fix demo (add gradio requirement and fix download)

app.py

@@ -17,7 +17,7 @@ args = parser.parse_args()
 os.environ["RWKV_JIT_ON"] = '1'
 
 from rwkv.model import RWKV
-model_path = hf_hub_download(repo_id="Synthia/ChatGalRWKV/", filename="rwkv-chatgal-v1-3B-ctx4096-epoch2.pth")
+model_path = hf_hub_download(repo_id="Synthia/ChatGalRWKV", filename="rwkv-chatgal-v1-3B-ctx4096-epoch2.pth")
 if os.environ['ON_COLAB'] == '1':
     os.environ["RWKV_CUDA_ON"] = '1' # if '1' then use CUDA kernel for seq mode (much faster)
     model = RWKV(model=model_path, strategy='cuda bf16')

requirements.txt

@@ -4,4 +4,5 @@ tokenizers
 torch
 rwkv
 pynvml
-huggingface_hub
+huggingface_hub
+gradio>=3.17.1