Merge pull request #4535 from google:lazy-simplification

PiperOrigin-RevId: 337183224

jax/experimental/jax2tf/jax2tf.py

@@ -25,7 +25,6 @@
 from jax import core
 from jax import custom_derivatives
 from jax import dtypes
-from jax import lax
 from jax import lax_linalg
 from jax import linear_util as lu
 from jax import numpy as jnp
@@ -38,8 +37,10 @@
 from jax.interpreters import partial_eval as pe
 from jax.interpreters import pxla
 from jax.interpreters import xla
+from jax.lax import lax
 from jax.lax import lax_control_flow
 from jax.lax import lax_fft
+from jax.lax import lax_parallel
 
 import numpy as np
 import tensorflow as tf  # type: ignore[import]
@@ -609,23 +610,24 @@ def _unexpected_primitive(p: core.Primitive, *args, **kwargs):
 tf_not_yet_impl = [
   lax.reduce_p, lax.rng_uniform_p,
 
-  lax.linear_solve_p,
+  lax_control_flow.linear_solve_p,
   lax_linalg.lu_p,
   lax_linalg.triangular_solve_p,
 
   lax.igamma_grad_a_p,
   lax.random_gamma_grad_p,
 
   # Not high priority?
-  lax.after_all_p, lax.all_to_all_p, lax.create_token_p, lax.cummax_p, lax.cummin_p,
-  lax.infeed_p, lax.outfeed_p, lax.pmax_p, lax.pmin_p, lax.ppermute_p, lax.psum_p,
-  lax.axis_index_p,
+  lax.after_all_p, lax_parallel.all_to_all_p, lax.create_token_p, lax.cummax_p,
+  lax.cummin_p, lax.infeed_p, lax.outfeed_p, lax_parallel.pmax_p,
+  lax_parallel.pmin_p, lax_parallel.ppermute_p, lax_parallel.psum_p,
+  lax_parallel.axis_index_p,
 
   pxla.xla_pmap_p,
 ]
 
 try:
-  tf_impl[lax.lax.tie_in_p] = lambda x, y: y
+  tf_impl[lax.tie_in_p] = lambda x, y: y
 except AttributeError:
   pass
 tf_impl[ad_util.stop_gradient_p] = tf.stop_gradient
@@ -688,6 +690,17 @@ def _population_count(x):
 tf_impl[lax.mul_p] = wrap_binary_op(tf.math.multiply)
 
 
+def _iota(*, dtype, shape, dimension):
+  size = shape[dimension]
+  # Some dtypes are unsupporetd, like uint32, so we just fall back to int32.
+  # TODO(mattjj, necula): improve tf.range dtype handling
+  vec = tf.range(tf.cast(size, tf.int32), dtype=tf.int32)
+  vec_shape = [-1 if i == dimension else 1 for i in range(len(shape))]
+  return tf.cast(tf.broadcast_to(tf.reshape(vec, vec_shape), shape), dtype)
+
+tf_impl[lax.iota_p] = _iota
+
+
 def _div(lhs, rhs):
   if lhs.dtype.is_integer:
     quotient = tf.math.floordiv(lhs, rhs)
@@ -1227,8 +1240,8 @@ def _select_and_gather_add(tangents: TfVal,
   const = lambda dtype, x: tf.constant(np.array(x), dtype)
 
   if double_word_reduction:
-    word_dtype = lax.lax._UINT_DTYPES[nbits]
-    double_word_dtype = lax.lax._UINT_DTYPES[nbits * 2]
+    word_dtype = lax._UINT_DTYPES[nbits]
+    double_word_dtype = lax._UINT_DTYPES[nbits * 2]
 
     # Packs two values into a tuple.
     def pack(a, b):
@@ -1536,7 +1549,7 @@ def _cond(index: TfVal, *operands: TfVal,
                  for jaxpr in branches]
   return tf.switch_case(index, branches_tf)
 
-tf_impl[lax.cond_p] = _cond
+tf_impl[lax_control_flow.cond_p] = _cond
 
 
 def _while(*args: TfVal, cond_nconsts: int, cond_jaxpr: core.ClosedJaxpr,
@@ -1603,10 +1616,10 @@ def select_one_carry(new_c: TfVal, c: TfVal) -> TfVal:
                                 (init_pred_b, *init_carry))
   return res_carry
 
-tf_impl[lax.while_p] = _while
+tf_impl[lax_control_flow.while_p] = _while
 
 # We use the scan impl rule to rewrite in terms of while.
-tf_impl_with_avals[lax.scan_p] = _convert_jax_impl(lax_control_flow._scan_impl)
+tf_impl_with_avals[lax_control_flow.scan_p] = _convert_jax_impl(lax_control_flow._scan_impl)
 
 def _top_k(operand: TfVal, k: int) -> Tuple[TfVal, TfVal]:
   # Some types originally incompatible with tf.math.top_k can be promoted

jax/experimental/jax2tf/tests/correctness_stats.py

@@ -91,7 +91,7 @@ def _to_np_dtype(dtype) -> NpDType:
       pass
     return np.dtype(dtype)
 
-  if args[0] is not core.unit:
+  if args and args[0] is not core.unit:
     np_dtype = _to_np_dtype(args[0].dtype)
   else:
     np_dtype = None

jax/lax/lax.py

@@ -1407,55 +1407,73 @@ def iota(dtype: DType, size: int) -> Array:
   <https://www.tensorflow.org/xla/operation_semantics#iota>`_
   operator.
   """
-  size = size if type(size) is masking.Poly else int(size)
-  shape = canonicalize_shape((size,))
-  dtype = dtypes.canonicalize_dtype(dtype)
-  lazy_expr = lazy.iota(dtype, shape[0])
-  aval = ShapedArray(shape, dtype)
-  return xla.DeviceArray(aval, None, lazy_expr, xla.DeviceConstant())
+  if config.omnistaging_enabled:
+    dtype = dtypes.canonicalize_dtype(dtype)
+    size = core.concrete_or_error(int, size, "size argument of lax.iota")
+    return iota_p.bind(dtype=dtype, shape=(size,), dimension=0)
+  else:
+    size = size if type(size) is masking.Poly else int(size)
+    shape = canonicalize_shape((size,))
+    dtype = dtypes.canonicalize_dtype(dtype)
+    lazy_expr = lazy.iota(dtype, shape[0])
+    aval = ShapedArray(shape, dtype)
+    return xla.DeviceArray(aval, None, lazy_expr, xla.DeviceConstant())
 
 def broadcasted_iota(dtype: DType, shape: Shape, dimension: int) -> Array:
   """Convenience wrapper around ``iota``."""
   dtype = dtypes.canonicalize_dtype(dtype)
   shape = canonicalize_shape(shape)
-  dimension = int(dimension)
-  return broadcast_in_dim(iota(dtype, shape[dimension]), shape, [dimension])
+  dimension = core.concrete_or_error(
+      int, dimension, "dimension argument of lax.broadcasted_iota")
+  return iota_p.bind(dtype=dtype, shape=shape, dimension=dimension)
 
 def _eye(dtype: DType, shape: Shape, offset: int) -> Array:
-  """Like numpy.eye, create a 2D array with ones on a diagonal.
-
-  This function exists for creating lazy identity matrices; that is,
-  materialization of the array is delayed and it may be fused into consumers to
-  avoid materialization at all."""
+  """Like numpy.eye, create a 2D array with ones on a diagonal."""
   N, M = tuple(map(int, shape))
   offset = int(offset)
   dtype = dtypes.canonicalize_dtype(dtype)
-  lazy_expr = lazy.eye(dtype, (N, M), offset)
-  aval = ShapedArray((N, M), dtype)
-  return xla.DeviceArray(aval, None, lazy_expr, xla.DeviceConstant())
+  if config.omnistaging_enabled:
+    bool_eye = eq(add(broadcasted_iota(np.int32, (N, M), 0), np.int32(offset)),
+                  broadcasted_iota(np.int32, (N, M), 1))
+    return convert_element_type_p.bind(bool_eye, new_dtype=dtype,
+                                       old_dtype=np.bool_)
+  else:
+    lazy_expr = lazy.eye(dtype, (N, M), offset)
+    aval = ShapedArray((N, M), dtype)
+    return xla.DeviceArray(aval, None, lazy_expr, xla.DeviceConstant())
 
 def _delta(dtype: DType, shape: Shape, axes: Sequence[int]) -> Array:
-  """This function exists for creating lazy Kronecker delta arrays, particularly
-  for use in jax.numpy.einsum to express traces. It differs from ``eye`` in that
-  it can create arrays of any rank, but doesn't allow offsets."""
+  """This utility function exists for creating Kronecker delta arrays."""
   shape = tuple(map(int, shape))
   axes = tuple(map(int, axes))
   dtype = dtypes.canonicalize_dtype(dtype)
   base_shape = tuple(np.take(shape, axes))
-  lazy_expr = lazy.broadcast(lazy.delta(dtype, base_shape), shape, axes)
-  aval = ShapedArray(shape, dtype)
-  return xla.DeviceArray(aval, None, lazy_expr, xla.DeviceConstant())
+  if config.omnistaging_enabled:
+    iotas = [broadcasted_iota(np.uint32, base_shape, i)
+             for i in range(len(base_shape))]
+    eyes = [eq(i1, i2) for i1, i2 in zip(iotas[:-1], iotas[1:])]
+    result = convert_element_type_p.bind(_reduce(operator.and_, eyes),
+                                         new_dtype=dtype, old_dtype=np.bool_)
+    return broadcast_in_dim(result, shape, axes)
+  else:
+    lazy_expr = lazy.broadcast(lazy.delta(dtype, base_shape), shape, axes)
+    aval = ShapedArray(shape, dtype)
+    return xla.DeviceArray(aval, None, lazy_expr, xla.DeviceConstant())
 
 def _tri(dtype: DType, shape: Shape, offset: int) -> Array:
-  """Like numpy.tri, create a 2D array with ones below a diagonal.
-  This function exists for creating lazy triangular matrices, particularly for
-  use in jax.numpy.tri."""
+  """Like numpy.tri, create a 2D array with ones below a diagonal."""
   N, M = tuple(map(int, shape))
   offset = int(offset)
   dtype = dtypes.canonicalize_dtype(dtype)
-  lazy_expr = lazy.tri(dtype, (N, M), offset)
-  aval = ShapedArray((N, M), dtype)
-  return xla.DeviceArray(aval, None, lazy_expr, xla.DeviceConstant())
+  if config.omnistaging_enabled:
+    bool_tri = ge(add(broadcasted_iota(np.int32, (N, M), 0), np.int32(offset)),
+                  broadcasted_iota(np.int32, (N, M), 1))
+    return convert_element_type_p.bind(bool_tri, old_dtype=np.int32,
+                                       new_dtype=dtype)
+  else:
+    lazy_expr = lazy.tri(dtype, (N, M), offset)
+    aval = ShapedArray((N, M), dtype)
+    return xla.DeviceArray(aval, None, lazy_expr, xla.DeviceConstant())
 
 def stop_gradient(x):
   """Stops gradient computation.
@@ -5779,6 +5797,7 @@ def _outfeed_translation_rule(c, token, *xs):
 outfeed_p.def_abstract_eval(_outfeed_abstract_eval)
 xla.translations[outfeed_p] = _outfeed_translation_rule
 
+
 def rng_uniform(a, b, shape):
   """Stateful PRNG generator. Experimental and its use is discouraged.
 
@@ -5811,6 +5830,30 @@ def _rng_uniform_translation_rule(c, a, b, *, shape):
 rng_uniform_p.def_abstract_eval(_rng_uniform_abstract_eval)
 xla.translations[rng_uniform_p] = _rng_uniform_translation_rule
 
+
+def _iota_abstract_eval(*, dtype, shape, dimension):
+  _check_shapelike("iota", "shape", shape)
+  if not any(dtypes.issubdtype(dtype, t) for t in _num):
+    msg = 'iota does not accept dtype {}. Accepted dtypes are subtypes of {}.'
+    typename = str(np.dtype(dtype).name)
+    accepted_typenames = (t.__name__ for t in _num)
+    raise TypeError(msg.format(typename, ', '.join(accepted_typenames)))
+  if not 0 <= dimension < len(shape):
+    raise ValueError("iota dimension must be between 0 and len(shape), got "
+                     f"dimension={dimension} for shape {shape}")
+  return ShapedArray(shape, dtype)
+
+def _iota_translation_rule(c, dtype, shape, dimension):
+  etype = xla_client.dtype_to_etype(dtype)
+  xla_shape = xc.Shape.array_shape(etype, shape)
+  return xops.Iota(c, xla_shape, dimension)
+
+iota_p = Primitive('iota')
+iota_p.def_impl(partial(xla.apply_primitive, iota_p))
+iota_p.def_abstract_eval(_iota_abstract_eval)
+xla.translations[iota_p] = _iota_translation_rule
+
+
 ### util
 
 _ndim = np.ndim

jax/random.py

@@ -304,6 +304,7 @@ def _fold_in(key, data):
   return threefry_2x32(key, PRNGKey(data))
 
 
+@partial(jit, static_argnums=(1, 2))
 def _random_bits(key, bit_width, shape):
   """Sample uniform random bits of given width and shape using PRNG key."""
   if not _is_prng_key(key):

tests/api_test.py

@@ -33,7 +33,7 @@
 import jax
 import jax.numpy as jnp
 from jax import float0, jit, grad, device_put, jacfwd, jacrev, hessian
-from jax import api, core, lax, lax_reference
+from jax import api, core, lax, lax_reference, lazy
 from jax.core import Primitive
 from jax.interpreters import ad
 from jax.interpreters import xla
@@ -1477,7 +1477,7 @@ def f(x):
   def test_dtype_warning(self):
     # cf. issue #1230
     if FLAGS.jax_enable_x64:
-      return  # test only applies when x64 is disabled
+      raise unittest.SkipTest("test only applies when x64 is disabled")
 
     def check_warning(warn, nowarn):
       with warnings.catch_warnings(record=True) as w:
@@ -2470,14 +2470,14 @@ def f(x):
       assert python_should_be_executing
       return jnp.sum(x)
 
-    x = jnp.arange(10, dtype=jnp.int32)
-    assert xla.is_device_constant(x)  # lazy iota
+    x = jnp.zeros(10, dtype=jnp.int32)
+    assert not lazy.is_trivial(x._lazy_expr)
 
     python_should_be_executing = True
     _ = f(x)
 
     python_should_be_executing = False  # should not recompile
-    x = np.arange(10, dtype=np.int32)
+    x = np.zeros(10, dtype=np.int32)
     _ = f(x)
 
   @parameterized.parameters(jtu.cases_from_list(range(10000)))

tests/lax_numpy_test.py

@@ -3651,8 +3651,9 @@ def testArange(self):
                       type(lax.iota(np.int32, 77)))
 
     # test laziness for int dtypes
-    self.assertTrue(xla.is_device_constant(jnp.arange(77)))
-    self.assertTrue(xla.is_device_constant(jnp.arange(77, dtype=jnp.int32)))
+    if not config.omnistaging_enabled:
+      self.assertTrue(xla.is_device_constant(jnp.arange(77)))
+      self.assertTrue(xla.is_device_constant(jnp.arange(77, dtype=jnp.int32)))
 
   def testArangeJit(self):
     ans = api.jit(lambda: jnp.arange(5))()