Xuanqing Liu

(Feature) Add default timezone in config file

5

## Summary Currently, the timezone seems to be based on geolocation. This is not very convenient because I sometimes want to schedule in another timezone. Allowing users to set up...

Demo for Java API v2.0?

1

Hi, thanks for the great work of Java API. I am able to follow the examples in compatibility mode. But since it is deprecated, could you also provide some examples...

[question] SAC target q nets may be updated many times in each round?

5

This line of code: https://github.com/hill-a/stable-baselines/blob/d288b6d911094268b17abb61dbe9ac3ca05e1ce3/stable_baselines/sac/sac.py#L464 since `step` is in `range(total_timesteps)` and `grad_step` is nested in `range(self.gradient_steps)`, then if self.gradient_steps is not 1, the target networks are updated repeatedly?

Any plan to continue training beyond 1T tokens?

4

Thanks for generously releasing the checkpoint. According to LLAMA paper, the downstream task performance keeps improving even at 1T tokens (figure below), I am wondering if you have the budget...

https://github.com/openvinotoolkit/openvino_contrib/blob/6c16acff2ae0578008b171706c776ee566211cd2/modules/java_api/src/main/java/org/intel/openvino/NativeLibrary.java#L89 Hi, I compiled the OpenVINO library together with the tbb bundled with openvino. I am on linux (ubuntu and centos) and confirmed that no libtbb.so.2 generated (only libtbb.so ->...

Could you share the training loss to improve reproducibility?

4

Hi, thanks for sharing the datasets! I'm trying to train a flan model using t5 and other backbone models. However i'm not confident enough on how well I reproduced your...

I really appreciate your code, it helps me a lot to reproduce the squad results, very good job! One thing I am wondering is that how you write the tokenizer...