testing

app.py

@@ -12,11 +12,12 @@ load_dotenv()
 bot = commands.Bot("", intents=discord.Intents.all())
 
 quantized = {
-    "mode": TORCH_QUANT,
+    "mode": TORCH,
     "runtimedtype": torch.float32,
     "useGPU": torch.cuda.is_available(),
-    "chunksize": 32,  # larger = more accurate, but more memory
-    "target": 100  # your gpu max size, excess vram offloaded to cpu
+    # "chunksize": 32,  # larger = more accurate, but more memory
+    # "target": 100,  # your gpu max size, excess vram offloaded to cpu
+    "dtype": torch.float32
 }
 
 # model_pipe = pickle.load(open('pipe.pkl', 'rb'))
@@ -33,7 +34,7 @@ useGPU = torch.cuda.is_available() # False
 async def on_ready():
     print(f'We have logged in as {bot.user}')
     global model
-    model = RWKV(path="https://huggingface.co/BlinkDL/rwkv-4-pile-1b5/resolve/main/RWKV-4-Pile-1B5-Instruct-test1-20230124.pth", 
+    model = RWKV(path="https://huggingface.co/BlinkDL/rwkv-4-pile-3b/resolve/main/RWKV-4a-Pile-3B-20230110-8020.pth", 
                  **quantized)