fix get_keys_to_not_convert() to return correct modules for full precision inference (#25105)

* add test for `get_keys_to_not_convert`

* add minimum patch to keep mpt lm_head from 8bit quantization

* add reivsion to

src/transformers/utils/bitsandbytes.py

@@ -265,17 +265,16 @@ def get_keys_to_not_convert(model):
         tied_keys = sum(tied_params, [])
     has_tied_params = len(tied_keys) > 0
 
-    # Check if it is a base model
-    is_base_model = not hasattr(model, model.base_model_prefix)
+    # If there is not tied weights, we want to keep the lm_head（output_embedding) in full precision
+    if not has_tied_params:
+        output_emb = model.get_output_embeddings()
+        if output_emb is not None:
+            list_last_module = [name for name, module in model.named_modules() if id(module) == id(output_emb)]
+            return list_last_module
 
-    # Ignore this for base models (BertModel, GPT2Model, etc.)
-    if (not has_tied_params) and is_base_model:
-        return []
-
-    # otherwise they have an attached head
+    # otherwise, no tied weights, no output embedding defined, simply keep the last module in full precision
     list_modules = list(model.named_parameters())
     list_last_module = [list_modules[-1][0]]
-
     # add last module together with tied weights
     intersection = set(list_last_module) - set(tied_keys)
     list_untouched = list(set(tied_keys)) + list(intersection)

tests/bnb/test_mixed_int8.py

@@ -124,6 +124,53 @@ class MixedInt8Test(BaseMixedInt8Test):
         gc.collect()
         torch.cuda.empty_cache()
 
+    def test_get_keys_to_not_convert(self):
+        r"""
+        Test the `get_keys_to_not_convert` function.
+        """
+        from accelerate import init_empty_weights
+
+        from transformers import AutoModelForMaskedLM, Blip2ForConditionalGeneration, MptForCausalLM, OPTForCausalLM
+        from transformers.utils.bitsandbytes import get_keys_to_not_convert
+
+        model_id = "mosaicml/mpt-7b"
+        config = AutoConfig.from_pretrained(
+            model_id, trust_remote_code=True, revision="72e5f594ce36f9cabfa2a9fd8f58b491eb467ee7"
+        )
+        with init_empty_weights():
+            model = AutoModelForCausalLM.from_config(config, trust_remote_code=True)
+        self.assertEqual(get_keys_to_not_convert(model), ["transformer.wte"])
+        # without trust_remote_code
+        config = AutoConfig.from_pretrained(model_id, revision="72e5f594ce36f9cabfa2a9fd8f58b491eb467ee7")
+        with init_empty_weights():
+            model = MptForCausalLM(config)
+        # The order of the keys does not matter, so we sort them before comparing, same for the other tests.
+        self.assertEqual(get_keys_to_not_convert(model).sort(), ["lm_head", "transformer.wte"].sort())
+
+        model_id = "Salesforce/blip2-opt-2.7b"
+        config = AutoConfig.from_pretrained(model_id, revision="1ef7f63a8f0a144c13fdca8103eb7b4691c74cec")
+        with init_empty_weights():
+            model = Blip2ForConditionalGeneration(config)
+        self.assertEqual(
+            get_keys_to_not_convert(model).sort(),
+            ["language_model.lm_head", "language_model.model.decoder.embed_tokens"].sort(),
+        )
+
+        model_id = "facebook/opt-350m"
+        config = AutoConfig.from_pretrained(model_id, revision="cb32f77e905cccbca1d970436fb0f5e6b58ee3c5")
+        with init_empty_weights():
+            model = OPTForCausalLM(config)
+        self.assertEqual(get_keys_to_not_convert(model).sort(), ["lm_head", "model.decoder.embed_tokens"].sort())
+
+        model_id = "roberta-large"
+        config = AutoConfig.from_pretrained(model_id, revision="716877d372b884cad6d419d828bac6c85b3b18d9")
+        with init_empty_weights():
+            model = AutoModelForMaskedLM.from_config(config)
+        self.assertEqual(
+            get_keys_to_not_convert(model).sort(),
+            ["'roberta.embeddings.word_embeddings', 'lm_head', 'lm_head.decoder"].sort(),
+        )
+
     def test_quantization_config_json_serialization(self):
         r"""
         A simple test to check if the quantization config is correctly serialized and deserialized