.

app.py

@@ -20,7 +20,7 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
 ctx_limit = 2500
 gen_limit = 500
-ENABLE_VISUAL = True
+ENABLE_VISUAL = False
 
 ########################## text rwkv ################################################################
 from rwkv.utils import PIPELINE, PIPELINE_ARGS
@@ -35,13 +35,13 @@ args = model_v6.args
 eng_name = 'rwkv-x060-eng_single_round_qa-3B-20240430-ctx1024'
 chn_name = 'rwkv-x060-chn_single_round_qa-3B-20240505-ctx1024'
 
-# state_eng_raw = torch.load(f'/mnt/e/RWKV-Runner/models/{eng_name}.pth')
-# state_chn_raw = torch.load(f'/mnt/e/RWKV-Runner/models/{chn_name}.pth')
+# state_eng_raw = torch.load(f'/mnt/e/RWKV-Runner/models/{eng_name}.pth', map_location=torch.device('cpu'))
+# state_chn_raw = torch.load(f'/mnt/e/RWKV-Runner/models/{chn_name}.pth', map_location=torch.device('cpu'))
 
 eng_file = hf_hub_download(repo_id="BlinkDL/temp-latest-training-models", filename=f"{eng_name}.pth")
 chn_file = hf_hub_download(repo_id="BlinkDL/temp-latest-training-models", filename=f"{chn_name}.pth")
-state_eng_raw = torch.load(eng_file)
-state_chn_raw = torch.load(chn_file)
+state_eng_raw = torch.load(eng_file, map_location=torch.device('cpu'))
+state_chn_raw = torch.load(chn_file, map_location=torch.device('cpu'))
 
 state_eng = [None] * args.n_layer * 3
 state_chn = [None] * args.n_layer * 3