【pir 】modify dy2static Sd and 3. Grounding DINO model

paddlemix/models/groundingdino/fuse_modules.py

@@ -15,6 +15,7 @@
 import paddle
 import paddle.nn as nn
 import paddle.nn.functional as F
+from paddle.framework import in_dynamic_mode
 from paddle.nn.initializer import Constant
 from paddlenlp.utils.initializer import constant_, xavier_uniform_
 
@@ -183,7 +184,7 @@ def forward(self, v, l, attention_mask_v=None, attention_mask_l=None):
         src_len = key_states.shape[1]
         attn_weights = paddle.bmm(query_states, key_states.transpose([0, 2, 1]))  # bs*nhead, nimg, ntxt
 
-        if attn_weights.shape != [bsz * self.num_heads, tgt_len, src_len]:
+        if in_dynamic_mode() and attn_weights.shape != [bsz * self.num_heads, tgt_len, src_len]:
             raise ValueError(
                 f"Attention weights should be of size {(bsz * self.num_heads, tgt_len, src_len)}, but is {attn_weights.shape}"
             )
@@ -236,12 +237,12 @@ def forward(self, v, l, attention_mask_v=None, attention_mask_l=None):
         attn_output_v = paddle.bmm(attn_probs_v, value_l_states)
         attn_output_l = paddle.bmm(attn_probs_l, value_v_states)
 
-        if attn_output_v.shape != [bsz * self.num_heads, tgt_len, self.head_dim]:
+        if in_dynamic_mode() and attn_output_v.shape != [bsz * self.num_heads, tgt_len, self.head_dim]:
             raise ValueError(
                 f"`attn_output_v` should be of size {(bsz, self.num_heads, tgt_len, self.head_dim)}, but is {attn_output_v.shape}"
             )
 
-        if attn_output_l.shape != [bsz * self.num_heads, src_len, self.head_dim]:
+        if in_dynamic_mode() and attn_output_l.shape != [bsz * self.num_heads, src_len, self.head_dim]:
             raise ValueError(
                 f"`attn_output_l` should be of size {(bsz, self.num_heads, src_len, self.head_dim)}, but is {attn_output_l.shape}"
             )

ppdiffusers/ppdiffusers/transformers/clip/modeling.py

@@ -20,6 +20,7 @@
 import paddle
 import paddle.nn.functional as F
 from paddle import nn
+from paddle.framework import in_dynamic_mode
 from paddlenlp.transformers.activations import ACT2FN
 from paddlenlp.transformers.model_outputs import (
     BaseModelOutput,
@@ -279,15 +280,15 @@ def forward(
         src_len = key_states.shape[1]
         attn_weights = paddle.matmul(query_states, key_states, transpose_y=True)
 
-        if attn_weights.shape != [bsz * self.num_heads, tgt_len, src_len]:
+        if in_dynamic_mode() and attn_weights.shape != [bsz * self.num_heads, tgt_len, src_len]:
             raise ValueError(
                 f"Attention weights should be of size {[bsz * self.num_heads, tgt_len, src_len]}, but is"
                 f" {attn_weights.shape}"
             )
 
         # apply the causal_attention_mask first
         if causal_attention_mask is not None:
-            if causal_attention_mask.shape != [bsz, 1, tgt_len, src_len]:
+            if in_dynamic_mode() and causal_attention_mask.shape != [bsz, 1, tgt_len, src_len]:
                 raise ValueError(
                     f"Attention mask should be of size {[bsz, 1, tgt_len, src_len]}, but is"
                     f" {causal_attention_mask.shape}"
@@ -296,7 +297,7 @@ def forward(
             attn_weights = attn_weights.reshape([bsz * self.num_heads, tgt_len, src_len])
 
         if attention_mask is not None:
-            if attention_mask.shape != [bsz, 1, tgt_len, src_len]:
+            if in_dynamic_mode() and attention_mask.shape != [bsz, 1, tgt_len, src_len]:
                 raise ValueError(
                     f"Attention mask should be of size {[bsz, 1, tgt_len, src_len]}, but is {attention_mask.shape}"
                 )
@@ -319,7 +320,7 @@ def forward(
 
         attn_output = paddle.matmul(attn_probs, value_states)
 
-        if attn_output.shape != [bsz * self.num_heads, tgt_len, self.head_dim]:
+        if in_dynamic_mode() and attn_output.shape != [bsz * self.num_heads, tgt_len, self.head_dim]:
             raise ValueError(
                 f"`attn_output` should be of size {[bsz, self.num_heads, tgt_len, self.head_dim]}, but is"
                 f" {attn_output.shape}"