Priya Goyal
Priya Goyal
we need the cuda stream information for tensors to be used in c2isl. The only way to do this right now is via TH state.
trying to add bitwise operators. so far they are parsed correctly in their precedence order. `tc2halide` conversion is also done. I am not sure about how to propagate these through...
There is some interest in being able to do nested function calls i.e. define one TC and call it from inside another one. ``` lang = """ def max() {...
PR https://github.com/facebookresearch/TensorComprehensions/pull/183 disables SpatialBatchNorm test. It has various issues: 1. dimension mismatching for weight 2. output doesn't match in the specific reference check this is related to changes in ATen...
In the next release, we want to improve the pybindings besides other things: - [ ] PyTorch now provides the cpp extensions that we can use to extract aten tensor...
the reason for having the LHS indirection is because let's say someone has the lookup table: ``` def lut(float(B, R) M, int32(B, N) I) -> (O) { O(b, n) +=!...
We can't use scalar inputs in the bounds inference right now. So for example: ``` LANG=""" def avgpool(float(B, C, H, W) input, float kH, float kW, float sH, float sW)...
is there a way for us to support variable tensor sizes? right now, if the tensor size changes, we have to recompile and cache. But often in computer vision, NLP,...
one of our users reported on slack channel that they were trying to translate the following Numpy code ``` def A_matvec_batch(A, X): n, m = X.shape Y = np.zeros((n, m))...
There is a big interest in having support for RNN loops in TC. Creating this master task to discuss further and track progress