Fine tune the Janus-Pro-7B model using ms swift

[furanger]

Fine tune the Janus-Pro-7B model using ms swift,
And add a proxy service interface,
Integrate Swift with fastapi_cient py
The client can directly use fastapi_cient Py doesn't need to be changed

[NTUYWANG103]

Thanks for your sharing with finetune readme, could u release a more detail version of "Fine tune the Janus-Pro-7B model using ms swift", thanks!

[GunsRuth]

Please tell your method to run this on local dataset

[furanger]

If there are multiple regulatory requirements for simultaneous training (seat belts, helmets, hot work, vehicles, etc.), Ensure that multiple standards have the same beginning and use concise language to guide the corresponding corpus into their own classification For example, employees need to comply with safety regulations in the work area and XXX in the driving area of construction vehicles Employees need to comply with safety regulations in the work area and work at heights in XXXXX Employees need to comply with safety regulations in the work area, and in the high-voltage equipment operation area XXXXX Using seat belts as an example to find regulations related to high-altitude operations, Extract the conditions that can be standardized through video, such as correctly wearing high-altitude seat belts, hanging them high, and using them low, Then, the conditions are formed into a description of the corpus, which can be divided into four situations Proper pairing High hanging low usage Low hanging high usage No accompanying accessories Then, according to the rules and regulations, provide a concise and direct explanation of the situation, btw: If you want to go further, you can first describe the image and then add the content of the standard image and a risk explanation at the beginning. This way, the results given will be more spiritual and not overfitting. My example is overfitting FURa0-ger From: Bitan Mallik Date: 2025-03-28 01:03 To: deepseek-ai/Janus CC: che.ender; Author Subject: Re: [deepseek-ai/Janus] Fine tune the Janus-Pro-7B model using ms swift (PR #189) Please tell your method to run this on local dataset — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***> GunsRuth left a comment (deepseek-ai/Janus#189) Please tell your method to run this on local dataset — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[furanger]

dataset images file 通过网盘分享的文件：安全帽 链接: https://pan.baidu.com/s/1S_IW4R1Rv25H33LbWVvTew 提取码: whdi FURanger From: FURanger Date: 2025-03-28 07:49 To: deepseek-ai/Janus Subject: Re: Re: [deepseek-ai/Janus] Fine tune the Janus-Pro-7B model using ms swift (PR #189) If there are multiple regulatory requirements for simultaneous training (seat belts, helmets, hot work, vehicles, etc.), Ensure that multiple standards have the same beginning and use concise language to guide the corresponding corpus into their own classification For example, employees need to comply with safety regulations in the work area and XXX in the driving area of construction vehicles Employees need to comply with safety regulations in the work area and work at heights in XXXXX Employees need to comply with safety regulations in the work area, and in the high-voltage equipment operation area XXXXX Using seat belts as an example to find regulations related to high-altitude operations, Extract the conditions that can be standardized through video, such as correctly wearing high-altitude seat belts, hanging them high, and using them low, Then, the conditions are formed into a description of the corpus, which can be divided into four situations Proper pairing High hanging low usage Low hanging high usage No accompanying accessories Then, according to the rules and regulations, provide a concise and direct explanation of the situation, btw: If you want to go further, you can first describe the image and then add the content of the standard image and a risk explanation at the beginning. This way, the results given will be more spiritual and not overfitting. My example is overfitting FURa0-ger From: Bitan Mallik Date: 2025-03-28 01:03 To: deepseek-ai/Janus CC: che.ender; Author Subject: Re: [deepseek-ai/Janus] Fine tune the Janus-Pro-7B model using ms swift (PR #189) Please tell your method to run this on local dataset — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***> GunsRuth left a comment (deepseek-ai/Janus#189) Please tell your method to run this on local dataset — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>