Info logs displayed as ERROR | stderr

Currently, a lot of logs which are clearly only info-level are displayed as ERROR | stderr, which probably shouldn't be the case.
Two examples:

$ python3 -m fastchat.serve.controller
2023-05-17 10:56:57 | INFO | controller | args: Namespace(host='localhost', port=21001, dispatch_method='shortest_queue')
2023-05-17 10:56:57 | INFO | controller | Init controller
2023-05-17 10:56:57 | ERROR | stderr | INFO:     Started server process [19332]
2023-05-17 10:56:57 | ERROR | stderr | INFO:     Waiting for application startup.
2023-05-17 10:56:57 | ERROR | stderr | INFO:     Application startup complete.
2023-05-17 10:56:57 | ERROR | stderr | INFO:     Uvicorn running on http://localhost:21001 (Press CTRL+C to quit)

> python3 -m fastchat.serve.model_worker --model-name 'vicuna-13b-v1.1' --model-path /path/to/vicuna
2023-05-17 10:57:44 | INFO | model_worker | args: Namespace(host='localhost', port=21002, worker_address='http://localhost:21002', controller_address='http://localhost:21001', model_path='/path/to/vicuna', device='cuda', gpus=None, num_gpus=1, max_gpu_memory=None, load_8bit=False, cpu_offloading=False, model_name='vicuna-13b-v1.1', limit_model_concurrency=5, stream_interval=2, no_register=False)
2023-05-17 10:57:44 | INFO | model_worker | Loading the model vicuna-13b-v1.1 on worker c7ee88 ...                                                                                                         
Downloading (…)okenizer_config.json:   0%|                                                                                                                                       | 0.00/727 [00:00<?, ?B/s]
Downloading (…)okenizer_config.json: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 727/727 [00:00<00:00, 199kB/s]
2023-05-17 10:57:44 | ERROR | stderr |                                                                                                                                                                     
Downloading tokenizer.model:   0%|                                                                                                                                              | 0.00/500k [00:00<?, ?B/s]
Downloading tokenizer.model: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 500k/500k [00:00<00:00, 7.57MB/s]
2023-05-17 10:57:44 | ERROR | stderr |                                                                                                                                                                     
Downloading (…)cial_tokens_map.json:   0%|                                                                                                                                       | 0.00/411 [00:00<?, ?B/s]
Downloading (…)cial_tokens_map.json: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 411/411 [00:00<00:00, 594kB/s]      

Start with the fastchat.serve.controller and fastchat.serve.model_worker entry points and reproduce the shown output. Trace how Uvicorn and model-download messages from stderr are captured and classified, then verify that informational messages are no longer labeled ERROR while genuine errors retain their severity.