Fix importing unofficial TF models with extra optimizer weights

Author: monologg

src/transformers/modeling_albert.py

@@ -117,7 +117,13 @@ def load_tf_weights_in_albert(model, config, tf_checkpoint_path):
         name = name.split("/")
 
         # Ignore the gradients applied by the LAMB/ADAM optimizers.
-        if "adam_m" in name or "adam_v" in name or "global_step" in name:
+        if (
+            "adam_m" in name
+            or "adam_v" in name
+            or "AdamWeightDecayOptimizer" in name
+            or "AdamWeightDecayOptimizer_1" in name
+            or "global_step" in name
+        ):
             logger.info("Skipping {}".format("/".join(name)))
             continue
 

src/transformers/modeling_bert.py

@@ -86,7 +86,10 @@ def load_tf_weights_in_bert(model, config, tf_checkpoint_path):
         name = name.split("/")
         # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v
         # which are not required for using pretrained model
-        if any(n in ["adam_v", "adam_m", "global_step"] for n in name):
+        if any(
+            n in ["adam_v", "adam_m", "AdamWeightDecayOptimizer", "AdamWeightDecayOptimizer_1", "global_step"]
+            for n in name
+        ):
             logger.info("Skipping {}".format("/".join(name)))
             continue
         pointer = model

src/transformers/modeling_t5.py

@@ -79,7 +79,10 @@ def load_tf_weights_in_t5(model, config, tf_checkpoint_path):
         name = txt_name.split("/")
         # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v
         # which are not required for using pretrained model
-        if any(n in ["adam_v", "adam_m", "global_step"] for n in name):
+        if any(
+            n in ["adam_v", "adam_m", "AdamWeightDecayOptimizer", "AdamWeightDecayOptimizer_1", "global_step"]
+            for n in name
+        ):
             logger.info("Skipping {}".format("/".join(name)))
             tf_weights.pop(txt_name, None)
             continue

templates/adding_a_new_model/modeling_xxx.py

@@ -76,7 +76,10 @@ def load_tf_weights_in_xxx(model, config, tf_checkpoint_path):
         name = name.split("/")
         # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v
         # which are not required for using pretrained model
-        if any(n in ["adam_v", "adam_m", "global_step"] for n in name):
+        if any(
+            n in ["adam_v", "adam_m", "AdamWeightDecayOptimizer", "AdamWeightDecayOptimizer_1", "global_step"]
+            for n in name
+        ):
             logger.info("Skipping {}".format("/".join(name)))
             continue
         pointer = model