Update app.py

app.py

@@ -19,12 +19,12 @@ gpu_h = nvmlDeviceGetHandleByIndex(0)
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
 ctx_limit = 4000
-gen_limit = 1500
+gen_limit = 1200
 
 ########################## text rwkv ################################################################
 from rwkv.utils import PIPELINE, PIPELINE_ARGS
 
-title_v6 = "rwkv7-g1-0.1b-20250307-ctx4096"
+title_v6 = "rwkv7-g1-0.4b-20250324-ctx4096"
 model_path_v6 = hf_hub_download(repo_id="BlinkDL/rwkv7-g1", filename=f"{title_v6}.pth")
 model_v6 = RWKV(model=model_path_v6.replace('.pth',''), strategy='cuda fp16')
 pipeline_v6 = PIPELINE(model_v6, "rwkv_vocab_v20230424")
@@ -126,7 +126,7 @@ with gr.Blocks(title=title_v6) as demo:
     gr.HTML(f"<div style=\"text-align: center;\">\n<h1>{title_v6}</h1>\n</div>")
 
     with gr.Tab("=== Base Model (Raw Generation) ==="):
-        gr.Markdown(f'This is [RWKV7 G1](https://huggingface.co/BlinkDL/rwkv7-g1) 0.1B (!!!) L12-D768 reasoning base LM - an attention-free pure RNN [RWKV-LM](https://github.com/BlinkDL/RWKV-LM). Supports 100+ world languages and code. Check [400+ Github RWKV projects](https://github.com/search?o=desc&p=1&q=rwkv&s=updated&type=Repositories). *** Can try examples (bottom of page) *** (can edit them). Demo limited to ctxlen {ctx_limit}.')
+        gr.Markdown(f'This is [RWKV7 G1](https://huggingface.co/BlinkDL/rwkv7-g1) 0.4B (!) L24-D1024 reasoning base LM - an attention-free pure RNN [RWKV-LM](https://github.com/BlinkDL/RWKV-LM). Supports 100+ world languages and code. Check [400+ Github RWKV projects](https://github.com/search?o=desc&p=1&q=rwkv&s=updated&type=Repositories). *** Can try examples (bottom of page) *** (can edit them). Demo limited to ctxlen {ctx_limit}.')
         with gr.Row():
             with gr.Column():
                 prompt = gr.Textbox(lines=6, label="Prompt", value="User: simulate SpaceX mars landing using python\n\nAssistant: <think")