize/ |
H A D | matrix_multiply_opt.py | 111 weight_shape = (out_channels // env.BLOCK_OUT, variable 127 weight = tvm.placeholder(weight_shape, name="weight", dtype=env.wgt_dtype) 133 weight_buf = tvm.compute(weight_shape,
|
/dports/misc/py-tvm/incubator-tvm-0.6.1/vta/tutorials/optimize/ |
H A D | matrix_multiply_opt.py | 111 weight_shape = (out_channels // env.BLOCK_OUT, variable 127 weight = tvm.placeholder(weight_shape, name="weight", dtype=env.wgt_dtype) 133 weight_buf = tvm.compute(weight_shape,
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/python/tvm/relay/data_dep_optimization/ |
H A D | bsr_dense.py | 52 func, relay.transform.DenseToSparse(weight_info.weight_name, weight_info.weight_shape)
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/vta/tutorials/optimize/ |
H A D | matrix_multiply_opt.py | 109 weight_shape = ( variable 124 weight = te.placeholder(weight_shape, name="weight", dtype=env.wgt_dtype) 128 weight_buf = te.compute(weight_shape, lambda *i: weight(*i), "weight_buf")
|
/dports/misc/tvm/incubator-tvm-0.6.1/python/tvm/relay/op/nn/ |
H A D | _nn.py | 165 weight_shape = get_const_tuple(inputs[1].shape) 167 return weight_shape[2] * weight_shape[3] 168 return weight_shape[0] * weight_shape[1] 930 def _dense_shape_func(data_shape, weight_shape): argument 934 out[out.shape[0] - 1] = weight_shape[0]
|
/dports/misc/py-tvm/incubator-tvm-0.6.1/python/tvm/relay/op/nn/ |
H A D | _nn.py | 165 weight_shape = get_const_tuple(inputs[1].shape) 167 return weight_shape[2] * weight_shape[3] 168 return weight_shape[0] * weight_shape[1] 930 def _dense_shape_func(data_shape, weight_shape): argument 934 out[out.shape[0] - 1] = weight_shape[0]
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/python/tvm/relay/transform/ |
H A D | transform.py | 984 def DenseToSparse(weight_name, weight_shape): argument 1003 return _ffi_api.DenseToSparse(weight_name, weight_shape)
|