ashutosh-arm commented on code in PR #12042:
URL: https://github.com/apache/tvm/pull/12042#discussion_r1018896087
##########
src/relay/qnn/op/dense.cc:
##########
@@ -47,12 +47,14 @@ bool QnnDenseRel(const Array<Type>& types, int num_inputs,
const Attrs& attrs,
if (data == nullptr || weight == nullptr) return false;
const auto* param = attrs.as<DenseAttrs>();
ICHECK(param != nullptr) << "DenseAttrs cannot be nullptr.";
- ICHECK(data->dtype == DataType::Int(8) || data->dtype == DataType::UInt(8))
- << "Expected quantized dense type(int8, uint8) for input but was " <<
data->dtype;
+ ICHECK(data->dtype == DataType::Int(8) || data->dtype == DataType::UInt(8) ||
+ data->dtype == DataType::Int(16) || data->dtype == DataType::UInt(16))
+ << "Expected quantized dense type(int8, uint8, int16, uint16) for input
but was "
+ << data->dtype;
ICHECK(weight->dtype == DataType::Int(8) || weight->dtype ==
DataType::UInt(8))
<< "Expected quantized dense type(int8, uint8) for weight but was " <<
weight->dtype;
- ICHECK(param->out_dtype == DataType::Int(32))
- << "Expected quantized dense type(int32) for output but was " <<
param->out_dtype;
+ ICHECK(param->out_dtype == DataType::Int(32) || param->out_dtype ==
DataType::Int(64))
+ << "Expected quantized dense type(int32, int64) for output but was " <<
param->out_dtype;
Review Comment:
Do we need additional dtype checks for bias as well based on output dtype?
If yes, then probably we should check the QNN Conv2D type checker too.
##########
tests/python/contrib/test_ethosn/test_convert_equivalents.py:
##########
@@ -227,7 +227,7 @@ def expected():
@requires_ethosn
@pytest.mark.parametrize(
"dtype,shape,constant_shape",
- [("int16", (1, 16, 12, 4), None)],
+ [("float32", (1, 16, 12, 4), None)],
Review Comment:
Just curious: why the dtype change?
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
