intelligent-machine-learning / intelligent-machine-learning/dlrover
while using megatron distributed flash-checkpoint to recovery, error ocurs when load_checkpoint
- Dominant language
- Python
- Stars
- 1.7k
- Forks
- 219
- Avg merge
- 10h 11m
- Merged PRs (30d)
- 11
Description
**Env:** 16GPUs + llama2 pretrain+ megatron-lm
**strategy:** TP 8 + PP 1 + DP 2
**case:** when killing a training proceess to retrigger fault-tollerence with megatron-distributed flash-checkpoint,the dp 1 group load_checkpoint failed with the following log,
```
WARNING: on rank 11 found iteration 15 in the metadata while max iteration across the ranks is 4160813071, replacing it with max iteration.WARNING: on rank 10 found iteration 15 in the metadata while max iteration across the ranks is 4160813071, replacing it with max iteration.WARNING: on rank 14 found iteration 15 in the metadata while max iteration across the ranks is 4160813071, replacing it with max iteration.
```
The reason is that dp 1 group load checkpoint from storage for no model in memory and uses allreduce when **read_metadata**, meanwhile dp 0 group only load from memory.
Contributor guide
Research direction
Start at the flash-checkpoint load_checkpoint path and its read_metadata operation, using the reported 16-GPU TP 8, PP 1, DP 2 setup and rank logs to reproduce the failure. Trace the differing storage-versus-memory loading paths and their allreduce behavior; done means fault-tolerant recovery completes with consistent checkpoint metadata across both data-parallel groups.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100