Platform For AI:LLM

Get started

Model Gallery offers a range of pre-trained large language models (LLMs) from the open-source community. In Model Gallery, you can perform model training (fine-tuning), distillation, compression, evaluation, and deployment without coding. Use cases:

• QuickStart: Deployment, fine-tuning, and evaluation of QwQ-32B

• Fine-tune, evaluate, and deploy a Qwen2.5 model

• Fine-tune, evaluate, compress, and deploy a Qwen2.5-Coder model

• Fine-tune, evaluate, compress, and deploy a DistilQwen2 model

• Deploy and fine-tune a Llama-3 model

• Deploy and fine-tune Qwen 1.5 models

• Deploy and fine-tune a Mixtral-8x7B MoE model

• Use QuickStart to fine-tune and deploy Llama 2 models

• LLM continued pre-training

• Develop a data augmentation and model distillation solution for LLMs

• Best practices for LLM evaluation

Advanced usage

If Model Gallery does not include the model you require, or if the training and deployment capabilities do not meet your needs, consider using Data Science Workshop (DSW) and Deep Learning Containers (DLC) for model fine-tuning and training. Then, deploy and run your model in Elastic Algorithm Service (EAS).

Lingjun intelligent computing service

Lingjun intelligent computing service is tailored for large-scale deep learning scenarios. It offers a comprehensive heterogeneous computing resource and AI engineering platform. Get started with Lingjun through the following use case:

• Develop Qwen models in PAI-Lingjun AI Computing Service

Data processing

Machine Learning Designer integrates a variety of advanced algorithms for processing text, video, and image data to improve the quality of training data. Text processing algorithms allow you to edit, transform, and deduplicate data samples, as well as filter out low-quality data samples. Video and image processing algorithms allow you to implement features such as data cleaning, content filtering, metadata extraction, and caption generation. In addition, the system provides a set of commonly used built-in data processing templates. You can select templates based on your business requirements and flexibly extend the templates through secondary development to meet the requirements of specific scenarios.

• Data Processing for LLM (Github Code)

• Data processing for LLM (web text data from Wikipedia)

• Data processing for LLM (thesis data from arXiv)

• Data processing for LLM (SFT data from Alpaca-CoT)

• LLM Data Processing-Alpaca-Cot (SFT Data)-DLC

• Data Processing for LLM (Github Code)-DLC

• E2E Development and Usage of LLM-DLC: Data Processing + Model Training + Model Inference

• E2E Development and Usage of LLM: Data Processing + Model Training + Model Inference

• Image-Text Filtering

• Video data filtering and labeling

More solutions

• Develop an LLM-based intent recognition solution