[`InstructBlip`] Fix int8/fp4 issues (#24888)

* fix dtype issue

* revert `.float()`

* fix copies

src/transformers/models/instructblip/modeling_instructblip.py

@@ -558,7 +558,6 @@ class InstructBlipVisionModel(InstructBlipPreTrainedModel):
         return self.embeddings
 
 
-# Copied from transformers.models.blip_2.modeling_blip_2.Blip2QFormerMultiHeadAttention with Blip2->InstructBlip
 class InstructBlipQFormerMultiHeadAttention(nn.Module):
     def __init__(self, config, is_cross_attention=False):
         super().__init__()
@@ -659,13 +658,14 @@ class InstructBlipQFormerMultiHeadAttention(nn.Module):
                 attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key
 
         attention_scores = attention_scores / math.sqrt(self.attention_head_size)
+        attention_scores_dtype = attention_scores.dtype
 
         if attention_mask is not None:
             # Apply the attention mask is (precomputed for all layers in BertModel forward() function)
             attention_scores = attention_scores + attention_mask
 
         # Normalize the attention scores to probabilities.
-        attention_probs = nn.Softmax(dim=-1)(attention_scores)
+        attention_probs = nn.Softmax(dim=-1)(attention_scores).to(attention_scores_dtype)
 
         if is_cross_attention and self.save_attention:
             self.save_attention_map(attention_probs)
@@ -1038,6 +1038,7 @@ class InstructBlipQFormerEmbeddings(nn.Module):
         else:
             embeddings = query_embeds
 
+        embeddings = embeddings.to(self.layernorm.weight.dtype)
         embeddings = self.layernorm(embeddings)
         embeddings = self.dropout(embeddings)
         return embeddings