[utils] integrated colotensor with lazy init context (#1324)

* [utils] integrated colotensor with lazy init context

* polish code

* polish code

* polish code

colossalai/utils/model/lazy_init_context.py

@@ -2,13 +2,13 @@
 # coding: utf-8
 
 import torch
-from colossalai.tensor import ColoParameter
+import torch.nn as nn
+from colossalai.tensor import ColoParameter, ColoTensor
+
 import types
 import inspect
-import typing
 from typing import List, Callable
 from colossalai.utils.model.utils import substitute_init_recursively
-import copy
 
 
 class LazyInitContext():
@@ -18,8 +18,7 @@ class LazyInitContext():
     
     Note:
         This API is only experimental and subject to future changes. 
-        It should be integrated with meta tensor initialization in the future.
-        
+
     Usage:
         with LazyInitContext() as ctx:
             model = nn.Linear(10, 10)
@@ -36,14 +35,17 @@ class LazyInitContext():
         assert not model.weight.is_meta and torch.all(model.weight == 0)
         
     Args:
+        to_meta (bool): optional, whether to initialize the model with meta tensors, default is False.
         extra_torch_tensor_func (List[str]): extra torch tensor functions related 
             to value setting, such as `zero_` and `triu_`. `zero_` is pre-added by default.
     """
 
-    tensor_set_value_func = ['zero_']
+    tensor_set_value_func = ['zero_', 'fill_']
 
-    def __init__(self, extra_torch_tensor_func: List[str] = None):
-        self._intercepted_init_func_cache = []
+    def __init__(self, to_meta: bool = False, extra_torch_tensor_func: List[str] = None):
+        # TODO: hijack the torch constructor functions as well
+        self._to_meta = to_meta
+        self._intercepted_nn_init_func_cache = {}
         self._nn_init_methods = self._get_nn_init_methods()
         self._torch_mod_cls = torch.nn.modules.module.Module
 
@@ -53,14 +55,20 @@ class LazyInitContext():
         else:
             self._torch_tensor_funcs = self.tensor_set_value_func
 
-    def _cache_func(self, func):
+    @property
+    def to_meta(self):
+        return self._to_meta
+
+    def _cache_init_func(self, func):
         """
-        This method wraps the ``torch.nn.init`` method so that the function call
-        is cached instead of being executed.
+        This method wraps the ``torch.nn.init`` method and torch tensor value-setting functions
+        so that the function call is cached instead of being executed.
         """
 
-        def wrapped_init_func(*args, **kwargs):
-            self._intercepted_init_func_cache.append(dict(func=func, args=args, kwargs=kwargs))
+        def wrapped_init_func(tensor, *args, **kwargs):
+            if tensor not in self._intercepted_nn_init_func_cache:
+                self._intercepted_nn_init_func_cache[tensor] = []
+            self._intercepted_nn_init_func_cache[tensor].append((func, args, kwargs))
 
         return wrapped_init_func
 
@@ -76,17 +84,10 @@ class LazyInitContext():
         for name in nn_init_method_names:
             nn_init_methods.append((name, getattr(torch.nn.init, name)))
 
-        def _has_tensor_in_arg(func):
-            hints = typing.get_type_hints(func)
-            for k, v in hints.items():
-                if v is torch.Tensor:
-                    return True
-            return False
-
         def _is_init_method(item):
             name, func = item
-            if (not isinstance(func, types.FunctionType) or name.startswith('_') or not name.endswith('_')
-                    or not _has_tensor_in_arg(func)):
+
+            if (not isinstance(func, types.FunctionType) or name.startswith('_') or not name.endswith('_')):
                 return False
             else:
                 return True
@@ -103,11 +104,13 @@ class LazyInitContext():
         has_device = 'device' in inspect.signature(func).parameters
 
         def layer_lazy_init(module, *args, **kwargs):
-            self._intercepted_init_func_cache.append(
-                dict(func=func, module=module, args=args, kwargs=copy.deepcopy(kwargs)))
+            # if this module contains device argument
+            # we set it to meta to initialize as meta backend
             if has_device:
                 kwargs['device'] = 'meta'
             func(module, *args, **kwargs)
+
+            # if device is not found, we intialize it and convert to meta
             if not has_device:
                 module.to('meta')
 
@@ -122,7 +125,7 @@ class LazyInitContext():
     def _patch_nn_init_funcs(self):
         # patch nn.init functions
         for name, func in self._nn_init_methods:
-            setattr(torch.nn.init, name, self._cache_func(func))
+            setattr(torch.nn.init, name, self._cache_init_func(func))
 
     def _unpatch_nn_init_funcs(self):
         # unpatch nn.init functions
@@ -150,7 +153,7 @@ class LazyInitContext():
             origin_func_name = self._get_tmp_origin_func_ref(func_name)
             origin_func = getattr(torch.Tensor, func_name)
             setattr(torch.Tensor, origin_func_name, origin_func)
-            setattr(torch.Tensor, func_name, self._cache_func(origin_func))
+            setattr(torch.Tensor, func_name, self._cache_init_func(origin_func))
 
     def _unpatch_torch_tensor_funcs(self):
         for func_name in self._torch_tensor_funcs:
@@ -159,17 +162,18 @@ class LazyInitContext():
             setattr(torch.Tensor, func_name, origin_func)
 
     def __enter__(self):
-        self._patch_submodule_init()
+        self._patch_torch_tensor_funcs()
+        self._patch_nn_init_funcs()
+
+        if self._to_meta:
+            self._patch_submodule_init()
         return self
 
     def __exit__(self, *args, **kwargs):
-        self._unpatch_submodule_init()
-        # build model_rebuild_dict in reverse order to make sure get correct init func for inherited class.
-        self.module_rebuild_dict = {}
-        self._intercepted_init_func_cache.reverse()
-        for cache in self._intercepted_init_func_cache:
-            self.module_rebuild_dict[cache['module']] = (cache['func'], cache['args'], cache['kwargs'])
-        self._intercepted_init_func_cache.reverse()
+        if self._to_meta:
+            self._unpatch_submodule_init()
+        self._unpatch_nn_init_funcs()
+        self._unpatch_torch_tensor_funcs()
 
     def lazy_init_parameters(self, model: torch.nn.Module, device='cpu', call_back: Callable = None):
         """
@@ -178,80 +182,56 @@ class LazyInitContext():
         Args:
             model (`torch.nn.Module`): the model instantiated under the context.
             device (str): the device on which weights are initialized
-            
+
         """
-        # build param mapping
-        param_id_to_name = dict()
-        for name, param in model.named_parameters():
-            param_id_to_name[id(param)] = name
-        for name, buffer in model.named_buffers():
-            param_id_to_name[id(buffer)] = name
-
-        assert model in self.module_rebuild_dict, 'We only support rebuild modules which intercepted during initializing by us.'
-
-        def _process_arg(arg):
-            """
-            Process args recursively. If arg is a torch.nn.Module instance in module_rebuild_dict, 
-            we need to rebuild it with real parameters. If arg is a tuple or list, we will process
-            the element of arg with this function again.
-            """
-            if torch.is_tensor(arg):
-                tensor_id = id(arg)
-                if tensor_id in param_id_to_name:
-                    arg = _replace_meta_param_with_real_param(arg)
-
-            elif isinstance(arg, torch.nn.Module):
-                if arg in self.module_rebuild_dict:
-                    arg = self.lazy_init_parameters(model=arg, device=device, call_back=call_back)
-
-            elif isinstance(arg, (tuple, list)):
-                rst_list = []
-                for element in arg:
-                    processed_element = _process_arg(element)
-                    rst_list.append(processed_element)
-                arg = rst_list
-            return arg
-
-        def _replace_meta_param_with_real_param(meta_param):
-            if meta_param.device != 'meta':
-                return meta_param
-            tensor_id = id(meta_param)
-            param_full_name = param_id_to_name[tensor_id]
-            real_param = torch.empty_like(meta_param, dtype=meta_param.dtype, device=device)
-            real_param = ColoParameter(real_param, requires_grad=meta_param.requires_grad)
-
-            if '.' in param_full_name:
-                submodule_name, param_name = param_full_name.rsplit('.', 1)
-                submodule = model.get_submodule(submodule_name)
-            else:
-                submodule = model
-                param_name = param_full_name
-            setattr(submodule, param_name, real_param)
-
-            # execute call_back function on the materailized tensor
-            # this can where sharding comes in
-            if call_back:
-                call_back(real_param)
-            return real_param
-
-        func, args, kwargs = self.module_rebuild_dict[model]
-        args = list(args)
-
-        # check args for parameter replacement
-        for idx, arg in enumerate(args):
-            arg = _process_arg(arg)
-            args[idx] = arg
-
-        # check kwargs for parameter replacement
-        for arg_name, arg in kwargs.items():
-            if arg_name == 'device':
-                arg = device
+
+        def _init_recursively(module: nn.Module):
+            # recursively initialize the module
+            for mod in module.children():
+                _init_recursively(mod)
+
+            # initialize and shard tensors directly attached to the current module
+            for name, param in module.named_parameters(recurse=False):
+                _init_and_shard(module, name, param)
+
+            for name, buf in module.named_buffers(recurse=False):
+                _init_and_shard(module, name, buf)
+
+        @torch.no_grad()
+        def _init_and_shard(module, name, tensor):
+            # check whether the tensor is a buffer or parameter
+            is_param = isinstance(tensor, nn.parameter.Parameter)
+
+            # get sharding spec
+            dist_spec = getattr(tensor, 'dist_spec', None)
+            pg = getattr(tensor, 'pg', None)
+
+            # convert the tensor from meta to materialized one
+            if tensor.is_meta:
+                materialized_tensor = torch.empty_like(tensor, device=device)
+                # if this tensor is a meta tensor, it must have an init function
+                assert tensor in self._intercepted_nn_init_func_cache
+                tensor = materialized_tensor
+
+            # apply init function
+            if tensor in self._intercepted_nn_init_func_cache:
+                init_func, args, kwargs = self._intercepted_nn_init_func_cache[tensor][-1]
+                init_func(tensor, *args, **kwargs)
+
+            # convert it to ColoTensor or ColoParameter
+            if is_param:
+                tensor = ColoParameter.from_torch_tensor(tensor, requires_grad=tensor.requires_grad)
             else:
-                arg = _process_arg(arg)
-            kwargs[arg_name] = arg
+                tensor = ColoTensor.from_torch_tensor(tensor)
+
+            # apply sharding
+            if dist_spec:
+                tensor = tensor.redistribute(dist_spec=dist_spec, pg=pg)
+
+            # override the original tensor
+            with torch.no_grad():
+                setattr(module, name, tensor)
 
-        # build user specified model
-        with torch.no_grad():
-            func(model, *args, **kwargs)
+        _init_recursively(model)
 
         return model

tests/test_utils/test_lazy_init_ctx.py

@@ -10,27 +10,42 @@ np.random.seed(MANUAL_SEED)
 torch.manual_seed(MANUAL_SEED)
 
 
-def test_lazy_init():
-    cpu_rng_state = torch.get_rng_state()
-    origin_model = resnet34(num_classes=10)
-    origin_param_dict = dict(origin_model.named_parameters())
-    torch.set_rng_state(cpu_rng_state)
-    ctx = LazyInitContext()
+def test_lazy_init_with_meta():
+    ctx = LazyInitContext(to_meta=True)
     with ctx:
         model = resnet34(num_classes=10)
+
     for param in model.parameters():
         assert param.is_meta
     for buffer in model.buffers():
         assert buffer.is_meta
+
     ctx.lazy_init_parameters(model)
+
+    for name, param in model.named_parameters():
+        assert not param.is_meta, name
+
+    for buffer in model.buffers():
+        assert not buffer.is_meta
+
+
+def test_lazy_init_without_meta():
+    ctx = LazyInitContext(to_meta=False)
+    with ctx:
+        model = resnet34(num_classes=10)
+
     for param in model.parameters():
         assert not param.is_meta
     for buffer in model.buffers():
         assert not buffer.is_meta
-    param_dict = dict(model.named_parameters())
-    for key in origin_param_dict.keys():
-        assert origin_param_dict[key].data.equal(param_dict[key].data)
+
+    conv1_weight_before_init = model.conv1.weight.clone()
+    ctx.lazy_init_parameters(model)
+    conv1_weight_after_init = model.conv1.weight.clone()
+
+    assert not torch.allclose(conv1_weight_after_init, conv1_weight_before_init)
 
 
 if __name__ == '__main__':
-    test_lazy_init()
+    test_lazy_init_with_meta()
+    test_lazy_init_without_meta()