Hard error when ignoring tensors. (#27484) (#29906)

* Hard error when ignoring tensors. (#27484)

* [WIP] Hard error when ignoring tensors.

* Better selection/error when saving a checkpoint.

- Find all names we should normally drop (those are in the transformers
  config)
- Find all disjoint tensors (for those we can safely trigger a copy to
  get rid of the sharing before saving)
- Clone those disjoint tensors getting rid of the issue
- Find all identical names (those should be declared in the config
  but we try to find them all anyway.)
- For all identical names:
  - If they are in the config, just ignore them everything is fine
  - If they are not, warn about them.
- For all remainder tensors which are shared yet neither identical NOR
  disjoint. raise a hard error.

* Adding a failing test on `main` that passes here.

* We don't need to keep the subfolder logic in this test.

* Apply suggestions from code review

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

---------

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* Add small tests.

* Dead variable.

* Fixup.

* Fixing tied_Weights_keys on generic models.

* Fixup + T5 encoder/decoder tying (with different layers)

* Code quality.

* Dynamic member.

* trigger

* Fixing encoder name for other types of encoder/decoder combos.

* Fix scoping.

* Update .github/workflows/self-scheduled.yml

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* Fixing the tied_weights after the call.

---------

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/modeling_utils.py

@@ -30,7 +30,7 @@ from contextlib import contextmanager
 from dataclasses import dataclass
 from functools import partial, wraps
 from threading import Thread
-from typing import Any, Callable, Dict, List, Optional, Tuple, Union
+from typing import Any, Callable, Dict, List, Optional, Set, Tuple, Union
 from zipfile import is_zipfile
 
 import torch
@@ -573,6 +573,79 @@ def set_initialized_submodules(model, state_dict_keys):
     return not_initialized_submodules
 
 
+def _end_ptr(tensor: torch.Tensor) -> int:
+    # extract the end of the pointer if the tensor is a slice of a bigger tensor
+    if tensor.nelement():
+        stop = tensor.view(-1)[-1].data_ptr() + tensor.element_size()
+    else:
+        stop = tensor.data_ptr()
+    return stop
+
+
+def _get_tied_weight_keys(module: nn.Module, prefix=""):
+    tied_weight_keys = []
+    if getattr(module, "_tied_weights_keys", None) is not None:
+        names = [f"{prefix}.{k}" if prefix else k for k in module._tied_weights_keys]
+        tied_weight_keys.extend(names)
+    if getattr(module, "_dynamic_tied_weights_keys", None) is not None:
+        names = [f"{prefix}.{k}" if prefix else k for k in module._dynamic_tied_weights_keys]
+        tied_weight_keys.extend(names)
+    for name, submodule in module.named_children():
+        local_prefix = f"{prefix}.{name}" if prefix else name
+        tied_weight_keys.extend(_get_tied_weight_keys(submodule, prefix=local_prefix))
+    return tied_weight_keys
+
+
+def _find_disjoint(tensors: List[Set[str]], state_dict: Dict[str, torch.Tensor]) -> Tuple[List[Set[str]], List[str]]:
+    filtered_tensors = []
+    for shared in tensors:
+        if len(shared) < 2:
+            filtered_tensors.append(shared)
+            continue
+
+        areas = []
+        for name in shared:
+            tensor = state_dict[name]
+            areas.append((tensor.data_ptr(), _end_ptr(tensor), name))
+        areas.sort()
+
+        _, last_stop, last_name = areas[0]
+        filtered_tensors.append({last_name})
+        for start, stop, name in areas[1:]:
+            if start >= last_stop:
+                filtered_tensors.append({name})
+            else:
+                filtered_tensors[-1].add(name)
+            last_stop = stop
+    disjoint_tensors = []
+    shared_tensors = []
+    for tensors in filtered_tensors:
+        if len(tensors) == 1:
+            disjoint_tensors.append(tensors.pop())
+        else:
+            shared_tensors.append(tensors)
+    return shared_tensors, disjoint_tensors
+
+
+def _find_identical(tensors: List[Set[str]], state_dict: Dict[str, torch.Tensor]) -> Tuple[List[Set[str]], Set[str]]:
+    shared_tensors = []
+    identical = []
+    for shared in tensors:
+        if len(shared) < 2:
+            continue
+
+        areas = collections.defaultdict(set)
+        for name in shared:
+            tensor = state_dict[name]
+            area = (tensor.device, tensor.data_ptr(), _end_ptr(tensor))
+            areas[area].add(name)
+        if len(areas) == 1:
+            identical.append(shared)
+        else:
+            shared_tensors.append(shared)
+    return shared_tensors, identical
+
+
 def _load_state_dict_into_model(model_to_load, state_dict, start_prefix):
     # Convert old format to new format if needed from a PyTorch state_dict
     old_keys = []
@@ -1646,15 +1719,24 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
         if getattr(self.config, "is_encoder_decoder", False) and getattr(self.config, "tie_encoder_decoder", False):
             if hasattr(self, self.base_model_prefix):
                 self = getattr(self, self.base_model_prefix)
-            self._tie_encoder_decoder_weights(self.encoder, self.decoder, self.base_model_prefix)
+            tied_weights = self._tie_encoder_decoder_weights(
+                self.encoder, self.decoder, self.base_model_prefix, "encoder"
+            )
+            # Setting a dynamic variable instead of `_tied_weights_keys` because it's a class
+            # attributed not an instance member, therefore modifying it will modify the entire class
+            # Leading to issues on subsequent calls by different tests or subsequent calls.
+            self._dynamic_tied_weights_keys = tied_weights
 
         for module in self.modules():
             if hasattr(module, "_tie_weights"):
                 module._tie_weights()
 
     @staticmethod
-    def _tie_encoder_decoder_weights(encoder: nn.Module, decoder: nn.Module, base_model_prefix: str):
+    def _tie_encoder_decoder_weights(
+        encoder: nn.Module, decoder: nn.Module, base_model_prefix: str, base_encoder_name: str
+    ):
         uninitialized_encoder_weights: List[str] = []
+        tied_weights: List[str] = []
         if decoder.__class__ != encoder.__class__:
             logger.info(
                 f"{decoder.__class__} and {encoder.__class__} are not equal. In this case make sure that all encoder"
@@ -1665,8 +1747,11 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
             decoder_pointer: nn.Module,
             encoder_pointer: nn.Module,
             module_name: str,
+            base_encoder_name: str,
             uninitialized_encoder_weights: List[str],
             depth=0,
+            total_decoder_name="",
+            total_encoder_name="",
         ):
             assert isinstance(decoder_pointer, nn.Module) and isinstance(
                 encoder_pointer, nn.Module
@@ -1674,8 +1759,10 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
             if hasattr(decoder_pointer, "weight"):
                 assert hasattr(encoder_pointer, "weight")
                 encoder_pointer.weight = decoder_pointer.weight
+                tied_weights.append(f"{base_encoder_name}{total_encoder_name}.weight")
                 if hasattr(decoder_pointer, "bias"):
                     assert hasattr(encoder_pointer, "bias")
+                    tied_weights.append(f"{base_encoder_name}{total_encoder_name}.bias")
                     encoder_pointer.bias = decoder_pointer.bias
                 return
 
@@ -1713,19 +1800,26 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                         decoder_modules[decoder_name],
                         encoder_modules[encoder_name],
                         module_name + "/" + name,
+                        base_encoder_name,
                         uninitialized_encoder_weights,
                         depth=depth + 1,
+                        total_encoder_name=f"{total_encoder_name}.{encoder_name}",
+                        total_decoder_name=f"{total_decoder_name}.{decoder_name}",
                     )
                     all_encoder_weights.remove(module_name + "/" + encoder_name)
 
                 uninitialized_encoder_weights += list(all_encoder_weights)
 
         # tie weights recursively
-        tie_encoder_to_decoder_recursively(decoder, encoder, base_model_prefix, uninitialized_encoder_weights)
+        tie_encoder_to_decoder_recursively(
+            decoder, encoder, base_model_prefix, base_encoder_name, uninitialized_encoder_weights
+        )
+
         if len(uninitialized_encoder_weights) > 0:
             logger.warning(
                 f"The following encoder weights were not tied to the decoder {uninitialized_encoder_weights}"
             )
+        return tied_weights
 
     def _tie_or_clone_weights(self, output_embeddings, input_embeddings):
         """Tie or clone module weights depending of whether we are using TorchScript or not"""
@@ -2402,34 +2496,49 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
 
             # These are all the pointers of shared tensors.
             shared_ptrs = {ptr: names for ptr, names in ptrs.items() if len(names) > 1}
-            warn_names = set()
+            error_names = []
+            to_delete_names = set()
+            # Recursively descend to find tied weight keys
+            _tied_weights_keys = _get_tied_weight_keys(self)
             for names in shared_ptrs.values():
                 # Removing the keys which are declared as known duplicates on
                 # load. This allows to make sure the name which is kept is consistent.
-                if self._tied_weights_keys is not None:
+                if _tied_weights_keys is not None:
                     found = 0
                     for name in sorted(names):
-                        matches_pattern = any(re.search(pat, name) for pat in self._tied_weights_keys)
+                        matches_pattern = any(re.search(pat, name) for pat in _tied_weights_keys)
                         if matches_pattern and name in state_dict:
                             found += 1
                             if found < len(names):
-                                del state_dict[name]
+                                to_delete_names.add(name)
+            # We are entering a place where the weights and the transformers configuration do NOT match.
+            shared_names, disjoint_names = _find_disjoint(shared_ptrs.values(), state_dict)
+            # Those are actually tensor sharing but disjoint from each other, we can safely clone them
+            # Reloaded won't have the same property, but it shouldn't matter in any meaningful way.
+            for name in disjoint_names:
+                state_dict[name] = state_dict[name].clone()
 
-                # When not all duplicates have been cleaned, still remove those keys, but put a clear warning.
+            # When not all duplicates have been cleaned, still remove those keys, but put a clear warning.
-                # If the link between tensors was done at runtime then `from_pretrained` will not get
+            # If the link between tensors was done at runtime then `from_pretrained` will not get
-                # the key back leading to random tensor. A proper warning will be shown
+            # the key back leading to random tensor. A proper warning will be shown
-                # during reload (if applicable), but since the file is not necessarily compatible with
+            # during reload (if applicable), but since the file is not necessarily compatible with
-                # the config, better show a proper warning.
+            # the config, better show a proper warning.
-                found = 0
+            shared_names, identical_names = _find_identical(shared_names, state_dict)
-                for name in names:
+            # delete tensors that have identical storage
-                    if name in state_dict:
+            for inames in identical_names:
-                        found += 1
+                known = inames.intersection(to_delete_names)
-                        if found > 1:
+                for name in known:
-                            del state_dict[name]
+                    del state_dict[name]
-                            warn_names.add(name)
+                unknown = inames.difference(to_delete_names)
-            if len(warn_names) > 0:
+                if len(unknown) > 1:
-                logger.warning_once(
+                    error_names.append(unknown)
-                    f"Removed shared tensor {warn_names} while saving. This should be OK, but check by verifying that you don't receive any warning while reloading",
+
+            if shared_names:
+                error_names.append(set(shared_names))
+
+            if len(error_names) > 0:
+                raise RuntimeError(
+                    f"The weights trying to be saved contained shared tensors {error_names} that are mismatching the transformers base configuration. Try saving using `safe_serialization=False` or remove this tensor sharing.",
                 )
 
         # Shard the model if it is too big.

src/transformers/models/bert/modeling_bert.py

@@ -15,7 +15,6 @@
 # limitations under the License.
 """PyTorch BERT model."""
 
-
 import math
 import os
 import warnings
@@ -1128,7 +1127,7 @@ class BertForPreTraining(BertPreTrainedModel):
     """Bert Model with a `language modeling` head on top for CLM fine-tuning.""", BERT_START_DOCSTRING
 )
 class BertLMHeadModel(BertPreTrainedModel):
-    _tied_weights_keys = ["predictions.decoder.bias", "cls.predictions.decoder.weight"]
+    _tied_weights_keys = ["cls.predictions.decoder.bias", "cls.predictions.decoder.weight"]
 
     def __init__(self, config):
         super().__init__(config)

src/transformers/models/encoder_decoder/modeling_encoder_decoder.py

@@ -262,9 +262,16 @@ class EncoderDecoderModel(PreTrainedModel):
         if self.config.tie_encoder_decoder:
             # tie encoder and decoder base model
             decoder_base_model_prefix = self.decoder.base_model_prefix
-            self._tie_encoder_decoder_weights(
+            tied_weights = self._tie_encoder_decoder_weights(
-                self.encoder, self.decoder._modules[decoder_base_model_prefix], self.decoder.base_model_prefix
+                self.encoder,
+                self.decoder._modules[decoder_base_model_prefix],
+                self.decoder.base_model_prefix,
+                "encoder",
             )
+            # Setting a dynamic variable instead of `_tied_weights_keys` because it's a class
+            # attributed not an instance member, therefore modifying it will modify the entire class
+            # Leading to issues on subsequent calls by different tests or subsequent calls.
+            self._dynamic_tied_weights_keys = tied_weights
 
     def get_encoder(self):
         return self.encoder

src/transformers/models/marian/modeling_marian.py

@@ -1343,7 +1343,13 @@ class MarianMTModel(MarianPreTrainedModel):
         if getattr(self.config, "is_encoder_decoder", False) and getattr(self.config, "tie_encoder_decoder", False):
             if hasattr(self, self.base_model_prefix):
                 self = getattr(self, self.base_model_prefix)
-            self._tie_encoder_decoder_weights(self.encoder, self.decoder, self.base_model_prefix)
+            tied_weights = self._tie_encoder_decoder_weights(
+                self.encoder, self.decoder, self.base_model_prefix, "encoder"
+            )
+            # Setting a dynamic variable instead of `_tied_weights_keys` because it's a class
+            # attributed not an instance member, therefore modifying it will modify the entire class
+            # Leading to issues on subsequent calls by different tests or subsequent calls.
+            self._dynamic_tied_weights_keys = tied_weights
 
         for module in self.modules():
             if hasattr(module, "_tie_weights"):

src/transformers/models/musicgen/modeling_musicgen.py

@@ -1891,9 +1891,16 @@ class MusicgenForConditionalGeneration(PreTrainedModel):
         if self.config.tie_encoder_decoder:
             # tie text encoder and decoder base model
             decoder_base_model_prefix = self.decoder.base_model_prefix
-            self._tie_encoder_decoder_weights(
+            tied_weights = self._tie_encoder_decoder_weights(
-                self.text_encoder, self.decoder._modules[decoder_base_model_prefix], self.decoder.base_model_prefix
+                self.text_encoder,
+                self.decoder._modules[decoder_base_model_prefix],
+                self.decoder.base_model_prefix,
+                "text_encoder",
             )
+            # Setting a dynamic variable instead of `_tied_weights_keys` because it's a class
+            # attributed not an instance member, therefore modifying it will modify the entire class
+            # Leading to issues on subsequent calls by different tests or subsequent calls.
+            self._dynamic_tied_weights_keys = tied_weights
 
     def get_audio_encoder(self):
         return self.audio_encoder

src/transformers/models/musicgen_melody/modeling_musicgen_melody.py

@@ -1810,9 +1810,16 @@ class MusicgenMelodyForConditionalGeneration(PreTrainedModel):
         if self.config.tie_encoder_decoder:
             # tie text encoder and decoder base model
             decoder_base_model_prefix = self.decoder.base_model_prefix
-            self._tie_encoder_decoder_weights(
+            tied_weights = self._tie_encoder_decoder_weights(
-                self.text_encoder, self.decoder._modules[decoder_base_model_prefix], self.decoder.base_model_prefix
+                self.text_encoder,
+                self.decoder._modules[decoder_base_model_prefix],
+                self.decoder.base_model_prefix,
+                "text_encoder",
             )
+            # Setting a dynamic variable instead of `_tied_weights_keys` because it's a class
+            # attributed not an instance member, therefore modifying it will modify the entire class
+            # Leading to issues on subsequent calls by different tests or subsequent calls.
+            self._dynamic_tied_weights_keys = tied_weights
 
     def get_text_encoder(self):
         return self.text_encoder

tests/test_modeling_utils.py

@@ -101,7 +101,7 @@ if is_torch_available():
         _prepare_4d_attention_mask,
         _prepare_4d_causal_attention_mask,
     )
-    from transformers.modeling_utils import shard_checkpoint
+    from transformers.modeling_utils import _find_disjoint, _find_identical, shard_checkpoint
 
     # Fake pretrained models for tests
     class BaseModel(PreTrainedModel):
@@ -256,6 +256,26 @@ class ModelUtilsTest(TestCasePlus):
 
         self.assertTrue(check_models_equal(model, model_loaded))
 
+    def test_model_manually_shared_disjointed_tensors_optimum(self):
+        config = BertConfig.from_pretrained("hf-internal-testing/tiny-random-bert")
+        model = BertModel(config)
+
+        # Let's fuse qkv
+        attn = model.encoder.layer[0].attention.self
+        q = attn.query.weight
+        k = attn.key.weight
+        v = attn.value.weight
+        # Force some shared storage
+        qkv = torch.stack([q, k, v], dim=0)
+        attn.query.weight = torch.nn.Parameter(qkv[0])
+        attn.key.weight = torch.nn.Parameter(qkv[1])
+        attn.value.weight = torch.nn.Parameter(qkv[2])
+        with tempfile.TemporaryDirectory() as tmp_dir:
+            model.save_pretrained(tmp_dir)
+            model_loaded = BertModel.from_pretrained(tmp_dir)
+
+        self.assertTrue(check_models_equal(model, model_loaded))
+
     def test_model_from_pretrained_subfolder_sharded(self):
         config = BertConfig.from_pretrained("hf-internal-testing/tiny-random-bert")
         model = BertModel(config)
@@ -2222,3 +2242,40 @@ class Mask4DTestHard(unittest.TestCase):
         ]
 
         self.assertEqual(decoded_0, decoded_1b)
+
+
+@require_torch
+class TestTensorSharing(TestCasePlus):
+    def test_disjoint(self):
+        main = torch.zeros(10)
+        a = main[:5]
+        b = main[5:]
+        state_dict = {"a": a, "b": b}
+
+        shared_names, disjoint_names = _find_disjoint([{"a", "b"}], state_dict)
+        self.assertEqual(shared_names, [])
+        self.assertEqual(disjoint_names, ["a", "b"])
+
+        a = main[::2]
+        b = main[1::2]
+        state_dict = {"a": a, "b": b}
+
+        shared_names, disjoint_names = _find_disjoint([{"a", "b"}], state_dict)
+        self.assertEqual(shared_names, [{"a", "b"}])
+        self.assertEqual(disjoint_names, [])
+
+    def test_identical(self):
+        a = torch.zeros(10)
+        b = a
+        state_dict = {"a": a, "b": b}
+
+        shared_names, identical_names = _find_identical([{"a", "b"}], state_dict)
+        self.assertEqual(shared_names, [])
+        self.assertEqual(identical_names, [{"a", "b"}])
+
+        b = a[:5]
+        state_dict = {"a": a, "b": b}
+
+        shared_names, identical_names = _find_identical([{"a", "b"}], state_dict)
+        self.assertEqual(shared_names, [{"a", "b"}])
+        self.assertEqual(identical_names, [])