fix(server): revert gpt-neox optims (#123)

router/src/infer.rs

@@ -229,7 +229,11 @@ async fn batching_task(
     shared: Arc<Shared>,
 ) {
     // Minimum batch size after which we try to add more requests
-    let limit_min_batch_size = (max_batch_size / 2) as u32;
+    let limit_min_batch_size = if max_batch_size > 1 {
+        (max_batch_size / 2) as u32
+    } else {
+        0
+    };
 
     // Infinite loop
     loop {

server/Makefile

@@ -1,4 +1,4 @@
-transformers_commit := 2f87dca1ca3e5663d0637da9bb037a6956e57a5e
+transformers_commit := 517563354a3226ecfc3dca6e7a38012668d7156a
 
 gen-server:
 	# Compile protos