Skip to content

Commit

Permalink
Add grok-vision-beta to xAI + Update grok-beta Features (#11004)
Browse files Browse the repository at this point in the history
  • Loading branch information
taowang1993 authored Nov 25, 2024
1 parent 56e361a commit 1065917
Show file tree
Hide file tree
Showing 3 changed files with 70 additions and 1 deletion.
5 changes: 4 additions & 1 deletion api/core/model_runtime/model_providers/x/llm/grok-beta.yaml
Original file line number Diff line number Diff line change
@@ -1,9 +1,12 @@
model: grok-beta
label:
en_US: Grok beta
en_US: Grok Beta
model_type: llm
features:
- agent-thought
- tool-call
- multi-tool-call
- stream-tool-call
model_properties:
mode: chat
context_size: 131072
Expand Down
64 changes: 64 additions & 0 deletions api/core/model_runtime/model_providers/x/llm/grok-vision-beta.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,64 @@
model: grok-vision-beta
label:
en_US: Grok Vision Beta
model_type: llm
features:
- agent-thought
- vision
model_properties:
mode: chat
context_size: 8192
parameter_rules:
- name: temperature
label:
en_US: "Temperature"
zh_Hans: "采样温度"
type: float
default: 0.7
min: 0.0
max: 2.0
precision: 1
required: true
help:
en_US: "The randomness of the sampling temperature control output. The temperature value is within the range of [0.0, 1.0]. The higher the value, the more random and creative the output; the lower the value, the more stable it is. It is recommended to adjust either top_p or temperature parameters according to your needs to avoid adjusting both at the same time."
zh_Hans: "采样温度控制输出的随机性。温度值在 [0.0, 1.0] 范围内,值越高,输出越随机和创造性;值越低,输出越稳定。建议根据需求调整 top_p 或 temperature 参数,避免同时调整两者。"

- name: top_p
label:
en_US: "Top P"
zh_Hans: "Top P"
type: float
default: 0.7
min: 0.0
max: 1.0
precision: 1
required: true
help:
en_US: "The value range of the sampling method is [0.0, 1.0]. The top_p value determines that the model selects tokens from the top p% of candidate words with the highest probability; when top_p is 0, this parameter is invalid. It is recommended to adjust either top_p or temperature parameters according to your needs to avoid adjusting both at the same time."
zh_Hans: "采样方法的取值范围为 [0.0,1.0]。top_p 值确定模型从概率最高的前p%的候选词中选取 tokens;当 top_p 为 0 时,此参数无效。建议根据需求调整 top_p 或 temperature 参数,避免同时调整两者。"

- name: frequency_penalty
use_template: frequency_penalty
label:
en_US: "Frequency Penalty"
zh_Hans: "频率惩罚"
type: float
default: 0
min: 0
max: 2.0
precision: 1
required: false
help:
en_US: "Number between 0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim."
zh_Hans: "介于0和2.0之间的数字。正值会根据新标记在文本中迄今为止的现有频率来惩罚它们,从而降低模型一字不差地重复同一句话的可能性。"

- name: user
use_template: text
label:
en_US: "User"
zh_Hans: "用户"
type: string
required: false
help:
en_US: "Used to track and differentiate conversation requests from different users."
zh_Hans: "用于追踪和区分不同用户的对话请求。"
2 changes: 2 additions & 0 deletions api/core/model_runtime/model_providers/x/llm/llm.py
Original file line number Diff line number Diff line change
Expand Up @@ -35,3 +35,5 @@ def _add_custom_parameters(credentials) -> None:
credentials["endpoint_url"] = str(URL(credentials["endpoint_url"])) or "https://api.x.ai/v1"
credentials["mode"] = LLMMode.CHAT.value
credentials["function_calling_type"] = "tool_call"
credentials["stream_function_calling"] = "support"
credentials["vision_support"] = "support"

0 comments on commit 1065917

Please sign in to comment.