jax_export_test.py

# Copyright 2023 The JAX Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import contextlib
import logging
import math
from typing import List
import unittest

from absl.testing import absltest, parameterized
import jax
from jax import numpy as jnp
from jax import tree_util
from jax.config import config
from jax.experimental.jax2tf import jax_export
try:
  from jax.experimental.jax2tf import jax2tf  # TODO: temporary
except ImportError:
  jax2tf = None  # type: ignore

from jax._src import core
from jax._src import test_util as jtu
from jax._src import xla_bridge as xb

import numpy as np


config.parse_flags_with_absl()


class JaxExportTest(jtu.JaxTestCase):

  def test_basic_export_only(self):
    def my_fun(x):
      return jnp.sin(x)
    exp = jax_export.export(my_fun)(jax.ShapeDtypeStruct((4,), dtype=np.float32))
    self.assertEqual("my_fun", exp.fun_name)
    self.assertEqual(jax_export.default_lowering_platform(), exp.lowering_platform)
    self.assertEqual(tree_util.tree_flatten(((1,), {}))[1], exp.in_tree)
    self.assertEqual((core.ShapedArray((4,), dtype=np.float32),), exp.in_avals)
    self.assertEqual((core.ShapedArray((4,), dtype=np.float32),), exp.out_avals)

  def test_pytree_export_only(self):
    a = np.arange(4, dtype=np.float32)
    b = np.arange(6, dtype=np.float32)
    def f(a_b_pair, *, a, b):
      return (dict(res=a_b_pair, a=a, b=b), jnp.sin(a), jnp.cos(b))

    exp = jax_export.export(f, lowering_platform="cpu")((a, b), a=a, b=b)
    a_aval = core.ShapedArray(a.shape, a.dtype)
    b_aval = core.ShapedArray(b.shape, b.dtype)
    self.assertEqual(exp.lowering_platform, "cpu")
    args = ((a, b),)
    kwargs = dict(a=a, b=b)
    self.assertEqual(exp.in_tree, tree_util.tree_flatten((args, kwargs))[1])
    self.assertEqual(exp.in_avals, (a_aval, b_aval, a_aval, b_aval))
    self.assertEqual(exp.out_tree, tree_util.tree_flatten(f(*args, **kwargs))[1])
    self.assertEqual(exp.out_avals, (a_aval, b_aval, a_aval, b_aval, a_aval, b_aval))

  def test_poly_export_only(self):
    a = np.arange(12, dtype=np.float32).reshape((3, 4))
    def f(a, b):  # a: f32[2w,h]  b: f32[w,h]
      return jnp.concatenate([a, b], axis=0)

    exp = jax_export.export(f)(
        jax_export.poly_spec(a.shape, a.dtype, "(2*w, h)"),
        jax_export.poly_spec(a.shape, a.dtype, "(w, h)"))
    self.assertEqual("(2*w, h)", str(exp.in_avals[0].shape))
    self.assertEqual("(w, h)", str(exp.in_avals[1].shape))
    self.assertEqual("(3*w, h)", str(exp.out_avals[0].shape))

  def test_poly_pytree_export_only(self):
    a = np.arange(12, dtype=np.float32).reshape((3, 4))
    def f(a0, a1, *, ak):
      return jnp.concatenate([a0, a1, ak], axis=0)

    a_poly_spec = jax_export.poly_spec(a.shape, a.dtype, "(w, h)")
    exp = jax_export.export(f)(a_poly_spec, a_poly_spec, ak=a_poly_spec)
    self.assertEqual("(w, h)", str(exp.in_avals[0].shape))
    self.assertEqual("(3*w, h)", str(exp.out_avals[0].shape))

  def test_basic(self):
    f = jnp.sin
    x = np.arange(4, dtype=np.float32)
    exp_f = jax_export.export(f)(x)

    f1 = jax_export.call_exported(exp_f)
    self.assertAllClose(f(x), f1(x))

  def test_call_exported_lambda(self):
    # When we export a lambda, the exported.fun_name is not a valid MLIR function name
    f = lambda x: jnp.sin(x)
    x = np.arange(4, dtype=np.float32)
    exp_f = jax_export.export(f)(x)
    f1 = jax_export.call_exported(exp_f)
    self.assertAllClose(f(x), f1(x))

  def test_call_twice_exported(self):
    def f(x): return jnp.sin(x)
    x = np.arange(4, dtype=np.float32)

    @jax.jit
    def f1(x):
      exp_f = jax_export.export(f)(x)
      return jax_export.call_exported(exp_f)(x) + jax_export.call_exported(exp_f)(x)

    self.assertAllClose(2. * f(x), f1(x))

  def test_unused_args(self):
    f = lambda x, y: jnp.sin(x)
    x = np.arange(4, dtype=np.float32)
    y = np.arange(6, dtype=np.float32)
    exp_f = jax_export.export(f)(x, y)

    f1 = jax_export.call_exported(exp_f)
    self.assertAllClose(f(x, y), f1(x, y))

  def test_pytree(self):
    a = np.arange(4, dtype=np.float32)
    b = np.arange(6, dtype=np.float32)
    def f(a_b_pair, a, b):
      return (dict(res=a_b_pair, a=a, b=b), jnp.sin(a), jnp.cos(b))

    exp_f = jax_export.export(f)((a, b), a=a, b=b)
    f1 = jax_export.call_exported(exp_f)
    self.assertAllClose(f((a, b), a=a, b=b),
                        f1((a, b), a=a, b=b))

  def test_error_wrong_intree(self):
    def f(a_b_pair, *, c):
      return jnp.sin(a_b_pair[0]) + jnp.cos(a_b_pair[1]) + c
    a = b = c = np.arange(4, dtype=np.float32)
    exp_f = jax_export.export(f)((a, b), c=c)

    with self.assertRaisesRegex(
        ValueError,
        "The invocation args and kwargs must have the same pytree structure"):
      jax_export.call_exported(exp_f)(a, b, c=(a, b))

  def test_error_wrong_avals(self):
    def f(a, *, b):  # a: f32[4] and b: f32[4]
      return jnp.sin(a) + jnp.cos(b)
    f32_4 = np.arange(4, dtype=np.float32)
    exp_f = jax_export.export(f)(f32_4, b=f32_4)

    with self.assertRaisesRegex(ValueError,
        r"Shape mismatch for args\[0\].shape\[0\]"):
      jax_export.call_exported(exp_f)(np.arange(6, dtype=np.float32), b=f32_4)

    with self.assertRaisesRegex(ValueError,
        r"Shape mismatch for kwargs\['b'\].shape\[0\]"):
      jax_export.call_exported(exp_f)(f32_4, b=np.arange(6, dtype=np.float32))

    with self.assertRaisesRegex(ValueError,
        r"Rank mismatch for args\[0\]"):
      jax_export.call_exported(exp_f)(f32_4.reshape((1, 4)), b=f32_4)

    with self.assertRaisesRegex(ValueError,
        r"Dtype mismatch for args\[0\]"):
      jax_export.call_exported(exp_f)(f32_4.astype(np.float16), b=f32_4)

  @parameterized.named_parameters(
      dict(testcase_name=p, platform=p)
      for p in ("cpu", "cuda", "rocm", "tpu"))
  def test_error_wrong_platform(self, platform):
    a = np.arange(4, dtype=np.float32)

    exp_f = jax_export.export(jnp.sin, lowering_platform=platform)(a)
    if xb.canonicalize_platform(jtu.device_under_test()) == platform:
      raise unittest.SkipTest("")

    with self.assertRaisesRegex(
        ValueError, "The exported function .* was lowered for platform"):
      jax_export.call_exported(exp_f)(a)

  def test_grad(self):
    f = lambda x: jnp.sum(jnp.sin(x))
    x = np.arange(4, dtype=np.float32)
    exp_f = jax_export.export(f)(x)

    f1 = jax_export.call_exported(exp_f)
    self.assertAllClose(jax.grad(f)(x), jax.grad(f1)(x))

  def test_pytree_vjp(self):
    def f(a_b_pair, *, a, b):
      return (dict(res=a_b_pair, a=2. * a, b=3. * b),
              jnp.sin(4. * a))

    a = np.arange(4, dtype=np.float32)
    b = np.arange(6, dtype=np.float32)
    exp_f = jax_export.export(f)((a, b), a=a, b=b)

    out_ct = f((a, b), a=a, b=b)  # The output has the right structure as the cotangent
    def f1_jax(a, b):  # For VJP, make a function without kwargs
      res = f((a, b), a=a, b=b)
      return res
    def f1_exp(a, b):  # For VJP, make a function without kwargs
      res = jax_export.call_exported(exp_f)((a, b), a=a, b=b)
      return res
    jax_vjp = jax.vjp(f1_jax, a, b)[1](out_ct)
    exp_vjp = jax.vjp(f1_exp, a, b)[1](out_ct)
    self.assertAllClose(jax_vjp, exp_vjp)

  def test_roundtrip(self):
    def f1(x):
      return jnp.sin(x)
    a = np.arange(4, dtype=np.float32)
    exp_f1 = jax_export.export(f1)(a)
    def f2(x):
      res1 = jax_export.call_exported(exp_f1)(x)
      res2 = jax_export.call_exported(exp_f1)(res1)
      return jnp.cos(res2)
    exp_f2 = jax_export.export(f2)(a)

    self.assertAllClose(jnp.cos(jnp.sin(jnp.sin(a))),
                        jax_export.call_exported(exp_f2)(a))

  # An inner function is exported with polymorphic shapes inner_poly_spec, and
  # is called from an outer function, which is exported with outer_poly_spec.
  @parameterized.named_parameters(
      dict(testcase_name=f"inner={d['inner_poly_spec']}_outer={d['outer_poly_spec']}",  # type: ignore
           **d)  # type: ignore
      for d in (
          dict(inner_poly_spec="3,a,a+b", outer_poly_spec="3,4,12"),
          dict(inner_poly_spec="3,a,a+b", outer_poly_spec="3,4,c"),
          dict(inner_poly_spec="3,a,a+b", outer_poly_spec="3,c,c",
               expect_error=(
                   r"Dimension variable 'b' must have integer value >= 1. "
                   r"Found 0 when solving a \+ b == args\[0\].shape\[2\]")),
          dict(inner_poly_spec="3,a,a+b", outer_poly_spec="c,4,12",
               expect_error=r"Shape mismatch for args\[0\].shape\[0\] \(expected constant\)"),
          dict(inner_poly_spec="3,a,a+b", outer_poly_spec="3,c+4,12"),  # TODO: This should be an error, c = 0
          dict(inner_poly_spec="3,4,3*a", outer_poly_spec="3,4,12"),
          dict(inner_poly_spec="3,4,5*a", outer_poly_spec="3,4,12",
               expect_error=(
                   r"Dimension variable 'a' must have integer value >= 1. "
                   r"Non-zero remainder 2 for factor 5 when solving 5\*a == args\[0\].shape\[2\]")),
          # dict(inner_poly_spec="3,4,5*a", outer_poly_spec="3,4,c"),  # TODO: there should be an error 5*a != c == 12
          # dict(inner_poly_spec="3,a,a", outer_poly_spec="3,a,a"),  # TODO: this should be a dynamic error
          dict(inner_poly_spec="3,a", inner_x_shape=(3, 4), outer_poly_spec="3,a,a",
               expect_error=r"Rank mismatch for args\[0\]"),
          dict(inner_poly_spec="3,a,a+b", inner_x_dtype=np.int32, outer_poly_spec="3,c,d",
               expect_error=r"Dtype mismatch for args\[0\]"),
      ))
  def test_poly(self, inner_poly_spec="3,a,a+b", inner_x_shape=(3, 4, 6),
                inner_x_dtype=np.float32,
                outer_poly_spec="3,c+4,12",  outer_x_shape=(3, 4, 12),
                expect_error=None):
    # Polymorphic export called with static or polymorphic shapes
    def inner(x):  # x: inner_poly_spec
      return jnp.reshape(x, (-1, x.shape[1]))

    inner_x = np.arange(np.prod(inner_x_shape),
                        dtype=inner_x_dtype).reshape(inner_x_shape)  # inner_x : f32[3,4,6]
    inner_exp = jax_export.export(inner)(
        jax_export.poly_spec(inner_x.shape, inner_x.dtype, inner_poly_spec))

    self.assertEqual(inner_exp.module_uses_dim_vars,
                     (inner_poly_spec != "3,4,12"))
    outer_x = np.arange(np.prod(outer_x_shape),
                        dtype=np.float32).reshape(outer_x_shape)  # outer_x : f32[3,4,12]
    def outer(x):  # x: outer_poly_spec
      # Use an addition to test that the shapes are refined properly for the
      # result of the call_exported.
      return jax_export.call_exported(inner_exp)(x) + inner(x)

    with contextlib.ExitStack() as stack:
      if expect_error is not None:
        stack.push(self.assertRaisesRegex(ValueError, expect_error))

      # Call it after exporting again, with polymorphic shapes
      outer_exp = jax_export.export(outer)(
          jax_export.poly_spec(outer_x.shape, outer_x.dtype, outer_poly_spec))
      self.assertEqual(outer_exp.module_uses_dim_vars,
                       (inner_poly_spec != "3,4,12" or outer_poly_spec != "3,4,12"))
      if not outer_exp.module_uses_dim_vars:
        res = jax_export.call_exported(outer_exp)(outer_x)
        self.assertAllClose(2. * inner(outer_x), res)
      else:
        # TODO: for now, we use XlaCallModule to run modules with polymorphic shapes
        # until we create the python bindings to invoke shape refinement.
        if jax2tf is not None:
          res = jax2tf._run_exported_as_tf([outer_x], outer_exp)[0].numpy()
          self.assertAllClose(2. * inner(outer_x), res)

  def test_call_poly(self):
    a_shape = (3, 4)
    a = np.arange(math.prod(a_shape), dtype=np.float32).reshape(a_shape)

    def f_inner(x):  # x: f32[w, h]
      return jnp.reshape(x, (-1,))

    exp_inner = jax_export.export(f_inner)(
        jax_export.poly_spec(a.shape, a.dtype, "w, h")
    )

    # There are dynamic shapes in the exported module
    self.assertIn("?x", exp_inner.mlir_module)
    self.assertIn("stablehlo.dynamic_reshape", exp_inner.mlir_module)

    # Add a wrapper "main" func with static shapes
    # TODO(necula): We will add this functionality to jax_export.
    from jax._src.interpreters import mlir
    from jax._src.lib.mlir import ir
    from jax._src.lib.mlir.dialects import hlo
    from jax._src.lib.mlir.dialects import func as func_dialect
    from jax.lib import xla_client as xc
    from jax._src.lib import xla_extension

    context = mlir.make_ir_context()
    with context, ir.Location.unknown(context):
      wrapped_module = ir.Module.parse(exp_inner.mlir_module)
      symbol_table = ir.SymbolTable(wrapped_module.operation)
      orig_main = symbol_table["main"]
      orig_main.attributes["sym_visibility"] = ir.StringAttr.get("private")
      symbol_table.set_symbol_name(orig_main, "_wrapped_jax_export_main")
      orig_main_name = ir.StringAttr(symbol_table.insert(orig_main)).value
      # Use static shapes
      new_main_input_types = [
          mlir.aval_to_ir_type(core.ShapedArray((3, 4), np.float32))
      ]
      orig_output_types = orig_main.type.results
      new_main_ftype = ir.FunctionType.get(
          new_main_input_types, orig_output_types
      )
      new_main_op = func_dialect.FuncOp(
          "main",
          new_main_ftype,
          ip=ir.InsertionPoint.at_block_begin(wrapped_module.body),
      )
      new_main_op.attributes["sym_visibility"] = ir.StringAttr.get("public")
      symbol_table.insert(new_main_op)
      entry_block = new_main_op.add_entry_block()
      with ir.InsertionPoint(entry_block):
        orig_main_args: List[ir.Value] = []
        for new_arg, orig_arg_type in zip(
            new_main_op.arguments, orig_main.type.inputs
        ):
          orig_main_args.append(hlo.ConvertOp(orig_arg_type, new_arg).result)
        call = func_dialect.CallOp(
            orig_output_types,
            ir.FlatSymbolRefAttr.get(orig_main_name),
            orig_main_args,
        )
        func_dialect.ReturnOp(call.results)
    symbol_table.set_symbol_name(new_main_op, "main")

    # TODO(necula): need conditionals until jaxlib 0.4.12 is the minimum version
    if xc.mlir_api_version >= 50:
      refined_module_str = xla_extension.mlir.refine_polymorphic_shapes(
          mlir.module_to_bytecode(wrapped_module)
      )
      context = mlir.make_ir_context()
      with context:
        refined_module = ir.Module.parse(refined_module_str)

      logging.info("Postprocessed module %s", str(refined_module))
      self.assertNotIn("?x", str(refined_module))
      self.assertNotIn("stablehlo.dynamic_reshape", str(refined_module))
      self.assertIn("stablehlo.reshape", str(refined_module))


if __name__ == "__main__":
  absltest.main(testLoader=jtu.JaxTestLoader())