MindSpore Transformers (MindFormers)

1. Introduction

The goal of the MindSpore Transformers suite is to build a full-process development suite for Large model pre-training, fine-tuning, evaluation, inference, and deployment. It provides mainstream Transformer-based Large Language Models (LLMs) and Multimodal Models (MMs). It is expected to help users easily realize the full process of large model development.

Based on MindSpore's built-in parallel technology and component-based design, the MindSpore Transformers suite has the following features:

• One-click initiation of single or multi card pre-training, fine-tuning, evaluation, inference, and deployment processes for large models;
• Provides rich multi-dimensional hybrid parallel capabilities for flexible and easy-to-use personalized configuration;
• System-level deep optimization on large model training and inference, native support for ultra-large-scale cluster efficient training and inference, rapid fault recovery;
• Support for configurable development of task components. Any module can be enabled by unified configuration, including model network, optimizer, learning rate policy, etc.;
• Provide real-time visualization of training accuracy/performance monitoring indicators.

For details about MindSpore Transformers tutorials and API documents, see MindSpore Transformers Documentation. The following are quick jump links to some of the key content:

• 📝 Pre-training
• 📝 Supervised Fine-Tuning
• 📝 Evaluation
• 📝 Service-oriented Deployment

If you have any suggestions on MindSpore Transformers, contact us through an issue, and we will address it promptly.

If you're interested in MindSpore Transformers technology or wish to contribute code, we welcome you to join the MindSpore Transformers SIG.

Models List

The following table lists models supported by MindSpore Transformers.

Model	Specifications	Model Type	Model Architecture	Latest Version
Qwen3	0.6B/1.7B/4B/8B/14B/32B	Dense LLM	Mcore	In-development version
Qwen3-MoE	30B-A3B/235B-A22B	Sparse LLM	Mcore	In-development version
DeepSeek-V3	671B	Sparse LLM	Legacy	1.6.0, In-development version
GLM4	9B	Dense LLM	Legacy	1.6.0, In-development version
Llama3.1	8B/70B	Dense LLM	Legacy	1.6.0, In-development version
Mixtral	8x7B	Sparse LLM	Legacy	1.6.0, In-development version
Qwen2.5	0.5B/1.5B/7B/14B/32B/72B	Dense LLM	Legacy	1.6.0, In-development version
TeleChat2	7B/35B/115B	Dense LLM	Legacy	1.6.0, In-development version
CodeLlama	34B	Dense LLM	Legacy	1.5.0
CogVLM2-Image	19B	MM	Legacy	1.5.0
CogVLM2-Video	13B	MM	Legacy	1.5.0
DeepSeek-V2	236B	Sparse LLM	Legacy	1.5.0
DeepSeek-Coder-V1.5	7B	Dense LLM	Legacy	1.5.0
DeepSeek-Coder	33B	Dense LLM	Legacy	1.5.0
GLM3-32K	6B	Dense LLM	Legacy	1.5.0
GLM3	6B	Dense LLM	Legacy	1.5.0
InternLM2	7B/20B	Dense LLM	Legacy	1.5.0
Llama3.2	3B	Dense LLM	Legacy	1.5.0
Llama3.2-Vision	11B	MM	Legacy	1.5.0
Llama3	8B/70B	Dense LLM	Legacy	1.5.0
Llama2	7B/13B/70B	Dense LLM	Legacy	1.5.0
Qwen2	0.5B/1.5B/7B/57B/57B-A14B/72B	Dense /Sparse LLM	Legacy	1.5.0
Qwen1.5	7B/14B/72B	Dense LLM	Legacy	1.5.0
Qwen-VL	9.6B	MM	Legacy	1.5.0
TeleChat	7B/12B/52B	Dense LLM	Legacy	1.5.0
Whisper	1.5B	MM	Legacy	1.5.0
Yi	6B/34B	Dense LLM	Legacy	1.5.0
YiZhao	12B	Dense LLM	Legacy	1.5.0
Baichuan2	7B/13B	Dense LLM	Legacy	1.3.2
GLM2	6B	Dense LLM	Legacy	1.3.2
GPT2	124M/13B	Dense LLM	Legacy	1.3.2
InternLM	7B/20B	Dense LLM	Legacy	1.3.2
Qwen	7B/14B	Dense LLM	Legacy	1.3.2
CodeGeex2	6B	Dense LLM	Legacy	1.1.0
WizardCoder	15B	Dense LLM	Legacy	1.1.0
Baichuan	7B/13B	Dense LLM	Legacy	1.0
Blip2	8.1B	MM	Legacy	1.0
Bloom	560M/7.1B/65B/176B	Dense LLM	Legacy	1.0
Clip	149M/428M	MM	Legacy	1.0
CodeGeex	13B	Dense LLM	Legacy	1.0
GLM	6B	Dense LLM	Legacy	1.0
iFlytekSpark	13B	Dense LLM	Legacy	1.0
Llama	7B/13B	Dense LLM	Legacy	1.0
MAE	86M	MM	Legacy	1.0
Mengzi3	13B	Dense LLM	Legacy	1.0
PanguAlpha	2.6B/13B	Dense LLM	Legacy	1.0
SAM	91M/308M/636M	MM	Legacy	1.0
Skywork	13B	Dense LLM	Legacy	1.0
Swin	88M	MM	Legacy	1.0
T5	14M/60M	Dense LLM	Legacy	1.0
VisualGLM	6B	MM	Legacy	1.0
Ziya	13B	Dense LLM	Legacy	1.0
Bert	4M/110M	Dense LLM	Legacy	0.8

indicates that the model has been offline from the main branch and can be used with the latest supported version.

The model maintenance strategy follows the Life Cycle And Version Matching Strategy of the corresponding latest supported version.

Model Level Introduction

The Mcore architecture model is divided into five levels for training and inference, respectively, representing different standards for model deployment. For details on the levels of different specifications of models in the library, please refer to the model documentation.

Training

• Released: Passed testing team verification, with loss and grad norm accuracy meeting benchmark alignment standards under deterministic conditions;
• Validated: Passed self-verification by the development team, with loss and grad norm accuracy meeting benchmark alignment standards under deterministic conditions;
• Preliminary: Passed preliminary self-verification by developers, with complete functionality and usability, normal convergence of training, but accuracy not strictly verified;
• Untested: Functionality is available but has not undergone systematic testing, with accuracy and convergence not verified, and support for user-defined development enablement;
• Community: Community-contributed MindSpore native models, developed and maintained by the community.

Inference

• Released: Passed testing team acceptance, with evaluation accuracy aligned with benchmark standards;
• Validated: Passed developer self-verification, with evaluation accuracy aligned with benchmark standards;
• Preliminary: Passed preliminary self-verification by developers, with complete functionality and usable for testing; inference outputs are logically consistent but accuracy has not been strictly verified;
• Untested: Functionality is available but has not undergone system testing; accuracy has not been verified; supports user-defined development enablement;
• Community: Community-contributed MindSpore native models, developed and maintained by the community.

2. Installation

Version Mapping

Currently supported hardware includes Atlas 800T A2, Atlas 800I A2, and Atlas 900 A3 SuperPoD.

Python 3.11.4 is recommended for the current suite.

MindSpore Transformers	MindSpore	CANN	Driver/Firmware
In-development version	In-development version	In-development version	In-development version

Historical Version Supporting Relationships:

MindSpore Transformers	MindSpore	CANN	Driver/Firmware
1.6.0	2.7.0	8.2.RC1	25.2.0
1.5.0	2.6.0-rc1	8.1.RC1	25.0.RC1
1.3.2	2.4.10	8.0.0	24.1.0
1.3.0	2.4.0	8.0.RC3	24.1.RC3
1.2.0	2.3.0	8.0.RC2	24.1.RC2

Installation Using the Source Code

Currently, MindSpore Transformers can be compiled and installed using the source code. You can run the following commands to install MindSpore Transformers:

git clone -b master https://gitee.com/mindspore/mindformers.git
cd mindformers
bash build.sh

3. User Guide

MindSpore Transformers supports distributed pre-training, supervised fine-tuning, and inference tasks for large models with one click. You can click the link of each model in Model List to see the corresponding documentation.

For more information about the functions of MindSpore Transformers, please refer to MindSpore Transformers Documentation.

4. Life Cycle And Version Matching Strategy

MindSpore Transformers version has the following five maintenance phases:

Status	Duration	Description
Plan	1-3 months	Planning function.
Develop	3 months	Build function.
Preserve	6 months	Incorporate all solved problems and release new versions.
No Preserve	0—3 months	Incorporate all the solved problems, there is no full-time maintenance team, and there is no plan to release a new version.
End of Life (EOL)	N/A	The branch is closed and no longer accepts any modifications.

MindSpore Transformers released version preservation policy:

MindSpore Transformers Version	Corresponding Label	Current Status	Release Time	Subsequent Status	EOL Date
1.7.0	v1.7.0	Preserve	2025/10/27	No preserve expected from 2025/04/27	2026/07/27
1.6.0	v1.6.0	Preserve	2025/07/29	No preserve expected from 2025/01/29	2026/04/29
1.5.0	v1.5.0	No Preserve	2025/04/29	End of Life expected from 2026/01/29	2026/01/29
1.3.2	v1.3.2	End of Life	2024/12/20	-	2025/09/20
1.2.0	v1.2.0	End of Life	2024/07/12	-	2025/04/12
1.1.0	v1.1.0	End of Life	2024/04/15	-	2025/01/15

5. Disclaimer

1. scripts/examples directory are provided as reference examples and do not form part of the commercially released products. They are only for users' reference. If it needs to be used, the user should be responsible for transforming it into a product suitable for commercial use and ensuring security protection. MindSpore Transformers does not assume security responsibility for the resulting security problems.
2. Regarding datasets, MindSpore Transformers only provides suggestions for datasets that can be used for training. MindSpore Transformers does not provide any datasets. Users who use any dataset for training must ensure the legality and security of the training data and assume the following risks:
   1. Data poisoning: Maliciously tampered training data may cause the model to produce bias, security vulnerabilities, or incorrect outputs.
   2. Data compliance: Users must ensure that data collection and processing comply with relevant laws, regulations, and privacy protection requirements.
3. If you do not want your dataset to be mentioned in MindSpore Transformers, or if you want to update the description of your dataset in MindSpore Transformers, please submit an issue to Gitee, and we will remove or update the description of your dataset according to your issue request. We sincerely appreciate your understanding and contribution to MindSpore Transformers.
4. Regarding model weights, users must verify the authenticity of downloaded and distributed model weights from trusted sources. MindSpore Transformers cannot guarantee the security of third-party weights. Weight files may be tampered with during transmission or loading, leading to unexpected model outputs or security vulnerabilities. Users should assume the risk of using third-party weights and ensure that weight files are verified for security before use.
5. Regarding weights, vocabularies, scripts, and other files downloaded from sources like openmind, users must verify the authenticity of downloaded and distributed model weights from trusted sources. MindSpore Transformers cannot guarantee the security of third-party files. Users should assume the risks arising from unexpected functional issues, outputs, or security vulnerabilities when using these files.
6. MindSpore Transformers saves weights or logs based on the path set by the user. Users should avoid using system file directories when configuring paths. If unexpected system issues arise due to improper path settings, users shall bear the risks themselves.

6. Contribution

We welcome contributions to the community. For details, see MindSpore Transformers Contribution Guidelines.

7. License

Apache 2.0 License