UPDATE: G1A 2.9B 1.5B 0.4B

config.production.yaml

@@ -4,8 +4,8 @@ STRATEGY: "cuda fp16"
 RWKV_CUDA_ON: True
 CHUNK_LEN: 256
 MODELS:
-  - SERVICE_NAME: "rwkv7-g1-2.9b-20250519-ctx4096"
-    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1-2.9b-20250519-ctx4096.pth"
+  - SERVICE_NAME: "rwkv7-g1a-2.9b-20250924-ctx4096"
+    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1a-2.9b-20250924-ctx4096.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-g1-1.5b-20250429-ctx4096"
-    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1-1.5b-20250429-ctx4096.pth"
+  - SERVICE_NAME: "rwkv7-g1a-1.5b-20250922-ctx4096"
+    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1a-1.5b-20250922-ctx4096.pth"
     DOWNLOAD_MODEL_REPO_ID: "BlinkDL/rwkv7-g1"
     REASONING: True
     DEFAULT_SAMPLER:
@@ -34,8 +34,8 @@ MODELS:
         - "\n\n"
       stop_tokens:
         - 0
-  - SERVICE_NAME: "rwkv7-g1-0.4b-20250324-ctx4096"
-    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1-0.4b-20250324-ctx4096.pth"
+  - SERVICE_NAME: "rwkv7-g1a-0.4b-20250905-ctx4096"
+    DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1a-0.4b-20250905-ctx4096.pth"
     DOWNLOAD_MODEL_REPO_ID: "BlinkDL/rwkv7-g1"
     REASONING: True
     DEFAULT_SAMPLER:

pyproject.toml

@@ -14,7 +14,7 @@ dependencies = [
     "pydantic-settings>=2.8.1",
     "pynvml>=12.0.0",
     "rich>=13.9.4",
-    "rwkv>=0.8.28",
+    "rwkv>=0.8.30",
     "setuptools>=75.8.2",
     "snowflake-id>=1.0.2",
     "modelscope>=1.23.0",