Upload MOJO

README.md

@@ -1,12 +1,12 @@
 ---
 library_name: transformers
 tags:
-  - bulk RNA-seq
-  - DNA methylation
-  - biology
-  - transcriptomics
-  - epigenomics
-  - multimodal
+- bulk RNA-seq
+- DNA methylation
+- biology
+- transcriptomics
+- epigenomics
+- multimodal
 ---
 
 # MOJO

config.json

@@ -39,6 +39,6 @@
   "stem_kernel_shape": 15,
   "token_embed_dim": 256,
   "torch_dtype": "float32",
-  "transformers_version": "4.37.2",
+  "transformers_version": "4.51.0",
   "use_gene_embedding": true
 }

mojo.py

@@ -78,14 +78,16 @@ class RotaryEmbedding(torch.nn.Module):
     ) -> Tuple[torch.Tensor, torch.Tensor]:
         if self.rescaling_factor is None:
             inv_freq = 1.0 / (
-                self.upper_freq ** (torch.arange(0, self.dim, 2).float() / self.dim)
+                self.upper_freq
+                ** (torch.arange(0, self.dim, 2, device=q.device).float() / self.dim)
             )
         else:
             updated_base = self.upper_freq * (
                 self.rescaling_factor ** (self.dim / (self.dim - 2))
             )
             inv_freq = 1.0 / (
-                updated_base ** (torch.arange(0, self.dim, 2).float() / self.dim)
+                updated_base
+                ** (torch.arange(0, self.dim, 2, device=q.device).float() / self.dim)
             )
 
         self._cos_cached, self._sin_cached = self._compute_cos_sin_tables(