search

iree-org / iree-turbine

IREE's PyTorch Frontend, based on Torch Dynamo.
Apache License 2.0
42 stars 21 forks source link

can not use torch.compile with simple gpt2 test #117

Open vivekvpandya opened 1 week ago

vivekvpandya commented 1 week ago 
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

@torch.compile(backend="turbine_cpu")
def test_gpt2_demo():
    tokenizer = AutoTokenizer.from_pretrained("gpt2")
    model = AutoModelForCausalLM.from_pretrained("gpt2").to(torch.float)
    prompt = "Thanks for"
    print("\nInput prompt: ", prompt, "\n")

    # run on CPU
    input_ids = tokenizer(prompt, return_tensors="pt").input_ids
    gen_tokens = model.generate(
        input_ids,
        max_length=4,
        do_sample=False,
        pad_token_id=tokenizer.eos_token_id,
    )
    gen_text = tokenizer.batch_decode(gen_tokens)[0]

    print("CPU output: ", gen_text, "\n")

test_gpt2_demo()

I am getting following error

/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/functions.py:663: UserWarning: Graph break due to unsupported builtin posix.fspath. This function is either a Python builtin (e.g. _warnings.warn) or a third-party C/C++ Python extension (perhaps created with pybind). If it is a Python builtin, please file an issue on GitHub so the PyTorch team can add support for it and see the next case for a workaround. If it is a third-party C/C++ Python extension, please either wrap it into a PyTorch-understood custom operator (see https://pytorch.org/tutorials/advanced/custom_ops_landing_page.html for more details) or, if it is traceable, use torch.compiler.allow_in_graph.
  torch._dynamo.utils.warn_once(msg)
/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
  warnings.warn(
W0904 11:46:21.351000 140371205161088 torch/_dynamo/convert_frame.py:762] [6/8] torch._dynamo hit config.cache_size_limit (8)
W0904 11:46:21.351000 140371205161088 torch/_dynamo/convert_frame.py:762] [6/8]    function: '__init_subclass__' (/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/utils/generic.py:324)
W0904 11:46:21.351000 140371205161088 torch/_dynamo/convert_frame.py:762] [6/8]    last reason: ___check_obj_id(L['cls'], 165879472)                        
W0904 11:46:21.351000 140371205161088 torch/_dynamo/convert_frame.py:762] [6/8] To log all recompilation reasons, use TORCH_LOGS="recompiles".
W0904 11:46:21.351000 140371205161088 torch/_dynamo/convert_frame.py:762] [6/8] To diagnose recompilation issues, see https://pytorch.org/docs/main/torch.compiler_troubleshooting.html.
/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/functions.py:663: UserWarning: Graph break due to unsupported builtin posix.stat. This function is either a Python builtin (e.g. _warnings.warn) or a third-party C/C++ Python extension (perhaps created with pybind). If it is a Python builtin, please file an issue on GitHub so the PyTorch team can add support for it and see the next case for a workaround. If it is a third-party C/C++ Python extension, please either wrap it into a PyTorch-understood custom operator (see https://pytorch.org/tutorials/advanced/custom_ops_landing_page.html for more details) or, if it is traceable, use torch.compiler.allow_in_graph.
  torch._dynamo.utils.warn_once(msg)
Traceback (most recent call last):
  File "/local/mnt/workspace/users/vpandya/iree-turbine/examples/simple_gpt2.py", line 24, in <module>
    test_gpt2_demo()
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/eval_frame.py", line 433, in _fn
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/examples/simple_gpt2.py", line 5, in test_gpt2_demo
    @torch.compile(backend="turbine_cpu")
  File "/local/mnt/workspace/users/vpandya/iree-turbine/examples/simple_gpt2.py", line 7, in torch_dynamo_resume_in_test_gpt2_demo_at_7
    tokenizer = AutoTokenizer.from_pretrained("gpt2")
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/models/auto/auto_factory.py", line 564, in from_pretrained
    return model_class.from_pretrained(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 2927, in from_pretrained
    @classmethod
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3245, in torch_dynamo_resume_in_from_pretrained_at_3245
    if not isinstance(config, PretrainedConfig):
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3352, in torch_dynamo_resume_in_from_pretrained_at_3352
    if not isinstance(config, PretrainedConfig):
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3377, in torch_dynamo_resume_in_from_pretrained_at_3377
    config = copy.deepcopy(config)
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3431, in torch_dynamo_resume_in_from_pretrained_at_3431
    is_local = os.path.isdir(pretrained_model_name_or_path)
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3509, in torch_dynamo_resume_in_from_pretrained_at_3509
    elif os.path.isfile(os.path.join(subfolder, pretrained_model_name_or_path)):
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3509, in torch_dynamo_resume_in_from_pretrained_at_3509
    elif os.path.isfile(os.path.join(subfolder, pretrained_model_name_or_path)):
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3512, in torch_dynamo_resume_in_from_pretrained_at_3512
    elif os.path.isfile(os.path.join(subfolder, pretrained_model_name_or_path + ".index")):
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3512, in torch_dynamo_resume_in_from_pretrained_at_3512
    elif os.path.isfile(os.path.join(subfolder, pretrained_model_name_or_path + ".index")):
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3837, in torch_dynamo_resume_in_from_pretrained_at_3520
    # Check first if we are `from_pt`
                ^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/models/gpt2/modeling_gpt2.py", line 1190, in __init__
    self.transformer = GPT2Model(config)
                       ^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/models/gpt2/modeling_gpt2.py", line 911, in __init__
    self.post_init()
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/convert_frame.py", line 1116, in __call__
    return self._torchdynamo_orig_callable(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/convert_frame.py", line 948, in __call__
    result = self._inner_convert(
             ^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/convert_frame.py", line 472, in __call__
    return _compile(
           ^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_utils_internal.py", line 84, in wrapper_function
    return StrobelightCompileTimeProfiler.profile_compile_time(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_strobelight/compile_time_profiler.py", line 129, in profile_compile_time
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.11/contextlib.py", line 81, in inner
    return func(*args, **kwds)
           ^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/convert_frame.py", line 817, in _compile
    guarded_code = compile_inner(code, one_graph, hooks, transform)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/utils.py", line 231, in time_wrapper
    r = func(*args, **kwargs)
        ^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/convert_frame.py", line 636, in compile_inner
    out_code = transform_code_object(code, transform)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/bytecode_transformation.py", line 1185, in transform_code_object
    transformations(instructions, code_options)
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/convert_frame.py", line 178, in _fn
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/convert_frame.py", line 582, in transform
    tracer.run()
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 2451, in run
    super().run()
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 893, in run
    while self.step():
          ^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 805, in step
    self.dispatch_table[inst.opcode](self, inst)
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 499, in wrapper
    return inner_fn(self, inst)
           ^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 2059, in CALL
    self.call_function(fn, args, kwargs)
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 743, in call_function
    self.push(fn.call_function(self, args, kwargs))
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/functions.py", line 344, in call_function
    return super().call_function(tx, args, kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/functions.py", line 293, in call_function
    return super().call_function(tx, args, kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/functions.py", line 90, in call_function
    return tx.inline_user_function_return(self, [*self.self_args(), *args], kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 749, in inline_user_function_return
    return InliningInstructionTranslator.inline_call(self, fn, args, kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 2666, in inline_call
    return cls.inline_call_(parent, func, args, kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 2782, in inline_call_
    tracer.run()
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 893, in run
    while self.step():
          ^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 805, in step
    self.dispatch_table[inst.opcode](self, inst)
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 1561, in LOAD_ATTR
    self._load_attr(inst)
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/symbolic_convert.py", line 1551, in _load_attr
    result = BuiltinVariable(getattr).call_function(
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/builtin.py", line 962, in call_function
    return handler(tx, args, kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/builtin.py", line 846, in builtin_dipatch
    rv = fn(tx, args, kwargs)
         ^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/builtin.py", line 764, in call_self_handler
    result = self_handler(tx, *args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/builtin.py", line 1626, in call_getattr
    return obj.var_getattr(tx, name)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/dicts.py", line 914, in var_getattr
    return ConstantVariable.create(getattr(self.obj, name))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local/mnt/workspace/users/vpandya/iree-turbine/install/torch/_dynamo/variables/constant.py", line 39, in create
    assert not isinstance(value, disallowed_type), reason
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: Dict types must use ConstDictVariable.

from user code:
   File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 1406, in post_init
    self.init_weights()
  File "/local/mnt/workspace/users/vpandya/.local/lib/python3.11/site-packages/transformers/modeling_utils.py", line 2316, in init_weights
    if self.config.pruned_heads:

Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information

You can suppress this exception and fall back to eager by setting:
    import torch._dynamo
    torch._dynamo.config.suppress_errors = True
ScottTodd commented 1 week ago

That error looks more related to PyTorch itself than iree-turbine. Do other backends for torch.compile work with that code? What versions of Python packages (notably torch) are you using?

vivekvpandya commented 1 week ago

That error looks more related to PyTorch itself than iree-turbine. Do other backends for torch.compile work with that code? What versions of Python packages (notably torch) are you using?

Python 3.11.9 (main, Apr 27 2024, 21:16:11) [GCC 13.2.0] on linux Type "help", "copyright", "credits" or "license" for more information.

import torch print(torch.version) 2.4.1+cpu