training checkpoints

Lists out the Checkpoints that have been generated for a Training Job.

Checkpoints are generated by the FlexAI runtime when a Training script’s code calls the torch.save() function 🔗.

Note

Visit the FlexAI Managed Checkpoints reference page to learn more about how FlexAI Managed Checkpoints work.

flexai training checkpoints <training_job_name>

Arguments

training_job_name

<string>

Required

Resource Name

The name of the Training Job to list checkpoints for.

Examples

• gpt2training-1

Flags

--json

<boolean>

Optional

Flag

Output the information in JSON format.

Examples

• --json

Example

flexai training checkpoints gpt2training-1

Which will output:

flexai training checkpoints gpt2training-1

 ID                                   │ NAME          │ NODE │ STEP │ TRAIN LOSS │ EVAL LOSS │ MODEL           │ VERSION │ INFERENCE READY │ TIMESTAMP
──────────────────────────────────────┼───────────────┼──────┼──────┼────────────┼───────────┼─────────────────┼─────────┼─────────────────┼───────────────────────────────────
 ce4d8def-b6bf-4cc6-8067-de9d312a82c5 │ checkpoint-50 │ 0    │ 50   │ 3.3707     │ 3.1356    │ GPT2LMHeadModel │ 4.53.2  │ true            │ 2025-07-15 09:30:10.549 +0000 UTC
 cabf2516-59f5-4ce7-8d14-220a8ca57ba7 │ checkpoint-99 │ 0    │ 90   │ 3.2857     │ 3.1356    │ GPT2LMHeadModel │ 4.53.2  │ true            │ 2025-07-15 09:30:51.46 +0000 UTC
 c26d1452-6d41-4b23-805e-30eeeadca729 │               │ 0    │ 99   │ 3.2857     │ 3.1356    │ GPT2LMHeadModel │ 4.53.2  │ true            │ 2025-07-15 09:30:51.459 +0000 UTC

Output Columns

Column	Description
ID	The unique identifier of the checkpoint.
NAME	The human-readable name of the checkpoint.
NODE	The Node where the checkpoint was created.
STEP	The training step at which the checkpoint was created.
TRAIN LOSS	The training loss at the time the checkpoint was created.
EVAL LOSS	The evaluation loss at the time the checkpoint was created.
MODEL	The name of the base model used in the checkpoint.
VERSION	The version of the model used in the checkpoint, such as 4.53.2.
INFERENCE READY	Indicates whether the checkpoint is ready for inference or not. Meaning it includes all necessary files and metadata required for inference to tasks.
TIMESTAMP	The ISO 8601 formatted timestamp of when the checkpoint was created.