Revert "Model: Skip empty token chunks"

This reverts commit 21516bd7b5.

This skips EOS and implementing it the proper way seems more
costly than necessary.

Signed-off-by: kingbri <bdashore3@proton.me>

backends/exllamav2/model.py

@@ -1185,15 +1185,13 @@ class ExllamaV2Container:
                 result_id = result.get("identifier")
 
                 if stage == "streaming" and result_id == job_id:
-                    chunk_tokens = result.get("token_ids")
-                    if chunk_tokens is None:
-                        continue
-                    else:
-                        generated_tokens += chunk_tokens.size(dim=0)
-
                     chunk = unwrap(result.get("text"), "")
                     full_response += chunk
 
+                    chunk_tokens = result.get("token_ids")
+                    if chunk_tokens is not None:
+                        generated_tokens += chunk_tokens.size(dim=0)
+
                     generation = {
                         "text": chunk,
                         "prompt_tokens": context_len,