Fix temb typecasting in GMQwen (#6)

2025-11-09 16:10:34 -08:00
parent ab3787647a
commit 9dff3fd21e
1 changed files with 1 additions and 1 deletions
@@ -174,7 +174,7 @@ class GMQwenImageTransformer2DModel(QwenImageTransformer2DModel):
                bs, k, 1, 1, h * self.patch_size, w * self.patch_size
            )[..., :x.shape[-2], :x.shape[-1]].log_softmax(dim=1)
        if self.constant_logstd is None:
-            out_logstds = self.proj_out_logstds(temb.detach()).reshape(bs, 1, 1, 1, 1, 1)
+            out_logstds = self.proj_out_logstds(temb.detach().float()).reshape(bs, 1, 1, 1, 1, 1)
        else:
            out_logstds = hidden_states.new_full((bs, 1, 1, 1, 1, 1), float(self.constant_logstd))