Dataset Viewer
Auto-converted to Parquet Duplicate
task_id
stringlengths
9
9
category
stringclasses
5 values
prompt
stringlengths
20
194
input_data
stringclasses
6 values
expected_output
stringlengths
51
7.9k
torch-000
tensor_ops
Reshape `x` to shape (4, 6) in row-major order.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [[-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197], [-1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012], [0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.913043498...
torch-001
tensor_ops
Return the transpose of `m` as a contiguous tensor.
{"m": "{\"data\": [[-1.0, -0.9428571462631226, -0.8857142925262451, -0.8285714387893677, -0.7714285850524902, -0.7142857313156128], [-0.6571428775787354, -0.6000000238418579, -0.5428571701049805, -0.48571425676345825, -0.4285714030265808, -0.37142854928970337], [-0.3142856955528259, -0.2571428418159485, -0.199999988079...
{"data": [[-1.0, -0.6571428775787354, -0.3142856955528259, 0.02857142686843872, 0.37142854928970337, 0.7142857313156128], [-0.9428571462631226, -0.6000000238418579, -0.2571428418159485, 0.08571428060531616, 0.4285714030265808, 0.7714285850524902], [-0.8857142925262451, -0.5428571701049805, -0.19999998807907104, 0.14285...
torch-002
tensor_ops
Return the row-wise sum of `b` as a 1-D tensor.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [1.0, 3.5, 6.0, 8.5, 11.0, 13.5, 16.0, 18.5], "dtype": "float32", "shape": [8]}
torch-003
tensor_ops
Return the column-wise mean of `b`.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [1.75, 1.8500001430511475, 1.9499999284744263, 2.0500001907348633, 2.1500000953674316], "dtype": "float32", "shape": [5]}
torch-004
tensor_ops
Clamp `x` to the range [-1, 1].
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.9130434989929199, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0], "dtype": "float32", "shape": [24]}
torch-005
tensor_ops
Return the elements of `x` that are strictly positive, in order.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.9130434989929199, 1.1739130020141602, 1.43478262424469, 1.6956522464752197, 1.95652174949646, 2.2173912525177, 2.4782609939575195, 2.7391304969787598, 3.0], "dtype": "float32", "shape": [12]}
torch-006
tensor_ops
Return `x` sorted in descending order (values only).
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [3.0, 2.7391304969787598, 2.4782609939575195, 2.2173912525177, 1.95652174949646, 1.6956522464752197, 1.43478262424469, 1.1739130020141602, 0.9130434989929199, 0.6521739959716797, 0.39130449295043945, 0.13043475151062012, -0.13043475151062012, -0.39130449295043945, -0.6521739959716797, -0.9130434989929199, -1.1...
torch-007
tensor_ops
Return the indices that sort `x` ascending, as int64.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23], "dtype": "int64", "shape": [24]}
torch-008
tensor_ops
Concatenate `b` with itself along dim 0.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.7000000476837158, 1.800...
torch-009
tensor_ops
Stack `b` with itself along a NEW leading dimension.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [[[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.7000000476837158, 1.80...
torch-010
tensor_ops
Return the matrix product of `m` with itself.
{"m": "{\"data\": [[-1.0, -0.9428571462631226, -0.8857142925262451, -0.8285714387893677, -0.7714285850524902, -0.7142857313156128], [-0.6571428775787354, -0.6000000238418579, -0.5428571701049805, -0.48571425676345825, -0.4285714030265808, -0.37142854928970337], [-0.3142856955528259, -0.2571428418159485, -0.199999988079...
{"data": [[1.077551007270813, 0.7836735844612122, 0.489795982837677, 0.19591835141181946, -0.09795932471752167, -0.3918367624282837], [0.7836736440658569, 0.6073469519615173, 0.43102046847343445, 0.2546938359737396, 0.07836739718914032, -0.09795913100242615], [0.4897959232330322, 0.4310203790664673, 0.3722448945045471,...
torch-011
tensor_ops
Normalise each row of `b` to unit L2 norm.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [[0.0, 0.18257418274879456, 0.3651483654975891, 0.547722578048706, 0.7302967309951782], [0.31311213970184326, 0.3757345974445343, 0.43835699558258057, 0.5009794235229492, 0.5636018514633179], [0.37011659145355225, 0.407128244638443, 0.44413992762565613, 0.4811515808105469, 0.5181632041931152], [0.3932418823242...
torch-012
tensor_ops
Add a leading batch dimension to `x` so its shape becomes (1, 24).
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [[-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.9130434989929...
torch-013
tensor_ops
Return the cumulative sum of `x` along dim 0.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-3.0, -5.73913049697876, -8.217391967773438, -10.434782981872559, -12.391304016113281, -14.086956977844238, -15.52173900604248, -16.69565200805664, -17.60869598388672, -18.2608699798584, -18.65217399597168, -18.782609939575195, -18.65217399597168, -18.2608699798584, -17.60869598388672, -16.69565200805664, -15...
torch-014
tensor_ops
Cast `x` to float64.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304349899291...
torch-015
tensor_ops
Return the pairwise maximum of `x` and its own reverse.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [3.0, 2.7391304969787598, 2.4782609939575195, 2.2173912525177, 1.95652174949646, 1.6956522464752197, 1.43478262424469, 1.1739130020141602, 0.9130434989929199, 0.6521739959716797, 0.39130449295043945, 0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.9130434989929199, 1.173913...
torch-016
autograd
Treat `x` as a leaf requiring grad, compute y = (x**2).sum(), backward, and return x.grad.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-6.0, -5.4782609939575195, -4.956521987915039, -4.4347825050354, -3.91304349899292, -3.3913044929504395, -2.86956524848938, -2.3478260040283203, -1.8260869979858398, -1.3043479919433594, -0.7826089859008789, -0.26086950302124023, 0.26086950302124023, 0.7826089859008789, 1.3043479919433594, 1.8260869979858398,...
torch-017
autograd
With `x` requiring grad, compute y = (x.sin() * x).sum(), backward, and return x.grad.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [2.828857421875, 2.128587245941162, 1.3369883298873901, 0.5377766489982605, -0.19042092561721802, -0.7810530662536621, -1.18531334400177, -1.3760414123535156, -1.349548101425171, -1.1252415180206299, -0.743121325969696, -0.2593919634819031, 0.2593919634819031, 0.743121325969696, 1.1252415180206299, 1.349548101...
torch-018
autograd
With `b` requiring grad, compute the mean of b.exp() clamped to a max of 10, backward, and return b.grad.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [[0.02500000037252903, 0.027629274874925613, 0.030535070225596428, 0.033746469765901566, 0.0372956208884716], [0.04121803119778633, 0.045552972704172134, 0.050343818962574005, 0.055638521909713745, 0.06149008497595787], [0.06795704364776611, 0.07510415464639664, 0.08300292491912842, 0.09173242747783661, 0.1013...
torch-019
autograd
Return the gradient of (x**3).sum() with respect to x, using torch.autograd.grad rather than backward.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [27.0, 22.508506774902344, 18.42533302307129, 14.750471115112305, 11.483932495117188, 8.625709533691406, 6.175803184509277, 4.134215354919434, 2.5009453296661377, 1.275992751121521, 0.4593576192855835, 0.0510396733880043, 0.0510396733880043, 0.4593576192855835, 1.275992751121521, 2.5009453296661377, 4.13421535...
torch-020
autograd
With `x` requiring grad, compute y = (x**2).sum() but DETACH x before squaring, then return a tensor of zeros shaped like x if no gradient flows (i.e. return the grad, or zeros when it is None).
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "dtype": "float32", "shape": [24]}
torch-021
autograd
Compute the second derivative of (x**3).sum() with respect to x (create_graph=True), returning d2y/dx2.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-18.0, -16.434783935546875, -14.869565963745117, -13.30434799194336, -11.739130020141602, -10.173913955688477, -8.608695983886719, -7.043478012084961, -5.4782609939575195, -3.913043975830078, -2.3478269577026367, -0.7826085090637207, 0.7826085090637207, 2.3478269577026367, 3.913043975830078, 5.478260993957519...
torch-022
nn_modules
Apply a Linear(5, 3) layer to `b` whose weight is arange(15).reshape(3,5)/10 (float32) and whose bias is tensor([0.1, 0.2, 0.3]). Return the output.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [[0.4000000059604645, 0.9999999403953552, 1.5999999046325684], [0.9000000357627869, 2.75, 4.600000381469727], [1.4000000953674316, 4.5, 7.600000381469727], [1.899999976158142, 6.25, 10.600000381469727], [2.4000000953674316, 8.0, 13.600001335144043], [2.9000000953674316, 9.75, 16.600000381469727], [3.4000000953...
torch-023
nn_modules
Apply torch.nn.functional.relu to `x`.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.9130434989929199, 1.1739130020141602, 1.43478262424469, 1.6956522464752197, 1.95652174949646, 2.2173912525177, 2.4782609939575195, 2.7391304969787598, 3.0], "dtype": "float32", "shape": ...
torch-024
nn_modules
Apply torch.nn.functional.gelu to `x` with the default approximation.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-0.004049867391586304, -0.0084369583055377, -0.016359606757760048, -0.029487209394574165, -0.04930786415934563, -0.07626339793205261, -0.10857655107975006, -0.1411219835281372, -0.1649046540260315, -0.16770288348197937, -0.13609030842781067, -0.05844927206635475, 0.07198548316955566, 0.2552141845226288, 0.484...
torch-025
nn_modules
Apply softmax along dim 1 to `l`.
{"l": "{\"data\": [[-2.0, -1.6363636255264282, -1.2727272510528564, -0.9090908765792847], [-0.5454545021057129, -0.1818181276321411, 0.1818181276321411, 0.5454545021057129], [0.9090908765792847, 1.2727272510528564, 1.6363636255264282, 2.0]], \"dtype\": \"float32\", \"shape\": [3, 4]}"}
{"data": [[0.13360126316547394, 0.19219224154949188, 0.2764783501625061, 0.39772820472717285], [0.13360127806663513, 0.19219225645065308, 0.2764783203601837, 0.39772817492485046], [0.13360126316547394, 0.19219224154949188, 0.2764783501625061, 0.39772820472717285]], "dtype": "float32", "shape": [3, 4]}
torch-026
nn_modules
Apply log_softmax along dim 1 to `l`.
{"l": "{\"data\": [[-2.0, -1.6363636255264282, -1.2727272510528564, -0.9090908765792847], [-0.5454545021057129, -0.1818181276321411, 0.1818181276321411, 0.5454545021057129], [0.9090908765792847, 1.2727272510528564, 1.6363636255264282, 2.0]], \"dtype\": \"float32\", \"shape\": [3, 4]}"}
{"data": [[-2.0128955841064453, -1.649259090423584, -1.2856228351593018, -0.9219864010810852], [-2.0128955841064453, -1.649259090423584, -1.2856228351593018, -0.9219865202903748], [-2.0128955841064453, -1.649259090423584, -1.2856228351593018, -0.9219864010810852]], "dtype": "float32", "shape": [3, 4]}
torch-027
nn_modules
Apply LayerNorm over the last dimension of `b` with eps=1e-5, no learnable affine parameters.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [[-1.4138602018356323, -0.7069301009178162, 0.0, 0.7069301009178162, 1.4138602018356323], [-1.4138604402542114, -0.7069302201271057, -4.213631541460927e-07, 0.7069298028945923, 1.4138600826263428], [-1.4138604402542114, -0.7069302201271057, 0.0, 0.7069302201271057, 1.4138596057891846], [-1.4138604402542114, -0...
torch-028
nn_modules
Apply 2x2 max pooling with stride 2 to the NCHW batch `i`.
{"i": "{\"data\": [[[[-2.0, -1.9895561933517456, -1.9791122674942017, -1.9686684608459473, -1.9582245349884033, -1.947780728340149, -1.937336802482605, -1.9268929958343506], [-1.9164490699768066, -1.9060052633285522, -1.8955613374710083, -1.885117530822754, -1.87467360496521, -1.8642297983169556, -1.8537858724594116, -...
{"data": [[[[-1.9060052633285522, -1.885117530822754, -1.8642297983169556, -1.8433420658111572], [-1.7389034032821655, -1.7180156707763672, -1.6971279382705688, -1.6762402057647705], [-1.5718015432357788, -1.5509138107299805, -1.5300260782241821, -1.5091383457183838], [-1.4046998023986816, -1.3838119506835938, -1.36292...
torch-029
nn_modules
Apply 2x2 average pooling with stride 2 to the NCHW batch `i`.
{"i": "{\"data\": [[[[-2.0, -1.9895561933517456, -1.9791122674942017, -1.9686684608459473, -1.9582245349884033, -1.947780728340149, -1.937336802482605, -1.9268929958343506], [-1.9164490699768066, -1.9060052633285522, -1.8955613374710083, -1.885117530822754, -1.87467360496521, -1.8642297983169556, -1.8537858724594116, -...
{"data": [[[[-1.953002691268921, -1.932114839553833, -1.9112271070480347, -1.8903393745422363], [-1.7859008312225342, -1.7650129795074463, -1.744125247001648, -1.7232375144958496], [-1.6187989711761475, -1.5979112386703491, -1.5770235061645508, -1.556135654449463], [-1.4516971111297607, -1.4308093786239624, -1.40992164...
torch-030
nn_modules
Apply a Conv2d(3, 2, kernel_size=3) to `i` with weight filled with 0.01 and bias filled with zero, no padding. Return the output.
{"i": "{\"data\": [[[[-2.0, -1.9895561933517456, -1.9791122674942017, -1.9686684608459473, -1.9582245349884033, -1.947780728340149, -1.937336802482605, -1.9268929958343506], [-1.9164490699768066, -1.9060052633285522, -1.8955613374710083, -1.885117530822754, -1.87467360496521, -1.8642297983169556, -1.8537858724594116, -...
{"data": [[[[-0.3341514468193054, -0.3313315808773041, -0.32851171493530273, -0.3256918787956238, -0.3228720426559448, -0.32005223631858826], [-0.311592698097229, -0.3087728023529053, -0.3059529960155487, -0.30313315987586975, -0.3003133237361908, -0.29749348759651184], [-0.2890339195728302, -0.28621408343315125, -0.28...
torch-031
nn_modules
Flatten `i` from dim 1 onward.
{"i": "{\"data\": [[[[-2.0, -1.9895561933517456, -1.9791122674942017, -1.9686684608459473, -1.9582245349884033, -1.947780728340149, -1.937336802482605, -1.9268929958343506], [-1.9164490699768066, -1.9060052633285522, -1.8955613374710083, -1.885117530822754, -1.87467360496521, -1.8642297983169556, -1.8537858724594116, -...
{"data": [[-2.0, -1.9895561933517456, -1.9791122674942017, -1.9686684608459473, -1.9582245349884033, -1.947780728340149, -1.937336802482605, -1.9268929958343506, -1.9164490699768066, -1.9060052633285522, -1.8955613374710083, -1.885117530822754, -1.87467360496521, -1.8642297983169556, -1.8537858724594116, -1.84334206581...
torch-032
nn_modules
Apply an Embedding(5, 4) to the index tensor `s`, where the embedding weight is arange(20).reshape(5,4) as float32. Return the output.
{"s": "{\"data\": [[0, 1, 2, 3, 4, 0], [1, 2, 3, 4, 0, 1], [2, 3, 4, 0, 1, 2], [3, 4, 0, 1, 2, 3]], \"dtype\": \"int64\", \"shape\": [4, 6]}"}
{"data": [[[0.0, 1.0, 2.0, 3.0], [4.0, 5.0, 6.0, 7.0], [8.0, 9.0, 10.0, 11.0], [12.0, 13.0, 14.0, 15.0], [16.0, 17.0, 18.0, 19.0], [0.0, 1.0, 2.0, 3.0]], [[4.0, 5.0, 6.0, 7.0], [8.0, 9.0, 10.0, 11.0], [12.0, 13.0, 14.0, 15.0], [16.0, 17.0, 18.0, 19.0], [0.0, 1.0, 2.0, 3.0], [4.0, 5.0, 6.0, 7.0]], [[8.0, 9.0, 10.0, 11.0...
torch-033
nn_modules
Apply sigmoid to `x`.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [0.04742587357759476, 0.06070346012711525, 0.07739628106355667, 0.09819959104061127, 0.1238439679145813, 0.15503396093845367, 0.19235458970069885, 0.2361484318971634, 0.2863774299621582, 0.3424997925758362, 0.40340331196784973, 0.4674374461174011, 0.5325625538825989, 0.5965967178344727, 0.657500147819519, 0.71...
torch-034
losses
Return the mean-squared-error between `b` and a tensor of zeros of the same shape, as a 1-element tensor.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [5.135000705718994], "dtype": "float32", "shape": [1]}
torch-035
losses
Return the cross-entropy loss of logits `l` against class targets [0, 3, 1] with mean reduction, as a 1-element tensor.
{"l": "{\"data\": [[-2.0, -1.6363636255264282, -1.2727272510528564, -0.9090908765792847], [-0.5454545021057129, -0.1818181276321411, 0.1818181276321411, 0.5454545021057129], [0.9090908765792847, 1.2727272510528564, 1.6363636255264282, 2.0]], \"dtype\": \"float32\", \"shape\": [3, 4]}"}
{"data": [1.5280470848083496], "dtype": "float32", "shape": [1]}
torch-036
losses
Return the PER-SAMPLE cross-entropy of logits `l` against targets [0, 3, 1] (reduction='none').
{"l": "{\"data\": [[-2.0, -1.6363636255264282, -1.2727272510528564, -0.9090908765792847], [-0.5454545021057129, -0.1818181276321411, 0.1818181276321411, 0.5454545021057129], [0.9090908765792847, 1.2727272510528564, 1.6363636255264282, 2.0]], \"dtype\": \"float32\", \"shape\": [3, 4]}"}
{"data": [2.0128955841064453, 0.9219865202903748, 1.649259090423584], "dtype": "float32", "shape": [3]}
torch-037
losses
Return the binary cross-entropy between sigmoid(`x`) and a target of all 0.5, mean reduction, as a 1-element tensor.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [1.0311164855957031], "dtype": "float32", "shape": [1]}
torch-038
losses
Return the L1 loss between `b` and b.roll(1, 0), mean reduction, as a 1-element tensor.
{"b": "{\"data\": [[0.0, 0.10000000149011612, 0.20000000298023224, 0.30000001192092896, 0.4000000059604645], [0.5, 0.6000000238418579, 0.699999988079071, 0.800000011920929, 0.9000000357627869], [1.0, 1.100000023841858, 1.2000000476837158, 1.3000000715255737, 1.399999976158142], [1.5, 1.600000023841858, 1.70000004768371...
{"data": [0.875], "dtype": "float32", "shape": [1]}
torch-039
losses
Return the smooth L1 loss (beta=1.0) between `x` and x.flip(0), mean reduction, as a 1-element tensor.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [2.6551668643951416], "dtype": "float32", "shape": [1]}
torch-040
optim
Take `x` as a parameter, run ONE SGD step with lr=0.1 on the loss (x**2).sum(), and return the updated parameter values.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-2.4000000953674316, -2.1913044452667236, -1.9826087951660156, -1.773913025856018, -1.56521737575531, -1.3565218448638916, -1.147826075553894, -0.939130425453186, -0.730434775352478, -0.5217391848564148, -0.31304359436035156, -0.10434780269861221, 0.10434780269861221, 0.31304359436035156, 0.5217391848564148, ...
torch-041
optim
Take `x` as a parameter, run ONE SGD step with lr=0.1 and momentum=0.9 on the loss (x**2).sum(), and return the updated parameter values.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-2.4000000953674316, -2.1913044452667236, -1.9826087951660156, -1.773913025856018, -1.56521737575531, -1.3565218448638916, -1.147826075553894, -0.939130425453186, -0.730434775352478, -0.5217391848564148, -0.31304359436035156, -0.10434780269861221, 0.10434780269861221, 0.31304359436035156, 0.5217391848564148, ...
torch-042
optim
Take `x` as a parameter, run ONE Adam step with lr=0.1 (default betas/eps) on the loss (x**2).sum(), and return the updated parameter values.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-2.9000000953674316, -2.6391305923461914, -2.378261089324951, -2.117391347885132, -1.856521725654602, -1.5956522226333618, -1.334782600402832, -1.0739129781723022, -0.813043475151062, -0.5521739721298218, -0.29130449891090393, -0.030434750020503998, 0.030434750020503998, 0.29130449891090393, 0.552173972129821...
torch-043
optim
Take `x` as a parameter, run THREE SGD steps with lr=0.05 on the loss (x**2).sum(), and return the final parameter values.
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-2.187000036239624, -1.9968260526657104, -1.806652307510376, -1.6164782047271729, -1.4263043403625488, -1.2361304759979248, -1.0459564924240112, -0.8557825684547424, -0.6656086444854736, -0.4754348397254944, -0.28526097536087036, -0.09508693218231201, 0.09508693218231201, 0.28526097536087036, 0.47543483972549...
torch-044
optim
Take `x` as a parameter and return the gradient after clipping the global grad-norm to 1.0, for the loss (x**2).sum().
{"x": "{\"data\": [-3.0, -2.7391304969787598, -2.4782609939575195, -2.2173912525177, -1.95652174949646, -1.6956522464752197, -1.43478262424469, -1.1739130020141602, -0.9130434989929199, -0.6521739959716797, -0.39130449295043945, -0.13043475151062012, 0.13043475151062012, 0.39130449295043945, 0.6521739959716797, 0.91304...
{"data": [-0.3391164541244507, -0.30962806940078735, -0.280139684677124, -0.2506512701511383, -0.22116290032863617, -0.19167451560497284, -0.1621861308813095, -0.13269773125648499, -0.10320935398340225, -0.07372097671031952, -0.04423259571194649, -0.014744189567863941, 0.014744189567863941, 0.04423259571194649, 0.07372...

pytorch-tasks-v1

Task dataset for a PyTorch RL / eval environment, in the shape used by the Prime Intellect Environments Hub.

45 PyTorch tasks across 5 categories. Each task gives the model one or more input tensors and an instruction; the answer is the tensor left in result, graded with torch.allclose against a reference. Grading is deterministic — no LLM judge, no external API, CPU only.

Category Tasks Covers
tensor_ops 16 reshape, transpose, axis reductions, clamp, boolean masking, sort/argsort, cat/stack, matmul, L2 row-normalise, unsqueeze, cumsum, dtype cast, elementwise max
nn_modules 12 Linear and Conv2d with explicit weights, relu/gelu/sigmoid, softmax and log_softmax, LayerNorm, max/avg pooling, flatten, Embedding
autograd 6 backward on scalar losses, torch.autograd.grad, gradient through sin, clamp-ed exp, detach semantics, second derivatives with create_graph
losses 6 MSE, cross-entropy (mean and per-sample), binary cross-entropy, L1, smooth L1
optim 5 one SGD step, SGD with momentum, one Adam step, three chained SGD steps, clip_grad_norm_

Fields

Field Description
task_id stable id, e.g. torch-017
category one of the five above
prompt the natural-language instruction shown to the model
input_data JSON object mapping tensor name (x, m, b, i, s, l) to a serialised tensor
expected_output the serialised reference result

Tensors serialise as {"data": <nested list>, "dtype": <torch dtype without the prefix>, "shape": [...]}.

How it was built, and why you can trust the answer key

Tasks are defined as (deterministic input tensors, instruction, reference solution). The expected output is computed by executing the reference, never written by hand, so the key cannot drift from the instruction.

Every task is then independently verified:

  • the reference runs and returns a real-valued tensor
  • it is deterministic (executed twice, compared exactly)
  • the result is finite — no NaN or inf in the answer key
  • the result is non-empty and not identical to its input
  • both the result and every input tensor survive the serialisation round-trip exactly, dtype and shape included

All 45 pass on torch 2.13.0+cpu. Builder and verifier: build_tasks.py.

Why the nn layers use explicit weights

Linear, Conv2d and Embedding here are filled with arange/linspace/constant weights rather than manual_seed plus default initialisation. Default initialisers and RNG stream details are implementation choices that have changed between PyTorch releases; arange cannot. A dataset whose labels silently change under pip install -U torch would be worse than no dataset.

A note on grading tolerance

Grading uses torch.allclose (rtol 1e-5, atol 1e-7) rather than exact equality: matmul, convolution and pooling dispatch to BLAS/oneDNN kernels whose last bits legitimately differ across builds and architectures. Shape is compared exactly, and an integer-valued reference requires an integer answer, so the tolerance never launders a wrong-shaped or wrong-typed result.

Downloads last month
-