check paddle version for v1 loader (#4473)

fastdeploy/model_executor/utils.py

@@ -205,6 +205,22 @@ def is_pre_sliced_weight(model_path):
     return len(rank_dirs) > 1
 
 
+def is_paddle_support_v1_loader():
+    src_shape = [32, 32]
+    tgt_shape = [1, 32, 64]
+    src_tensor = paddle.ones(src_shape, dtype="float32")
+    tgt_tensor = paddle.zeros(tgt_shape, dtype="float32")
+    for exp_id in range(tgt_shape[0]):
+        # gate
+        gate_tgt = tgt_tensor[exp_id][..., : tgt_shape[2] // 2]
+        gate_tgt.copy_(src_tensor, False)
+        # up
+        up_tgt = tgt_tensor[exp_id][..., tgt_shape[2] // 2 :]
+        up_tgt.copy_(src_tensor, False)
+    is_same = bool(paddle.all(tgt_tensor == 1))
+    return is_same
+
+
 def v1_loader_support(fd_config):
     _v1_no_support_archs = ["Qwen2VLForConditionalGeneration", "Qwen2_5_VLForConditionalGeneration"]
 
@@ -242,6 +258,10 @@ def v1_loader_support(fd_config):
     if fd_config.model_config.architectures[0] in _v1_no_support_archs:
         _err_msg(f"v1 loader currently does not support {fd_config.model_config.architectures[0]}")
         return False
+
+    if not is_paddle_support_v1_loader():
+        _err_msg("The installed Paddle does not support v1 loader")
+        return False
     return True