Fix load_log_files UnboundLocalError when all FileNotFoundError retries fail

[Chessing234]

Bug

fastchat/serve/monitor/basic_stats.py::load_log_files opens the log file inside a 5-attempt retry loop and then iterates lines:

def load_log_files(filename):
    data = []
    for retry in range(5):
        try:
            lines = open(filename).readlines()
            break
        except FileNotFoundError:
            time.sleep(2)

    for l in lines:
        ...

If the file never becomes available (wrong path, producer hasn't written it yet, stale watch), every iteration hits the except branch. The loop completes without executing lines = ..., so the following for l in lines: raises UnboundLocalError: local variable 'lines' referenced before assignment. That error buries the real cause — FileNotFoundError — and points at a line where nothing is actually wrong.

load_log_files_parallel calls this through a Pool.imap, so on a missing shard the traceback surfaced to the caller is UnboundLocalError with no filename, instead of a clear "this file is missing."

Root cause

The retry loop swallows FileNotFoundError but has no terminal branch for "all retries exhausted." Python doesn't care that the symbol is annotated in the try; the binding only happens on a successful open().

Why the fix is correct

Attach a for/else whose else only runs when the loop wasn't broken out of — i.e. when every retry failed — and re-raise the real exception with the filename:

     for retry in range(5):
         try:
             lines = open(filename).readlines()
             break
         except FileNotFoundError:
             time.sleep(2)
+    else:
+        raise FileNotFoundError(f"Failed to open {filename} after 5 retries")

     for l in lines:

• Happy path: break skips the else, behaviour is unchanged.
• Failure path: the UnboundLocalError is replaced by FileNotFoundError naming the actual file, matching the exception that the loop is explicitly retrying.

Two added lines; no other changes in this function or file.