UPDATE: g1a4-2.9b-20251118, g1a3-1.5b-20251015

config.production.yaml

@@ -4,8 +4,8 @@ STRATEGY: "cuda fp16"
 RWKV_CUDA_ON: True
 CHUNK_LEN: 256
 MODELS:
-  - SERVICE_NAME: "rwkv7-g1a-2.9b-20250924-ctx4096"
-    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1a-2.9b-20250924-ctx4096.pth"
+  - SERVICE_NAME: "rwkv7-g1a4-2.9b-20251118-ctx8192"
+    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1a4-2.9b-20251118-ctx8192.pth"
     DOWNLOAD_MODEL_REPO_ID: "BlinkDL/rwkv7-g1"
     REASONING: True
     DEFAULT_SAMPLER:
@@ -19,8 +19,8 @@ MODELS:
         - "\n\n"
       stop_tokens:
         - 0
-  - SERVICE_NAME: "rwkv7-g1a-1.5b-20250922-ctx4096"
-    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1a-1.5b-20250922-ctx4096.pth"
+  - SERVICE_NAME: "rwkv7-g1a3-1.5b-20251015-ctx8192"
+    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1a3-1.5b-20251015-ctx8192.pth"
     DOWNLOAD_MODEL_REPO_ID: "BlinkDL/rwkv7-g1"
     REASONING: True
     DEFAULT_SAMPLER: