Use dynamic CUDA check instead of hardcoded device in emb_extractor.py

Fixes CPU-only inference, which currently fails with "AssertionError: Torch not compiled with CUDA enabled" because device="cuda" is hardcoded.

Replaces the hardcoded "cuda" with a dynamic check ("cuda" if torch.cuda.is_available() else "cpu"), matching the pattern already used elsewhere in the codebase (e.g. perturber_utils.py line 194).

See discussion #592: https://huggingface.co/ctheodoris/Geneformer/discussions/592

geneformer/emb_extractor.py

@@ -98,7 +98,7 @@ def get_embs(
         minibatch = filtered_input_data.select([i for i in range(i, max_range)])
 
         max_len = int(max(minibatch["length"]))
-        original_lens = torch.tensor(minibatch["length"], device="cuda")
+        original_lens = torch.tensor(minibatch["length"], device="cuda" if torch.cuda.is_available() else "cpu")
         minibatch.set_format(type="torch")
 
         input_data_minibatch = minibatch["input_ids"]
@@ -108,7 +108,7 @@ def get_embs(
 
         with torch.no_grad():
             outputs = model(
-                input_ids=input_data_minibatch.to("cuda"),
+                original_lens = torch.tensor(minibatch["length"], device="cuda" if torch.cuda.is_available() else "cpu")
                 attention_mask=pu.gen_attention_mask(minibatch),
             )