Any plan to release training details such as hyper-parameters(epoch, batch, learning rate) and loss curve

[purezhanghan]

Thanks for open-sourcing the project and model.

[Peng-fei-Wang]

Hi @purezhanghan , thanks for your interest in our work!

We plan to include more training details, such as the hyper-parameters you mentioned, in a future revision of our paper. Please keep an eye on the paper for updates.

[Peng-fei-Wang]

@purezhanghan
Detailed hyperparameters and model configurations are provided in the table below:

Category	Hyperparameter	Generator	Selector
Data & Model	Base Model	OmniSQL-32B	Qwen2.5-Coder-32B
	Max Prompt Length	9216	9216
	Max Response Length	2048	2048
GRPO Algorithm	Reward Function	syntax, execution	selection
	Mini-Batch Size	64	128
	Micro-Batch Size / GPU	1	2
	KL Divergence Coefficient	0.001	0.001
	Learning Rate	1e-6	1e-6
	LR Warmup Steps Ratio	0.1	0.1
Rollout	Rollout Engine	SGLang	SGLang
	GPU Memory Utilization	0.5	0.8
	N	16	16
	Log-Prob Micro-Batch Size / GPU	1	2
	Ref Model Log-Prob Micro-Batch / GPU	1	2
	Tensor Parallel Size (TP)	8	8
Training	Total Epochs	3	5
	Total GPUs	32	32
	GPUs per Node	8	8
	Total Nodes	4	4
	GPU Type	A100	A100