[GH-ISSUE #625] Please increase limit of Epoches over 1000, and allows use of GPU Patch Size over 40 #361

New issue

Closed

opened 2026-10-20 09:58:32 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-20 09:58:32 +00:00

Owner

Originally created by @takilung on GitHub (Jun 24, 2023).
Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/625

MY GPU and PC can handle it at max, I would like to do at least 2000 Epoches for closest voice to original VA, and GPU being 40 is just too low for patch size. RTX 4090 can handle way more than that thank you please

Originally created by @takilung on GitHub (Jun 24, 2023). Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/625 MY GPU and PC can handle it at max, I would like to do at least 2000 Epoches for closest voice to original VA, and GPU being 40 is just too low for patch size. RTX 4090 can handle way more than that thank you please

bsakaguchi

2026-10-20 09:58:32 +00:00

closed this issue
added the
stale
label

bsakaguchi commented

2026-10-20 09:58:33 +00:00

Author

Owner

@TripleKiller666 commented on GitHub (Jun 28, 2023):

EZ, you just change the below Values in your infer-web.py file:

            with gr.Group():
                gr.Markdown(value=i18n("step3: 填写训练设置, 开始训练模型和索引"))
                with gr.Row():
                    save_epoch10 = gr.Slider(
                        minimum=0,
                        maximum=50,
                        step=1,
                        label=i18n("保存频率save_every_epoch"),
                        value=5,
                        interactive=True,
                    )
                    total_epoch11 = gr.Slider(
                        minimum=0,
                        maximum=1000,
                        step=1,
                        label=i18n("总训练轮数total_epoch"),
                        value=20,
                        interactive=True,
                    )
                    batch_size12 = gr.Slider(
                        minimum=1,
                        maximum=40,
                        step=1,
                        label=i18n("每张显卡的batch_size"),
                        value=default_batch_size,
                        interactive=True,
                    )
                    if_save_latest13 = gr.Radio(
                        label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"),
                        choices=[i18n("是"), i18n("否")],
                        value=i18n("否"),
                        interactive=True,
                    )

For example 9999 Epochs max:

                    total_epoch11 = gr.Slider(
                        minimum=0,
                        maximum=9999,
                        step=1,
                        label=i18n("总训练轮数total_epoch"),

Edit:
The value part is the Default Value that gets selected when you start/restart the Gradio UI

@TripleKiller666 commented on GitHub (Jun 28, 2023): EZ, you just change the below Values in your infer-web.py file: ``` with gr.Group(): gr.Markdown(value=i18n("step3: 填写训练设置, 开始训练模型和索引")) with gr.Row(): save_epoch10 = gr.Slider( minimum=0, maximum=50, step=1, label=i18n("保存频率save_every_epoch"), value=5, interactive=True, ) total_epoch11 = gr.Slider( minimum=0, maximum=1000, step=1, label=i18n("总训练轮数total_epoch"), value=20, interactive=True, ) batch_size12 = gr.Slider( minimum=1, maximum=40, step=1, label=i18n("每张显卡的batch_size"), value=default_batch_size, interactive=True, ) if_save_latest13 = gr.Radio( label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"), choices=[i18n("是"), i18n("否")], value=i18n("否"), interactive=True, ) ``` For example 9999 Epochs max: ``` total_epoch11 = gr.Slider( minimum=0, maximum=9999, step=1, label=i18n("总训练轮数total_epoch"), ``` Edit: The value part is the Default Value that gets selected when you start/restart the Gradio UI

bsakaguchi commented

2026-10-20 09:58:33 +00:00

Author

Owner

@github-actions[bot] commented on GitHub (Apr 29, 2024):

This issue was closed because it has been inactive for 15 days since being marked as stale.

@github-actions[bot] commented on GitHub (Apr 29, 2024): This issue was closed because it has been inactive for 15 days since being marked as stale.