fix: torch.int32 instead of torch.torch.int32 (#28883)

src/transformers/models/bark/modeling_bark.py

@@ -75,7 +75,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/bart/modeling_bart.py

@@ -89,7 +89,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/distilbert/modeling_distilbert.py

@@ -82,7 +82,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/falcon/modeling_falcon.py

@@ -122,7 +122,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/gpt_bigcode/modeling_gpt_bigcode.py

@@ -92,7 +92,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/gpt_neo/modeling_gpt_neo.py

@@ -80,7 +80,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/gpt_neox/modeling_gpt_neox.py

@@ -63,7 +63,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/llama/modeling_llama.py

@@ -63,7 +63,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/mbart/modeling_mbart.py

@@ -72,7 +72,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/mistral/modeling_mistral.py

@@ -62,7 +62,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/mixtral/modeling_mixtral.py

@@ -155,7 +155,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/opt/modeling_opt.py

@@ -77,7 +77,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/phi/modeling_phi.py

@@ -70,7 +70,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/qwen2/modeling_qwen2.py

@@ -69,7 +69,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,

src/transformers/models/whisper/modeling_whisper.py

@@ -70,7 +70,7 @@ def _get_unpad_data(attention_mask):
     seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
     indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
     max_seqlen_in_batch = seqlens_in_batch.max().item()
-    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
     return (
         indices,
         cu_seqlens,