v2.2.3: fix contiguous, add msg if point vanish

CHANGELOG.md

@@ -1,4 +1,9 @@
 # Changelog
+## [2.2.3] - 2022-9-28
+### Fixed 
+- Fix missing .contiguous for input feature
+- Add some debug msg if points vanished.
+
 ## [2.2.2] - 2022-9-25
 ### Fixed 
 - Fix CI problem: main function too long and cause OOM in CI vm.

docs/COMMON_PROBLEMS.md

@@ -27,7 +27,7 @@ Your coordinates generate nothing with some conv params. Modify your conv params
 Example:
 
 Conv Params:
-```spatial shape=[8, 200, 200],ksize=[3, 3, 3],stride=[2, 2, 2],padding=[0, 1, 1],dilation=[1, 1, 1]```
+```spatial shape=[8, 200, 200],ksize=3,stride=2,padding=[0, 1, 1],dilation=1```
 Coordinates:
 ```
 [[0, 7, 153, 142]]

example/libspconv/README.md

@@ -6,6 +6,8 @@ libspconv + pybindings = "core_cc.so" in spconv python package.
 
 run ```run_build.sh``` to get ```libspconv.so```.
 
+see [inference code example](main.cu)
+
 ## libspconv API
 
 currently not available, but you can check python code to understand how to use C++ apis, spconv python and libspconv use same c++ code.

spconv/csrc/sparse/all.py

@@ -1856,7 +1856,28 @@ def get_indice_pairs_implicit_gemm(self):
                 }}
             }}
             // tv::ssprint("HASH SIZE", hash_size, num_act_out);
-
+            if (num_act_out == 0){{
+                std::stringstream ss;
+                ss << R"(Your points vanished here, this usually because you provide 
+conv params that may ignore some input points. Example: 
+    spatial_shape=[8, 200, 200]
+    ksize=3
+    stride=2
+    padding=[0, 1, 1]
+    dilation=1
+    Coordinates=[[0, 7, 153, 142]]
+these params will cause ALL points in z == 7 dropped because of padding_z=0.
+enlarge your spatial shape or change your conv param to make sure 
+every input point has a corresponding output point.
+Your Conv Params: )" << "\\n";
+                tv::sstream_print<'\\0'>(ss, "    spatial_shape=", input_dims, "\\n");
+                tv::sstream_print<'\\0'>(ss, "    ksize=", ksize, "\\n");
+                tv::sstream_print<'\\0'>(ss, "    stride=", stride, "\\n");
+                tv::sstream_print<'\\0'>(ss, "    padding=", padding, "\\n");
+                tv::sstream_print<'\\0'>(ss, "    dilation=", dilation, "\\n");
+                tv::ssprint(ss.str());
+                throw std::runtime_error(ss.str());
+            }}
             if (num_out_act_bound > 0 && num_act_out > num_out_act_bound){{
                 num_act_out = num_out_act_bound;
             }}
@@ -2092,6 +2113,29 @@ def get_indice_pairs(self):
 
                     // TODO pytorch unique may be faster?
                     num_act_out = apply_thrust_unique_to_indice_pairs_uniq(indice_pairs_uniq, thrustalloc, stream_int);
+                    if (num_act_out == 0){{
+                        std::stringstream ss;
+                        ss << R"(Your points vanished here, this usually because you provide 
+conv params that may ignore some input points. Example: 
+    spatial_shape=[8, 200, 200]
+    ksize=3
+    stride=2
+    padding=[0, 1, 1]
+    dilation=1
+    Coordinates=[[0, 7, 153, 142]]
+these params will cause ALL points in z == 7 dropped because of padding_z=0.
+enlarge your spatial shape or change your conv param to make sure 
+every input point has a corresponding output point.
+Your Conv Params: )" << "\\n";
+                        tv::sstream_print<'\\0'>(ss, "    spatial_shape=", input_dims, "\\n");
+                        tv::sstream_print<'\\0'>(ss, "    ksize=", ksize, "\\n");
+                        tv::sstream_print<'\\0'>(ss, "    stride=", stride, "\\n");
+                        tv::sstream_print<'\\0'>(ss, "    padding=", padding, "\\n");
+                        tv::sstream_print<'\\0'>(ss, "    dilation=", dilation, "\\n");
+                        tv::ssprint(ss.str());
+                        throw std::runtime_error(ss.str());
+                    }}
+
                     bool use_bound_algo = false;
                     int64_t num_out_bounded = num_act_out;
                     if (num_out_act_bound > 0 && num_act_out > num_out_act_bound){{

spconv/pytorch/conv.py

@@ -327,7 +327,9 @@ def forward(self, input: SparseConvTensor):
             out_tensor.spatial_shape = out_spatial_shape
             return out_tensor
         indice_dict = input.indice_dict.copy()
-
+        # only support contiguous tensor for now
+        if not features.is_contiguous():
+            features = features.contiguous()
         algo = self.algo
         if self.indice_key is not None:
             datas = input.find_indice_pair(self.indice_key)

spconv/pytorch/ops.py

@@ -51,6 +51,24 @@
 DEBUG_INT64_HASH_K = False
 INT32_MAX = SpconvOps.get_int32_max()
 
+_POINT_VANISH_MSG = """Your points vanished here, this usually because you provide 
+conv params that may ignore some input points. Example: 
+    spatial_shape=[8, 200, 200]
+    ksize=3
+    stride=2
+    padding=[0, 1, 1]
+    dilation=1
+    Coordinates=[[0, 7, 153, 142]]
+these params will cause ALL points in z == 7 dropped because of padding_z=0.
+enlarge your spatial shape or change your conv param to make sure 
+every input point has a corresponding output point.
+Your Conv Params: 
+    spatial_shape={}
+    ksize={}
+    stride={}
+    padding={}
+    dilation={}"""
+
 
 def get_conv_output_size(input_size, kernel_size, stride, padding, dilation):
     ndim = len(input_size)
@@ -239,6 +257,9 @@ def get_indice_pairs(indices: torch.Tensor,
                                                 stream_int=stream)
             uniq_res = indice_pairs_uniq.unique()
             num_act_out = uniq_res.shape[0] - 1
+            if (num_act_out == 0):
+                msg = _POINT_VANISH_MSG.format(spatial_shape, ksize, stride, padding, dilation)
+                raise ValueError(msg)
             use_bound_algo = False
             if num_out_act_bound > 0 and num_act_out > num_out_act_bound:
                 num_act_out = num_out_act_bound
@@ -611,6 +632,9 @@ def get_indice_pairs_implicit_gemm(
                 num_act_out = uniq_res.shape[0] - 1
                 uniq_out_indices_offset_tv = torch_tensor_to_tv(uniq_res)
                 raw_out_indices_offset_tv = indice_pairs_uniq_tv
+            if (num_act_out == 0):
+                msg = _POINT_VANISH_MSG.format(spatial_shape, ksize, stride, padding, dilation)
+                raise ValueError(msg)
 
         if num_out_act_bound > 0 and num_act_out > num_out_act_bound:
             num_act_out = num_out_act_bound
@@ -1080,6 +1104,15 @@ def indice_conv_backward(features: torch.Tensor,
                          algo: ConvAlgo = ConvAlgo.Native,
                          timer: CUDAKernelTimer = CUDAKernelTimer(False)):
     # print(out_bp.mean(), out_bp.max(), out_bp.min())
+    filters_shape = filters.shape
+    # TODO handle this in nn.Module to make sure features in backward is contiguous
+    if not features.is_contiguous():
+        features = features.contiguous()
+    if not out_bp.is_contiguous():
+        out_bp = out_bp.contiguous()
+    assert out_bp.is_contiguous()
+    assert filters.is_contiguous()
+    assert features.is_contiguous()
 
     if SPCONV_CPP_GEMM and GEMM_CPP is not None:
         alloc = TorchAllocator(features.device)
@@ -1110,16 +1143,6 @@ def indice_conv_backward(features: torch.Tensor,
         df = alloc.allocated[AllocKeys.DFilters]
         return din, df
 
-    filters_shape = filters.shape
-    # TODO handle this in nn.Module to make sure features in backward is contiguous
-    if not features.is_contiguous():
-        features = features.contiguous()
-    if not out_bp.is_contiguous():
-        out_bp = out_bp.contiguous()
-    assert out_bp.is_contiguous()
-    assert filters.is_contiguous()
-    assert features.is_contiguous()
-
     if not ALL_WEIGHT_IS_KRSC:
         kv_dim = 0
         is_KC_not_CK = not FILTER_HWIO
@@ -1438,6 +1461,10 @@ def implicit_gemm(features: torch.Tensor,
     if bias is not None:
         bias_tv = torch_tensor_to_tv(bias)
 
+    if not features.is_contiguous():
+        features = features.contiguous()
+    assert features.is_contiguous()
+    assert filters.is_contiguous()
 
     if SPCONV_CPP_GEMM and CONV_CPP is not None:
         alloc = TorchAllocator(features.device)
@@ -1477,10 +1504,6 @@ def implicit_gemm(features: torch.Tensor,
     # CONV.stream_synchronize(stream)
 
     # t = time.time()
-    if not features.is_contiguous():
-        features = features.contiguous()
-    assert features.is_contiguous()
-    assert filters.is_contiguous()
 
     if features.dtype == torch.int8 or features.dtype == torch.qint8:
         raise NotImplementedError("work in progress")

version.txt

@@ -1 +1 @@
-2.2.2
+2.2.3