Summary:
Currently, arrow iterator can only read arrow files. However, the pyarrow library can read
other formats, including jsonlines. This allows the same ArrowIterator to read from jsonlines,
so we can read from the original source data, and simply omit the entropy column when doing so
Test Plan:
Run train script until dataloader starts
Summary:
Currently, arrow iterator can only read arrow files. However, the pyarrow library can read
other formats, including jsonlines. This allows the same ArrowIterator to read from jsonlines,
so we can read from the original source data, and simply omit the entropy column when doing so
Test Plan:
Run train script until dataloader starts
Summary:
- Make the data/checkpoint code fsspec compatible
- Still will not work with s3 saves, due to `torch.distributed.checkpoint.save` not being out of the box workable with `fsspec`. Will implement in followup PR
Test Plan:
Run unit tests and the commands below
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100
```
These currently won't work due to the torch distributed save, but theses hould be tested at a later date
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100 dump_dir=s3://blt/scratch/checkpoint-test/
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100 dump_dir=s3://blt/scratch/checkpoint-test/
```
Summary:
- Make the data/checkpoint code fsspec compatible
- Still will not work with s3 saves, due to `torch.distributed.checkpoint.save` not being out of the box workable with `fsspec`. Will implement in followup PR
Test Plan:
Run unit tests and the commands below
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100
```
These currently won't work due to the torch distributed save, but theses hould be tested at a later date
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100 dump_dir=s3://blt/scratch/checkpoint-test/
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100 dump_dir=s3://blt/scratch/checkpoint-test/
```
Summary:
- Make the data/checkpoint code fsspec compatible
- Still will not work with s3 saves, due to `torch.distributed.checkpoint.save` not being out of the box workable with `fsspec`. Will implement in followup PR
Test Plan:
Run unit tests and the commands below
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100
```
These currently won't work due to the torch distributed save, but theses hould be tested at a later date
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100 dump_dir=s3://blt/scratch/checkpoint-test/
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100 dump_dir=s3://blt/scratch/checkpoint-test/
```
Summary:
With >1 GPU, but only 1 node, all reduces fail when inputs are not bf16. This uses a modified copy of torch's grad norm to avoid failures
Test Plan:
- Run unit tests:
- Run single gpu training: `python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100`
- Run 1 node, multi-gpu training `torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100`
Summary:
With >1 GPU, but only 1 node, all reduces fail when inputs are not bf16. This uses a modified copy of torch's grad norm to avoid failures
Test Plan:
- Run unit tests:
- Run single gpu training: `python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100`
- Run 1 node, multi-gpu training `torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100`
Summary:
With >1 GPU, but only 1 node, all reduces fail when inputs are not bf16. This uses a modified copy of torch's grad norm to avoid failures
Test Plan:
- Run unit tests:
- Run single gpu training: `python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100`
- Run 1 node, multi-gpu training `torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100`
Summary:
- Make arrow iterator able to read from jsonl files, the entropies are omitted in this case
- Make the data/checkpoint code fsspec compatible
- Fix issues with all reduce with non-bf16 in dist_sum and norm computation.
- Minimal fixes to get eval to run, it is slow currently
- Add bpb numbers during training
Test Plan:
Run unit tests and the commands below
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100
```
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100 dump_dir=s3://blt/scratch/checkpoint-test/
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null checkpoint.dump.every=100 dump_dir=s3://blt/scratch/checkpoint-test/
```
Summary:
- Make arrow iterator able to read from jsonl files, the entropies are omitted in this case
- Make the data/checkpoint code fsspec compatible
- Fix issues with all reduce with non-bf16 in dist_sum and norm computation.
- Minimal fixes to get eval to run, it is slow currently
- Add bpb numbers during training
Test Plan:
Run
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/entropy_model.yaml eval=null max_steps=10100
```
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null
```
Summary:
- Make arrow iterator able to read from jsonl files, the entropies are omitted in this case
- Make the data/checkpoint code fsspec compatible
- Fix issues with all reduce with non-bf16 in dist_sum and norm computation.
- Minimal fixes to get eval to run, it is slow currently
- Add bpb numbers during training
Test Plan:
Run
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/entropy_model.yaml eval=null max_steps=10100
```
```
python -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null
```
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/s3_debug.yaml eval=null
```
Summary:
- Make arrow iterator able to read from jsonl files, the entropies are omitted in this case
- Make the data/checkpoint code fsspec compatible
- Fix issues with all reduce with non-bf16 in dist_sum and norm computation.
- Minimal fixes to get eval to run, it is slow currently
- Add bpb numbers during training
Test Plan:
Run
```
torchrun --nproc-per-node 8 -m bytelatent.train config=internal/configs/entropy_model.yaml eval=null max_steps=10100
```
Summary:
- Make arrow iterator able to read from jsonl files, the entropies are omitted in this case
- Make the data/checkpoint code fsspec compatible
- Fix issues with all reduce with non-bf16 in dist_sum and norm computation.
- Minimal fixes to get eval to run, it is slow currently
- Add bpb numbers during training
Test Plan:
