[megatron] support bailing_v25 megatron

[Jintao-Huang]

modelscope/mcore-bridge#85

[gemini-code-assist]

Code Review

This pull request updates the supported models documentation in both Chinese and English. It adds three new MiniCPM5-1B models from OpenBMB and marks several Ling and Ring models from inclusionAI as supported. The reviewer noted a potential typo in the required transformers library version for the new MiniCPM5 models, suggesting it should be >=4.46 instead of >=5.6.

[Jintao-Huang]

/gemini review

[gemini-code-assist]

Code Review

This pull request adds support for MiniCPM5-1B models and updates the support status of several Ling and Ring models in the documentation. It also updates the Megatron conversion utility to handle padding-free arguments by duplicating inputs. Feedback highlights a potential issue in convert_utils.py where multiplying a list of mutable dictionaries ([inputs] * N) can lead to unintended side effects, suggesting the use of copy.deepcopy instead. Additionally, the documentation changes contain a typo specifying a non-existent transformers>=5.6 version, which should be corrected to a valid 4.x version.

[gemini-code-assist]

Duplicating a list containing a mutable dictionary using [inputs] * N creates a list of references to the exact same dictionary object. If template.data_collator modifies the input dictionaries in-place (e.g., popping keys, adding padding, or converting lists to tensors in-place), this will lead to unexpected side effects, duplicate modifications, or runtime errors (such as KeyError when popping keys). To prevent this, use a list comprehension with copy.deepcopy(inputs) to ensure each element is a completely independent copy.

Suggested change

	mg_inputs = to_device(
	template.data_collator([inputs] * (2 if args.padding_free else 1), padding_to=get_padding_to(args)), 'cuda')
	import copy
	mg_inputs = to_device(
	template.data_collator([copy.deepcopy(inputs) for _ in range(2 if args.padding_free else 1)], padding_to=get_padding_to(args)), 'cuda')

[gemini-code-assist]

The Hugging Face transformers library does not currently have a version 5.6 (the latest is in the 4.x range). This is likely a typo for transformers>=4.40 or similar. Please verify and correct the required version.

Suggested change

	|[OpenBMB/MiniCPM5-1B](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B)|llama|minicpm5|transformers>=5.6|✔|-|[openbmb/MiniCPM5-1B](https://huggingface.co/openbmb/MiniCPM5-1B)|
	|[OpenBMB/MiniCPM5-1B-Base](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B-Base)|llama|minicpm5|transformers>=5.6|✔|-|[openbmb/MiniCPM5-1B-Base](https://huggingface.co/openbmb/MiniCPM5-1B-Base)|
	|[OpenBMB/MiniCPM5-1B-SFT](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B-SFT)|llama|minicpm5|transformers>=5.6|✔|-|[openbmb/MiniCPM5-1B-SFT](https://huggingface.co/openbmb/MiniCPM5-1B-SFT)|
	|[OpenBMB/MiniCPM5-1B](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B)|llama|minicpm5|transformers>=4.40|✔|-|[openbmb/MiniCPM5-1B](https://huggingface.co/openbmb/MiniCPM5-1B)|
	|[OpenBMB/MiniCPM5-1B-Base](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B-Base)|llama|minicpm5|transformers>=4.40|✔|-|[openbmb/MiniCPM5-1B-Base](https://huggingface.co/openbmb/MiniCPM5-1B-Base)|
	|[OpenBMB/MiniCPM5-1B-SFT](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B-SFT)|llama|minicpm5|transformers>=4.40|✔|-|[openbmb/MiniCPM5-1B-SFT](https://huggingface.co/openbmb/MiniCPM5-1B-SFT)|

[gemini-code-assist]

The Hugging Face transformers library does not currently have a version 5.6 (the latest is in the 4.x range). This is likely a typo for transformers>=4.40 or similar. Please verify and correct the required version.

Suggested change

	|[OpenBMB/MiniCPM5-1B](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B)|llama|minicpm5|transformers>=5.6|✔|-|[openbmb/MiniCPM5-1B](https://huggingface.co/openbmb/MiniCPM5-1B)|
	|[OpenBMB/MiniCPM5-1B-Base](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B-Base)|llama|minicpm5|transformers>=5.6|✔|-|[openbmb/MiniCPM5-1B-Base](https://huggingface.co/openbmb/MiniCPM5-1B-Base)|
	|[OpenBMB/MiniCPM5-1B-SFT](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B-SFT)|llama|minicpm5|transformers>=5.6|✔|-|[openbmb/MiniCPM5-1B-SFT](https://huggingface.co/openbmb/MiniCPM5-1B-SFT)|
	|[OpenBMB/MiniCPM5-1B](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B)|llama|minicpm5|transformers>=4.40|✔|-|[openbmb/MiniCPM5-1B](https://huggingface.co/openbmb/MiniCPM5-1B)|
	|[OpenBMB/MiniCPM5-1B-Base](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B-Base)|llama|minicpm5|transformers>=4.40|✔|-|[openbmb/MiniCPM5-1B-Base](https://huggingface.co/openbmb/MiniCPM5-1B-Base)|
	|[OpenBMB/MiniCPM5-1B-SFT](https://modelscope.cn/models/OpenBMB/MiniCPM5-1B-SFT)|llama|minicpm5|transformers>=4.40|✔|-|[openbmb/MiniCPM5-1B-SFT](https://huggingface.co/openbmb/MiniCPM5-1B-SFT)|