# 介绍 MindSpore Transformers套件的目标是构建一个大模型训练、微调、评估、推理、部署的全流程开发套件： 提供业内主流的Transformer类预训练模型和SOTA下游任务应用，涵盖丰富的并行特性。期望帮助用户轻松的实现大模型训练和创新研发。 MindSpore Transformers套件基于MindSpore内置的并行技术和组件化设计，具备如下特点： - 一行代码实现从单卡到大规模集群训练的无缝切换； - 提供灵活易用的个性化并行配置； - 能够自动进行拓扑感知，高效地融合数据并行和模型并行策略； - 一键启动任意任务的单卡/多卡训练、微调、评估、推理流程； - 支持用户进行组件化配置任意模块，如优化器、学习策略、网络组装等； - 提供Trainer、pipeline、AutoClass等高阶易用性接口； - 提供预置SOTA权重自动下载及加载功能； - 支持人工智能计算中心无缝迁移部署； 如果您对MindSpore Transformers有任何建议，请通过issue与我们联系，我们将及时处理。 - **[MindFromers教程文档](https://mindformers.readthedocs.io/zh_CN/latest)** - [模型README](https://gitee.com/mindspore/mindformers/tree/dev/docs/model_cards) - [任务README](https://gitee.com/mindspore/mindformers/tree/dev/docs/task_cards) - [MindPet指导教程](docs/feature_cards/Pet_Tuners.md) - [AICC指导教程](docs/readthedocs/source_zh_cn/docs/practice/AICC.md) 目前支持的模型列表如下： | 模型 | 任务（task name） | 模型（model name） | | :------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | [BERT](docs/model_cards/bert.md) | masked_language_modeling
[text_classification](docs/task_cards/text_classification.md)
[token_classification](docs/task_cards/token_classification.md)
[question_answering](docs/task_cards/question_answering.md) | bert_base_uncased
txtcls_bert_base_uncased
txtcls_bert_base_uncased_mnli
tokcls_bert_base_chinese
tokcls_bert_base_chinese_cluener
qa_bert_base_uncased
qa_bert_base_chinese_uncased | | [T5](docs/model_cards/t5.md) | translation | t5_small | | [GPT2](docs/model_cards/gpt2.md) | [text_generation](docs/task_cards/text_generation.md) | gpt2_small
gpt2_13b
gpt2_52b | | [PanGuAlpha](docs/model_cards/pangualpha.md) | [text_generation](docs/task_cards/text_generation.md) | pangualpha_2_6_b
pangualpha_13b | | [GLM](docs/model_cards/glm.md) | [text_generation](docs/task_cards/text_generation.md) | glm_6b
glm_6b_lora | | [GLM2](docs/model_cards/glm2.md) | [text_generation](docs/task_cards/text_generation.md) | glm2_6b
glm2_6b_lora | | [LLama](docs/model_cards/llama.md) | [text_generation](docs/task_cards/text_generation.md) | llama_7b
llama_13b
llama_65b
llama_7b_lora | | [LLama2](docs/model_cards/llama.md) | [text_generation](docs/task_cards/text_generation.md) | llama2_7b
llama2_13b | | [Bloom](docs/model_cards/bloom.md) | [text_generation](docs/task_cards/text_generation.md) | bloom_560m
bloom_7.1b
bloom_65b
bloom_176b | | [MAE](docs/model_cards/mae.md) | masked_image_modeling | mae_vit_base_p16 | | [VIT](docs/model_cards/vit.md) | [image_classification](docs/task_cards/image_classification.md) | vit_base_p16 | | [Swin](docs/model_cards/swin.md) | [image_classification](docs/task_cards/image_classification.md) | swin_base_p4w7 | | [CLIP](docs/model_cards/clip.md) | [contrastive_language_image_pretrain](docs/task_cards/contrastive_language_image_pretrain.md)
[zero_shot_image_classification](docs/task_cards/zero_shot_image_classification.md) | clip_vit_b_32
clip_vit_b_16
clip_vit_l_14
clip_vit_l_14@336 | | [BLIP2](docs/model_cards/blip2.md) | [contrastive_language_image_pretrain](docs/task_cards/contrastive_language_image_pretrain.md)
[zero_shot_image_classification](docs/task_cards/zero_shot_image_classification.md) | blip2_stage1_vit_g | 目前在research中支持的模型列表如下： | 模型 | 任务（task name） | 模型（model name） | | :------------------------------------------: | :---------------------------------------------------: | :------------------------------ | | [Baichuan](research/baichuan/baichuan.md) | [text_generation](docs/task_cards/text_generation.md) | baichuan_7b
baichuan_13b | | [Baichuan2](research/baichuan2/baichuan2.md) | [text_generation](docs/task_cards/text_generation.md) | baichuan2_7b
baichuan2_13b | | [Internlm](research/internlm/internlm.md) | [text_generation](docs/task_cards/text_generation.md) | InternLM-7B | | [ziya](research/ziya/ziya.md) | [text_generation](docs/task_cards/text_generation.md) | ziya-13B |