Op info test for logspace .. masked.amax

[qihqi]

Fix the Op info test for logspace .. masked.amax

1. Find the lines 153 to 157 of test_ops.py and remove
   logspace .. masked.amax from skip_list
2. Run op_info test with pytest test/test_ops.py
3. Fix the failure.

Please refer to
this guide as guide to fix:

Also refer to these PRs:

• Add bitwise_left_shift, bitwise_right_shift #7343
• Add support for baddbmm, bincount #7312
• Add support for trig, binary, algebraic ops #7325

[ManfeiBai]

actual work will be "logspace", "lu", "lu_solve", "lu_unpack"

[ManfeiBai]

Hi, @barney-s, thanks for contribution! This issue would be a new first issue, please feel free to reply this issue, so that we could assign it to you

[barney-s]

Please assign it to me.

[barney-s]

Looking at logspace

logspace is decomposing to linspace. We will fix linspace #7505 and revisit this again.

Script:

import torch
import torch_xla2

env = torch_xla2.default_env()
env.config.debug_print_each_op = True
env.config.debug_accuracy_for_each_op = True

with env:
  y = torch.tensor([1, 5, 10])
  print(torch.logspace(start=-10, end=10, steps=5))

Output:

FUNCTION: tensor
/usr/local/google/home/barni/workspace/pytorch/xla/experimental/torch_xla2/torch_xla2/ops/jtorch.py:48: UserWarning: Explicitly requested dtype <class 'jax.numpy.int64'> requested in array is not available, and will be truncated to dtype int32. To enable more dtypes, set the jax_enable_x64 configuration option or the JAX_ENABLE_X64 shell environment variable. See https://github.com/google/jax#current-gotchas for more.
  return jnp.array(
FUNCTION: logspace
 DISPATCH: aten::logspace
  FUNCTION: linspace
   DISPATCH: aten::linspace
    FUNCTION: arange
     DISPATCH: aten::arange.start_step
    FUNCTION: lt
     DISPATCH: aten::lt.Scalar

[barney-s]

#8236

[ManfeiBai]

reopen due to we still have lu_unpack

[barney-s]

#8262 for lu_unpack