Amber-Believe comments

Results 12 comments of


                                            Amber-Believe

[Bug] 更改session_len无法解决输出截断的问题

好的问题已解决

[Performance]: V0.6.0 version of the model, benchmarking found that the number of successful responses accounted for half of the number of requests, which is why

![企业微信截图_17259622432984](https://github.com/user-attachments/assets/35b3d728-a6ff-4d1a-ba2d-8fefeef0e6ac) In the v0.6.0 model, the total number of requests is 1000, -- the request-rate is 100, 1) only half of the requests are successfully responded to, why do so...

[Performance]: V0.6.0 version of the model, benchmarking found that the number of successful responses accounted for half of the number of requests, which is why

> @Amber-Believe确保你关注了#8176 Hello, so the meaning here is that the number of successful responses is only half due to the problem of the multi-step scheduler (--num-scheduler-steps 10)?

[Performance]: V0.6.0 version of the model, benchmarking found that the number of successful responses accounted for half of the number of requests, which is why

Okay, I'll try the new version of vllm

[Bug]对VLM来说，是否支持多个lora adapter使用

> 是说 vision encoder 加了 lora 了么？这个还没有支持使用lora训练多个领域的能力，然后对应多个adapter，想要根据不同的场景，加载不同的adapter，然后通过调用API的方式使用，这种支持吗（多模态）

[Bug]对VLM来说，是否支持多个lora adapter使用

有点像动态lora，通过专用 API 端点在运行时动态配置 LoRA 适配器，可以动态更改模型这种

模型训练，使用的是哪一方的资源

Alice方（服务方）： #ray start --head --node-ip-address="ip" --port="3000" --include-dashboard=False --disable-usage-stats import secretflow as sf import spu from secretflow.ml.boost.homo_boost import SFXgboost from secretflow.security.aggregation import SPUAggregator, SecureAggregator print('The version of SecretFlow: {}'.format(sf.__version__)) sf.shutdown() cluster_config...