Bladeren bron

fix: moonshot context size error (#2406)

takatost 1 jaar geleden
bovenliggende
commit
3242cf5384

+ 1 - 0
api/core/model_runtime/model_providers/_position.yaml

@@ -14,6 +14,7 @@
 - minimax
 - tongyi
 - wenxin
+- moonshot
 - jina
 - chatglm
 - xinference

+ 3 - 7
api/core/model_runtime/model_providers/moonshot/llm/moonshot-v1-128k.yaml

@@ -7,21 +7,17 @@ features:
   - agent-thought
 model_properties:
   mode: chat
-  context_size: 4096
+  context_size: 128000
 parameter_rules:
   - name: temperature
     use_template: temperature
   - name: top_p
     use_template: top_p
-  - name: presence_penalty
-    use_template: presence_penalty
-  - name: frequency_penalty
-    use_template: frequency_penalty
   - name: max_tokens
     use_template: max_tokens
-    default: 512
+    default: 1024
     min: 1
-    max: 4096
+    max: 128000
 pricing:
   input: '0.06'
   output: '0.06'

+ 3 - 7
api/core/model_runtime/model_providers/moonshot/llm/moonshot-v1-32k.yaml

@@ -7,21 +7,17 @@ features:
   - agent-thought
 model_properties:
   mode: chat
-  context_size: 4096
+  context_size: 32000
 parameter_rules:
   - name: temperature
     use_template: temperature
   - name: top_p
     use_template: top_p
-  - name: presence_penalty
-    use_template: presence_penalty
-  - name: frequency_penalty
-    use_template: frequency_penalty
   - name: max_tokens
     use_template: max_tokens
-    default: 512
+    default: 1024
     min: 1
-    max: 4096
+    max: 32000
 pricing:
   input: '0.024'
   output: '0.024'

+ 2 - 6
api/core/model_runtime/model_providers/moonshot/llm/moonshot-v1-8k.yaml

@@ -7,21 +7,17 @@ features:
   - agent-thought
 model_properties:
   mode: chat
-  context_size: 4096
+  context_size: 8192
 parameter_rules:
   - name: temperature
     use_template: temperature
   - name: top_p
     use_template: top_p
-  - name: presence_penalty
-    use_template: presence_penalty
-  - name: frequency_penalty
-    use_template: frequency_penalty
   - name: max_tokens
     use_template: max_tokens
     default: 512
     min: 1
-    max: 4096
+    max: 8192
 pricing:
   input: '0.012'
   output: '0.012'