tkonolige commented on a change in pull request #6685:
URL: https://github.com/apache/incubator-tvm/pull/6685#discussion_r506601559
##########
File path: tests/python/frontend/tensorflow/test_forward.py
##########
@@ -1750,6 +1750,64 @@ def test_forward_batch_matmul():
_test_batch_matmul((2, 3, 4, 2, 3, 4, 5, 6), (2, 3, 4, 2, 3, 4, 5, 6),
"float32", False, True)
+#######################################################################
+# SparseTensorDenseMatMul
+# ----------------------------------
+
+
+def _test_sparse_dense_matmul(indices, values, A_shape, B_shape, dtype,
flip=False):
Review comment:
You pass in indices and values here, but then you don't use them? Or am
I missing something?
##########
File path: python/tvm/relay/frontend/tensorflow.py
##########
@@ -890,6 +890,44 @@ def _impl(inputs, attr, params, mod):
return _impl
+def _sparse_tensor_dense_matmul():
+ # Sparse utility from Numpy
+ from scipy import sparse
+
+ def _impl(inputs, attr, params, mod):
+ assert len(inputs) == 4, "There should be 4 input tensors"
+
+ indices_tensor = _infer_value(inputs[0], params, mod).asnumpy()
+ values_tensor = _infer_value(inputs[1], params, mod).asnumpy()
+ dense_shape_tensor = _infer_value(inputs[2], params, mod).asnumpy()
+
+ data = inputs[3]
+
+ rows = [x[0] for x in indices_tensor]
+ cols = [x[1] for x in indices_tensor]
+
+ # Create Numpy sparse Tensor(CSR)
+ weight_sp = sparse.csr_matrix(
+ (values_tensor, (rows, cols)),
shape=tuple(dense_shape_tensor.tolist())
+ )
+ weight_sp = sparse.csr_matrix(weight_sp.transpose())
+
+ weight_data = _expr.const(weight_sp.data, weight_sp.data.dtype)
+ weight_indptrs = _expr.const(weight_sp.indptr, weight_sp.indptr.dtype)
+ weight_indices = _expr.const(weight_sp.indices,
weight_sp.indices.dtype)
+
+ ret = _op.nn.sparse_dense(data, [weight_data, weight_indices,
weight_indptrs])
Review comment:
Reading the Tensorflow docs
(https://www.tensorflow.org/api_docs/python/tf/sparse/sparse_dense_matmul), it
looks like Tensorflow is computing A*B where A is sparse. Our sparse_dense
computes B*A^T. You've done the transposition, but the order seems wrong.
##########
File path: python/tvm/topi/cuda/sparse.py
##########
@@ -367,9 +367,14 @@ def _alter_sparse_dense_layout(_attrs, inputs, _tinfos,
_out_type):
and isinstance(inputs[2], relay.Constant)
and isinstance(inputs[3], relay.Constant)
):
- sparse_matrix = sp.bsr_matrix(
- (inputs[1].data.asnumpy(), inputs[2].data.asnumpy(),
inputs[3].data.asnumpy())
- )
+ if len(inputs[1].data.asnumpy().shape) == 1:
Review comment:
Why this change? If you pass a 1D array to `bsr_matrix` it work just
fine.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
