xinference 加载的rerank模型 添加进来后无法调用

[sinopec]

例行检查

• [x ] 我已确认目前没有类似 issue
• [ x] 我已确认我已升级到最新版本
• [ x] 我已完整查看过项目 README，尤其是常见问题部分
• [x ] 我理解并愿意跟进此 issue，协助测试和提供反馈
• [x ] 我理解并认可上述内容，并理解项目维护者精力有限，不遵循规则的 issue 可能会被无视或直接关闭

问题描述
通过xinference 加载的rerank模型bge-reranker-v2-m3 ，使用curl脚本测试正常。

但是通过渠道添加进new-api后，使用同样的脚本，调用返回
{"error":{"message":"not implemented (request id: 202502261550548185154107K8AGBRS)","type":"new_api_error","param":"","code":"convert_request_failed"}}

查看后台log，记录如下：
new-api | [SYS] 2025/02/26 - 16:32:33 | model ratio not found: bge-reranker-v2-m3
new-api | [INFO] 2025/02/26 - 16:32:33 | 20250226163233233467847qyMQRCIm | user 1 with unlimited token has enough quota ＄199.710980, trusted and no need to pre-consume
new-api | [GIN] 2025/02/26 - 16:32:33 | 20250226163233233467847qyMQRCIm | 500 | 765.567µs | 10.30.0.123 | POST /v1/rerank
new-api | [ERR] 2025/02/26 - 16:32:33 | 20250226163233233467847qyMQRCIm | relay error (channel #2, status code: 500): not implemented (request id: 20250226163233233467847qyMQRCIm)

复现步骤
xinference 启动 rerank模型，
将xinference渠道添加new-api
使用下面脚本测试rereank功能
curl -X 'POST'
'http:/10.255.114.31:3000/v1/rerank'
-H 'accept: application/json'
-H 'Content-Type: application/json'
-H 'Authorization: Bearer sk-aaaa'
-d '{
"model": "bge-reranker-v2-m3",
"query": "A man is eating pasta.",
"documents": [
"A man is eating food.",
"A man is eating a piece of bread.",
"The girl is carrying a baby.",
"A man is riding a horse.",
"A woman is playing violin."
]
}'

预期结果

如同直接访问xinfeence端点一样正常返回结果
{
"id": "9c815c12-f45b-11ef-a5a9-bff6a7a8f883",
"results": [
{
"index": 0,
"relevance_score": 0.999574601650238,
"document": null
},
{
"index": 1,
"relevance_score": 0.07814765721559525,
"document": null
},
{
"index": 3,
"relevance_score": 1.7700713215162978e-05,
"document": null
},
{
"index": 2,
"relevance_score": 1.63753629749408e-05,
"document": null
},
{
"index": 4,
"relevance_score": 1.6318974303430878e-05,
"document": null
}
],
"meta": {
"api_version": null,
"billed_units": null,
"tokens": null,
"warnings": null
}
}

相关截图

渠道设置如图：

[Calcium-Ion]

你用openai渠道类型接入当然不能用了，况且我们也没有兼容xinference的rerank