replaced assert with raise ValueError for t5, switch_transformers, pix2struct,...

replaced assert with raise ValueError for t5, switch_transformers, pix2struct, mt5, longt5, gptsan_japanese. (#23273)

* replaced assert with raise ValueError

* one liner

* reverse one liner and cache-decoder check

src/transformers/models/gptsan_japanese/modeling_gptsan_japanese.py

@@ -767,9 +767,10 @@ class GPTSanJapanesePreTrainedModel(PreTrainedModel):
         decoder_start_token_id = self.config.decoder_start_token_id
         pad_token_id = self.config.pad_token_id
 
-        assert decoder_start_token_id is not None, (
+        if decoder_start_token_id is None:
+            raise ValueError(
                 "self.model.config.decoder_start_token_id has to be defined. In T5 it is usually set to the pad_token_id."
-            " See T5 docs for more information"
+                "See T5 docs for more information."
             )
 
         # shift inputs to the right
@@ -782,7 +783,8 @@ class GPTSanJapanesePreTrainedModel(PreTrainedModel):
             shifted_input_ids[..., 1:] = input_ids[..., :-1].clone()
             shifted_input_ids[..., 0] = decoder_start_token_id
 
-        assert pad_token_id is not None, "self.model.config.pad_token_id has to be defined."
+        if pad_token_id is None:
+            raise ValueError("self.model.config.pad_token_id has to be defined.")
         # replace possible -100 values in labels by `pad_token_id`
         shifted_input_ids.masked_fill_(shifted_input_ids == -100, pad_token_id)
 

src/transformers/models/longt5/modeling_longt5.py

@@ -451,9 +451,10 @@ class LongT5Attention(nn.Module):
         real_seq_length = seq_length
 
         if past_key_value is not None:
-            assert (
-                len(past_key_value) == 2
-            ), f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+            if len(past_key_value) != 2:
+                raise ValueError(
+                    f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+                )
             real_seq_length += past_key_value[0].shape[2] if query_length is None else query_length
 
         key_length = real_seq_length if key_value_states is None else key_value_states.shape[1]
@@ -1349,9 +1350,10 @@ class LongT5PreTrainedModel(PreTrainedModel):
         decoder_start_token_id = self.config.decoder_start_token_id
         pad_token_id = self.config.pad_token_id
 
-        assert decoder_start_token_id is not None, (
+        if decoder_start_token_id is None:
+            raise ValueError(
                 "self.model.config.decoder_start_token_id has to be defined. In LongT5 it is usually set to the pad_token_id."
-            " See LongT5 docs for more information"
+                "See LongT5 docs for more information."
             )
 
         # shift inputs to the right
@@ -1364,7 +1366,8 @@ class LongT5PreTrainedModel(PreTrainedModel):
             shifted_input_ids[..., 1:] = input_ids[..., :-1].clone()
             shifted_input_ids[..., 0] = decoder_start_token_id
 
-        assert pad_token_id is not None, "self.model.config.pad_token_id has to be defined."
+        if pad_token_id is None:
+            raise ValueError("self.model.config.pad_token_id has to be defined.")
         # replace possible -100 values in labels by `pad_token_id`
         shifted_input_ids.masked_fill_(shifted_input_ids == -100, pad_token_id)
 

src/transformers/models/mt5/modeling_mt5.py

@@ -333,9 +333,10 @@ class MT5Attention(nn.Module):
         real_seq_length = seq_length
 
         if past_key_value is not None:
-            assert (
-                len(past_key_value) == 2
-            ), f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+            if len(past_key_value) != 2:
+                raise ValueError(
+                    f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+                )
             real_seq_length += past_key_value[0].shape[2] if query_length is None else query_length
 
         key_length = real_seq_length if key_value_states is None else key_value_states.shape[1]
@@ -818,9 +819,10 @@ class MT5PreTrainedModel(PreTrainedModel):
         decoder_start_token_id = self.config.decoder_start_token_id
         pad_token_id = self.config.pad_token_id
 
-        assert decoder_start_token_id is not None, (
+        if decoder_start_token_id is None:
+            raise ValueError(
                 "self.model.config.decoder_start_token_id has to be defined. In MT5 it is usually set to the pad_token_id."
-            " See MT5 docs for more information"
+                "See MT5 docs for more information."
             )
 
         # shift inputs to the right
@@ -833,7 +835,8 @@ class MT5PreTrainedModel(PreTrainedModel):
             shifted_input_ids[..., 1:] = input_ids[..., :-1].clone()
             shifted_input_ids[..., 0] = decoder_start_token_id
 
-        assert pad_token_id is not None, "self.model.config.pad_token_id has to be defined."
+        if pad_token_id is None:
+            raise ValueError("self.model.config.pad_token_id has to be defined.")
         # replace possible -100 values in labels by `pad_token_id`
         shifted_input_ids.masked_fill_(shifted_input_ids == -100, pad_token_id)
 
@@ -952,7 +955,8 @@ class MT5Stack(MT5PreTrainedModel):
             raise ValueError(f"You have to specify either {err_msg_prefix}input_ids or {err_msg_prefix}inputs_embeds")
 
         if inputs_embeds is None:
-            assert self.embed_tokens is not None, "You have to initialize the model with valid token embeddings"
+            if self.embed_tokens is None:
+                raise ValueError("You have to initialize the model with valid token embeddings")
             inputs_embeds = self.embed_tokens(input_ids)
 
         batch_size, seq_length = input_shape
@@ -961,7 +965,8 @@ class MT5Stack(MT5PreTrainedModel):
         mask_seq_length = past_key_values[0][0].shape[2] + seq_length if past_key_values is not None else seq_length
 
         if use_cache is True:
-            assert self.is_decoder, f"`use_cache` can only be set to `True` if {self} is used as a decoder"
+            if not self.is_decoder:
+                raise ValueError(f"`use_cache` can only be set to `True` if {self} is used as a decoder")
 
         if attention_mask is None:
             attention_mask = torch.ones(batch_size, mask_seq_length, device=inputs_embeds.device)
@@ -1852,8 +1857,14 @@ class MT5ForConditionalGeneration(MT5PreTrainedModel):
                     layer_past_state.index_select(0, beam_idx.to(layer_past_state.device)),
                 )
 
-            assert reordered_layer_past_states[0].shape == layer_past_states[0].shape
-            assert len(reordered_layer_past_states) == len(layer_past_states)
+            if reordered_layer_past_states[0].shape != layer_past_states[0].shape:
+                raise ValueError(
+                    f"reordered_layer_past_states[0] shape {reordered_layer_past_states[0].shape} and layer_past_states[0] shape {layer_past_states[0].shape} mismatched"
+                )
+            if len(reordered_layer_past_states) != len(layer_past_states):
+                raise ValueError(
+                    f"length of reordered_layer_past_states {len(reordered_layer_past_states)} and length of layer_past_states {len(layer_past_states)} mismatched"
+                )
 
             reordered_decoder_past = reordered_decoder_past + (reordered_layer_past_states,)
         return reordered_decoder_past

src/transformers/models/pix2struct/modeling_pix2struct.py

@@ -479,9 +479,10 @@ class Pix2StructPreTrainedModel(PreTrainedModel):
         decoder_start_token_id = self.config.decoder_start_token_id
         pad_token_id = self.config.pad_token_id
 
-        assert decoder_start_token_id is not None, (
+        if decoder_start_token_id is None:
+            raise ValueError(
                 "self.model.config.decoder_start_token_id has to be defined. In Pix2Struct it is usually set to the pad_token_id."
-            " See Pix2Struct docs for more information"
+                "See Pix2Struct docs for more information."
             )
 
         # shift inputs to the right
@@ -494,7 +495,8 @@ class Pix2StructPreTrainedModel(PreTrainedModel):
             shifted_input_ids[..., 1:] = input_ids[..., :-1].clone()
             shifted_input_ids[..., 0] = decoder_start_token_id
 
-        assert pad_token_id is not None, "self.model.config.pad_token_id has to be defined."
+        if pad_token_id is None:
+            raise ValueError("self.model.config.pad_token_id has to be defined.")
         # replace possible -100 values in labels by `pad_token_id`
         shifted_input_ids.masked_fill_(shifted_input_ids == -100, pad_token_id)
 
@@ -1356,8 +1358,14 @@ class Pix2StructTextModel(Pix2StructPreTrainedModel):
                     layer_past_state.index_select(0, beam_idx.to(layer_past_state.device)),
                 )
 
-            assert reordered_layer_past_states[0].shape == layer_past_states[0].shape
-            assert len(reordered_layer_past_states) == len(layer_past_states)
+            if reordered_layer_past_states[0].shape != layer_past_states[0].shape:
+                raise ValueError(
+                    f"reordered_layer_past_states[0] shape {reordered_layer_past_states[0].shape} and layer_past_states[0] shape {layer_past_states[0].shape} mismatched"
+                )
+            if len(reordered_layer_past_states) != len(layer_past_states):
+                raise ValueError(
+                    f"length of reordered_layer_past_states {len(reordered_layer_past_states)} and length of layer_past_states {len(layer_past_states)} mismatched"
+                )
 
             reordered_decoder_past = reordered_decoder_past + (reordered_layer_past_states,)
         return reordered_decoder_past

src/transformers/models/switch_transformers/modeling_switch_transformers.py

@@ -515,9 +515,10 @@ class SwitchTransformersAttention(nn.Module):
         real_seq_length = seq_length
 
         if past_key_value is not None:
-            assert (
-                len(past_key_value) == 2
-            ), f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+            if len(past_key_value) != 2:
+                raise ValueError(
+                    f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+                )
             real_seq_length += past_key_value[0].shape[2] if query_length is None else query_length
 
         key_length = real_seq_length if key_value_states is None else key_value_states.shape[1]

src/transformers/models/t5/modeling_t5.py

@@ -163,9 +163,8 @@ def load_tf_weights_in_t5(model, config, tf_checkpoint_path):
             logger.info(f"Transposing numpy weight of shape {array.shape} for {name}")
             array = np.transpose(array)
         try:
-            assert (
-                pointer.shape == array.shape
-            ), f"Pointer shape {pointer.shape} and array shape {array.shape} mismatched"
+            if pointer.shape != array.shape:
+                raise ValueError(f"Pointer shape {pointer.shape} and array shape {array.shape} mismatched")
         except AssertionError as e:
             e.args += (pointer.shape, array.shape)
             raise
@@ -473,9 +472,10 @@ class T5Attention(nn.Module):
         real_seq_length = seq_length
 
         if past_key_value is not None:
-            assert (
-                len(past_key_value) == 2
-            ), f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+            if len(past_key_value) != 2:
+                raise ValueError(
+                    f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
+                )
             real_seq_length += past_key_value[0].shape[2] if query_length is None else query_length
 
         key_length = real_seq_length if key_value_states is None else key_value_states.shape[1]
@@ -848,9 +848,10 @@ class T5PreTrainedModel(PreTrainedModel):
         decoder_start_token_id = self.config.decoder_start_token_id
         pad_token_id = self.config.pad_token_id
 
-        assert decoder_start_token_id is not None, (
+        if decoder_start_token_id is None:
+            raise ValueError(
                 "self.model.config.decoder_start_token_id has to be defined. In T5 it is usually set to the pad_token_id."
-            " See T5 docs for more information"
+                "See T5 docs for more information."
             )
 
         # shift inputs to the right
@@ -863,7 +864,8 @@ class T5PreTrainedModel(PreTrainedModel):
             shifted_input_ids[..., 1:] = input_ids[..., :-1].clone()
             shifted_input_ids[..., 0] = decoder_start_token_id
 
-        assert pad_token_id is not None, "self.model.config.pad_token_id has to be defined."
+        if pad_token_id is None:
+            raise ValueError("self.model.config.pad_token_id has to be defined.")
         # replace possible -100 values in labels by `pad_token_id`
         shifted_input_ids.masked_fill_(shifted_input_ids == -100, pad_token_id)
 
@@ -981,7 +983,8 @@ class T5Stack(T5PreTrainedModel):
             raise ValueError(f"You have to specify either {err_msg_prefix}input_ids or {err_msg_prefix}inputs_embeds")
 
         if inputs_embeds is None:
-            assert self.embed_tokens is not None, "You have to initialize the model with valid token embeddings"
+            if self.embed_tokens is None:
+                raise ValueError("You have to initialize the model with valid token embeddings")
             inputs_embeds = self.embed_tokens(input_ids)
 
         batch_size, seq_length = input_shape
@@ -990,7 +993,8 @@ class T5Stack(T5PreTrainedModel):
         mask_seq_length = past_key_values[0][0].shape[2] + seq_length if past_key_values is not None else seq_length
 
         if use_cache is True:
-            assert self.is_decoder, f"`use_cache` can only be set to `True` if {self} is used as a decoder"
+            if not self.is_decoder:
+                raise ValueError(f"`use_cache` can only be set to `True` if {self} is used as a decoder")
 
         if attention_mask is None:
             attention_mask = torch.ones(batch_size, mask_seq_length, device=inputs_embeds.device)
@@ -1817,8 +1821,14 @@ class T5ForConditionalGeneration(T5PreTrainedModel):
                     layer_past_state.index_select(0, beam_idx.to(layer_past_state.device)),
                 )
 
-            assert reordered_layer_past_states[0].shape == layer_past_states[0].shape
-            assert len(reordered_layer_past_states) == len(layer_past_states)
+            if reordered_layer_past_states[0].shape != layer_past_states[0].shape:
+                raise ValueError(
+                    f"reordered_layer_past_states[0] shape {reordered_layer_past_states[0].shape} and layer_past_states[0] shape {layer_past_states[0].shape} mismatched"
+                )
+            if len(reordered_layer_past_states) != len(layer_past_states):
+                raise ValueError(
+                    f"length of reordered_layer_past_states {len(reordered_layer_past_states)} and length of layer_past_states {len(layer_past_states)} mismatched"
+                )
 
             reordered_decoder_past = reordered_decoder_past + (reordered_layer_past_states,)
         return reordered_decoder_past