Built-In Training Image List
Last updated: 2025-08-14 15:38:37
Overview
TI-ONE includes mainstream deep learning and machine learning frameworks such as TensorFlow, Pytorch, and PySpark. It also introduced the self-developed training acceleration framework Angel. Among them, tilearn-llm is a training acceleration component customized for large models, built into the platform's common training image, and supports both Notebook and task-based modeling.
Built-In Image List
Framework | Image Name | Supported Training Mode | Remarks |
PyTorch | tilearn-llm0.9-torch2.3-py3.10-cuda12.4-gpu | DDP | Supported core libraries: Python 3.10, CUDA 12.4, jupyterlab 2.3.2, torch 2.3.0a0+40ec155e58.nv24.3, transformers 4.39.3, deepspeed 0.13.4, tilearn-llm 0.9.9, tilearn.ops 0.2.2.175, angel-vllm 0.4.2 Supported GPU types: PNV6, H100, H800, PNV5b, L40, A100, A800, A10, V100, T4 Supported modules: Task-Based Modeling and Development Machine |
| tilearn-llm0.9-torch2.1-py3.10-cuda12.1-gpu | DDP | Supported core libraries: Python 3.10, CUDA 12.1, jupyterlab 2.3.2, torch 2.1.2, transformers 4.39.3, deepspeed 0.14.0, tilearn-llm 0.9.3.3, tilearn.ops 0.2.1.172, angel-vllm 0.3.4 Supported GPU types: H100, H800, PNV5b, L40, A100, A800, A10, V100, T4 Supported modules: Task-Based Modeling and Development Machine |
| tilearn-llm0.8-torch2.1-py3.10-cuda12.1-gpu | DDP | Supported core libraries: Python 3.10, CUDA 12.1, jupyterlab 2.3.2, torch 2.1.0a0+b5021ba, transformers 4.31.0, deepspeed 0.10.0, tilearn-llm 0.8.3, tilearn.ops 0.2.0.1 Supported GPU types: H100, H800, PNV5b, L40, A100, A800, A10, V100, T4 Supported modules: Task-Based Modeling and Development Machine |
| ti-acc2.5-torch1.9-py3.8-cuda11.1-gpu | DDP | Supported core libraries: Python 3.8, CUDA 11.1, torch 1.9.0+cu111, tiacc-training.torch 2.5.1.dev13 Supported GPU types: A100, A10, V100, T4 Supported modules: Task-Based Modeling |
| torch1.9-py3.8-cuda11.1-gpu | DDP,MPI,Horovod | Supported core libraries: Python 3.8, CUDA 11.1, torch 1.9.0+cu111 Supported GPU types: A100, A10, V100, T4 Supported modules: Task-Based Modeling |
| ti-acc2.5-torch1.12-tf1.15-tf2.4-pyspark2.4.5-py3.8-cuda11.3-gpu | - | Supported core libraries: jupyterlab 3.6.1, multiple conda environments, including: - pyspark: Python 3.7, spark-2.4.5-bin-hadoop2.7 - pyspark3: Python 3.8, spark-3.3.1-bin-hadoop3 - pytorch_py3: Python 3.8, CUDA 11.1, torch 1.9.0+cu111 - tiacc_pytorch_py3: Python 3.8, CUDA 11.1, torch 1.12.1+cu113, tiacc-training.torch 2.5.1.dev10 - tf_py3: Python 3.7, CUDA 10.0, tensorflow-gpu 1.15.0 - tiacc_tf_py3: Python 3.7, CUDA 10.0, tensorflow-gpu 1.15.0 - tf2_py3: Python 3.8, CUDA 11.0, tensorflow 2.4.0 Supported GPU types: V100, T4, partial kernel support for A100/A10 Supported modules: Development Machine |
TensorFlow | ti-acc1.0-tf1.15-py3.6-cuda10.0-gpu | PS-Worker | - |
| tf1.15-py3.7-cpu | PS-Worker,MPI,Horovod | - |
| tf1.15-py3.7-cuda10.0-gpu | | |
| tf2.4-py3.8-cpu | | |
| tf2.4-py3.8-cuda11.1-gpu | | |
Spark | spark2.4.5-cpu | Spark | - |
PySpark | spark2.4.5-py3.6-cpu | Spark | - |
Others | py3.8-cpu | - | Miniforge3, jupyterlab 4.3, tensorboard 2.18.0, Python 3.8 conda environment Supported modules: Development Machine |
| py3.10-cpu | - | Miniforge3, jupyterlab 4.3, tensorboard 2.18.0, Python 3.10 conda environment Supported modules: Development Machine |
Was this page helpful?
You can also Contact Sales or Submit a Ticket for help.
Yes
No
Feedback