fix qwen2_vl in orange pi

mindnlp/core/__init__.py

@@ -53,6 +53,7 @@
 from ._bind import get_default_dtype, set_default_dtype, get_default_device
 from .amp import autocast, GradScaler
 from .func import vmap
+from .configs import set_pyboost
 
 from . import profiler, cuda, optim, amp, compiler, jit, version, __future__, overrides, \
     return_types, linalg, fx, backends, testing, nn, fft, _jit_internal, utils

mindnlp/core/nn/functional.py

@@ -849,7 +849,7 @@ def conv3d(input, weight, bias=None, stride=1, padding=0, dilation=1, groups=1):
                             mode=1,
                             pad_mode=pad_mode,
                             pad=pad,
-                            stride=stride,
+                            stride=tuple(stride),
                             dilation=dilation,
                             group=groups)
     output = conv3d_op(input, weight)

mindnlp/core/ops/comparison.py

@@ -183,7 +183,7 @@ def sort(input, *, dim=-1, descending=False, stable=False):
 # topk
 has_topk = hasattr(mindspore.mint, 'topk')
 def topk(input, k, dim=-1, largest=True, sorted=True):
-    if use_pyboost() and has_topk:
+    if use_pyboost() and has_topk and not ON_ORANGE_PI:
         out = mindspore.mint.topk(input, int(k), dim, largest, sorted)
     else:
         out = ops.topk(input, k, dim, largest, sorted)

mindnlp/core/ops/pointwise.py

@@ -334,9 +334,9 @@ def erfc(input, *, out=None):
 
 
 def erfinv(input, *, out=None):
+    if ON_ORANGE_PI:
+        return erfinv_torch(input)
     if use_pyboost() and has_erfinv:
-        if ON_ORANGE_PI:
-            return erfinv_torch(input)
         return call_ms_func(mindspore.mint.erfinv, input, out=out)
     return call_ms_func(ops.erfinv, input, out=out)