update wenxin llm (#2929)

2 years ago · 5becb4c43a
parent 13694293e3
commit 5becb4c43a
16 changed files with 357 additions and 3 deletions
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-3.5-4k-0205.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-3.5-4k-0205.yaml
@ -0,0 +1,37 @@
+model: ernie-3.5-8k
+label:
+  en_US: Ernie-3.5-8K
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 4096
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.8
+  - name: top_p
+    use_template: top_p
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 2048
+  - name: presence_penalty
+    use_template: presence_penalty
+  - name: frequency_penalty
+    use_template: frequency_penalty
+  - name: response_format
+    use_template: response_format
+  - name: disable_search
+    label:
+      zh_Hans: 禁用搜索
+      en_US: Disable Search
+    type: boolean
+    help:
+      zh_Hans: 禁用模型自行进行外部搜索。
+      en_US: Disable the model to perform external search.
+    required: false
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-3.5-8k-0205.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-3.5-8k-0205.yaml
@ -0,0 +1,37 @@
+model: ernie-3.5-8k-0205
+label:
+  en_US: Ernie-3.5-8K-0205
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.8
+  - name: top_p
+    use_template: top_p
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 2048
+  - name: presence_penalty
+    use_template: presence_penalty
+  - name: frequency_penalty
+    use_template: frequency_penalty
+  - name: response_format
+    use_template: response_format
+  - name: disable_search
+    label:
+      zh_Hans: 禁用搜索
+      en_US: Disable Search
+    type: boolean
+    help:
+      zh_Hans: 禁用模型自行进行外部搜索。
+      en_US: Disable the model to perform external search.
+    required: false
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-3.5-8k-1222.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-3.5-8k-1222.yaml
@ -0,0 +1,37 @@
+model: ernie-3.5-8k-1222
+label:
+  en_US: Ernie-3.5-8K-1222
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.8
+  - name: top_p
+    use_template: top_p
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 2048
+  - name: presence_penalty
+    use_template: presence_penalty
+  - name: frequency_penalty
+    use_template: frequency_penalty
+  - name: response_format
+    use_template: response_format
+  - name: disable_search
+    label:
+      zh_Hans: 禁用搜索
+      en_US: Disable Search
+    type: boolean
+    help:
+      zh_Hans: 禁用模型自行进行外部搜索。
+      en_US: Disable the model to perform external search.
+    required: false
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-3.5-8k.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-3.5-8k.yaml
@ -0,0 +1,37 @@
+model: ernie-3.5-8k
+label:
+  en_US: Ernie-3.5-8K
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.8
+  - name: top_p
+    use_template: top_p
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 2048
+  - name: presence_penalty
+    use_template: presence_penalty
+  - name: frequency_penalty
+    use_template: frequency_penalty
+  - name: response_format
+    use_template: response_format
+  - name: disable_search
+    label:
+      zh_Hans: 禁用搜索
+      en_US: Disable Search
+    type: boolean
+    help:
+      zh_Hans: 禁用模型自行进行外部搜索。
+      en_US: Disable the model to perform external search.
+    required: false
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-4.0-8k.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-4.0-8k.yaml
@ -0,0 +1,37 @@
+model: ernie-4.0-8k
+label:
+  en_US: Ernie-4.0-8K
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.8
+  - name: top_p
+    use_template: top_p
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 2048
+  - name: presence_penalty
+    use_template: presence_penalty
+  - name: frequency_penalty
+    use_template: frequency_penalty
+  - name: response_format
+    use_template: response_format
+  - name: disable_search
+    label:
+      zh_Hans: 禁用搜索
+      en_US: Disable Search
+    type: boolean
+    help:
+      zh_Hans: 禁用模型自行进行外部搜索。
+      en_US: Disable the model to perform external search.
+    required: false
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-bot-4.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-bot-4.yaml
@ -36,3 +36,4 @@ parameter_rules:
      zh_Hans: 禁用模型自行进行外部搜索。
      en_US: Disable the model to perform external search.
    required: false
+deprecated: true
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-bot-8k.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-bot-8k.yaml
@ -36,3 +36,4 @@ parameter_rules:
      zh_Hans: 禁用模型自行进行外部搜索。
      en_US: Disable the model to perform external search.
    required: false
+deprecated: true
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-bot-turbo.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-bot-turbo.yaml
@ -27,3 +27,4 @@ parameter_rules:
    use_template: frequency_penalty
  - name: response_format
    use_template: response_format
+deprecated: true
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-bot.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-bot.yaml
@ -36,3 +36,4 @@ parameter_rules:
    required: false
  - name: response_format
    use_template: response_format
+deprecated: true
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-lite-8k-0308.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-lite-8k-0308.yaml
@ -0,0 +1,30 @@
+model: ernie-lite-8k-0308
+label:
+  en_US: ERNIE-Lite-8K-0308
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.95
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1.0
+    default: 0.7
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 1024
+  - name: presence_penalty
+    use_template: presence_penalty
+    default: 1.0
+    min: 1.0
+    max: 2.0
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-lite-8k-0922.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-lite-8k-0922.yaml
@ -0,0 +1,30 @@
+model: ernie-lite-8k-0922
+label:
+  en_US: ERNIE-Lite-8K-0922
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.95
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1.0
+    default: 0.7
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 1024
+  - name: presence_penalty
+    use_template: presence_penalty
+    default: 1.0
+    min: 1.0
+    max: 2.0
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-speed-128k.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-speed-128k.yaml
@ -0,0 +1,30 @@
+model: ernie-speed-128k
+label:
+  en_US: ERNIE-Speed-128K
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 128000
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.95
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1.0
+    default: 0.7
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 1024
+  - name: presence_penalty
+    use_template: presence_penalty
+    default: 1.0
+    min: 1.0
+    max: 2.0
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-speed-8k.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-speed-8k.yaml
@ -0,0 +1,30 @@
+model: ernie-speed-8k
+label:
+  en_US: ERNIE-Speed-8K
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.95
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1.0
+    default: 0.7
+  - name: max_tokens
+    use_template: max_tokens
+    default: 1024
+    min: 2
+    max: 1024
+  - name: presence_penalty
+    use_template: presence_penalty
+    default: 1.0
+    min: 1.0
+    max: 2.0
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie-speed-appbuilder.yaml
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie-speed-appbuilder.yaml
@ -0,0 +1,25 @@
+model: ernie-speed-appbuilder
+label:
+  en_US: ERNIE-Speed-AppBuilder
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0.1
+    max: 1.0
+    default: 0.95
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1.0
+    default: 0.7
+  - name: presence_penalty
+    use_template: presence_penalty
+    default: 1.0
+    min: 1.0
+    max: 2.0
--- a/api/core/model_runtime/model_providers/wenxin/llm/ernie_bot.py
+++ b/api/core/model_runtime/model_providers/wenxin/llm/ernie_bot.py
@ -121,15 +121,29 @@ class ErnieMessage:

 class ErnieBotModel:
    api_bases = {
-        'ernie-bot': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions',
+        'ernie-bot': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-3.5-4k-0205',
        'ernie-bot-4': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions_pro',
-        'ernie-bot-8k': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie_bot_8k',
+        'ernie-bot-8k': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions',
        'ernie-bot-turbo': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/eb-instant',
+        'ernie-3.5-8k': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions',
+        'ernie-3.5-8k-0205': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-3.5-8k-0205',
+        'ernie-3.5-8k-1222': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-3.5-8k-1222',
+        'ernie-3.5-4k-0205': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-3.5-4k-0205',
+        'ernie-4.0-8k': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions_pro',
+        'ernie-speed-8k': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie_speed',
+        'ernie-speed-128k': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-speed-128k',
+        'ernie-speed-appbuilder': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ai_apaas',
+        'ernie-lite-8k-0922': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/eb-instant',
+        'ernie-lite-8k-0308': 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-lite-8k',
    }

    function_calling_supports = [
        'ernie-bot',
        'ernie-bot-8k',
+        'ernie-3.5-8k',
+        'ernie-3.5-8k-0205',
+        'ernie-3.5-8k-1222',
+        'ernie-3.5-4k-0205'
    ]

    api_key: str = ''
@ -285,6 +299,12 @@ class ErnieBotModel:
            **parameters
        }

+        if 'max_tokens' in parameters and type(parameters['max_tokens']) == int:
+            body['max_output_tokens'] = parameters['max_tokens']
+
+        if 'presence_penalty' in parameters and type(parameters['presence_penalty']) == float:
+            body['penalty_score'] = parameters['presence_penalty']
+
        if system_message:
            body['system'] = system_message

--- a/api/tests/integration_tests/model_runtime/wenxin/test_llm.py
+++ b/api/tests/integration_tests/model_runtime/wenxin/test_llm.py
@ -154,7 +154,7 @@ def test_invoke_stream_model():
    model = ErnieBotLargeLanguageModel()

    response = model.invoke(
-        model='ernie-bot',
+        model='ernie-3.5-8k',
        credentials={
            'api_key': os.environ.get('WENXIN_API_KEY'),
            'secret_key': os.environ.get('WENXIN_SECRET_KEY')