Skip to content

Commit aadf2f2

Browse files
committed
feat: Implement long context extension with yarn
1 parent b856127 commit aadf2f2

11 files changed

Lines changed: 964 additions & 11 deletions

config_files/training/config_example_coca.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,7 @@ settings:
1111
world_size: ${cuda_env:WORLD_SIZE}
1212
paths:
1313
checkpoint_saving_path: data/checkpoints
14+
experiments_root_path: ${modalities_env:experiments_root_path}
1415
train_dataset_path: ./data/lorem_ipsum.pbin
1516
intervals:
1617
training_log_interval_in_steps: 2

config_files/training/config_lorem_ipsum_long_fsdp1.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,7 @@ settings:
1111
world_size: ${cuda_env:WORLD_SIZE}
1212
paths:
1313
checkpoint_saving_path: data/checkpoints
14+
experiments_root_path: ${modalities_env:experiments_root_path}
1415
train_dataset_path: ./data/lorem_ipsum_long.pbin
1516
test_dataset_path: ./data/lorem_ipsum.pbin
1617
intervals:

config_files/training/config_lorem_ipsum_long_fsdp1_warmstart.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,7 @@ settings:
1111
world_size: ${cuda_env:WORLD_SIZE}
1212
paths:
1313
checkpoint_saving_path: data/checkpoints
14+
experiments_root_path: ${modalities_env:experiments_root_path}
1415
train_dataset_path: ./data/lorem_ipsum_long.pbin
1516
test_dataset_path: ./data/lorem_ipsum.pbin
1617
intervals:

config_files/training/config_lorem_ipsum_long_fsdp2.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,7 @@ settings:
1111
world_size: ${cuda_env:WORLD_SIZE}
1212
paths:
1313
checkpoint_saving_path: data/checkpoints
14+
experiments_root_path: ${modalities_env:experiments_root_path}
1415
train_dataset_path: ./data/lorem_ipsum_long.pbin
1516
test_dataset_path: ./data/lorem_ipsum.pbin
1617
intervals:

config_files/training/config_lorem_ipsum_long_fsdp2_pp.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,7 @@ settings:
1111
world_size: ${cuda_env:WORLD_SIZE}
1212
paths:
1313
checkpoint_saving_path: data/checkpoints
14+
experiments_root_path: ${modalities_env:experiments_root_path}
1415
train_dataset_path: ./data/lorem_ipsum_long.pbin
1516
test_dataset_path: ./data/lorem_ipsum.pbin
1617
intervals:

config_files/training/config_lorem_ipsum_long_fsdp2_pp_tp.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,7 @@ settings:
1111
world_size: ${cuda_env:WORLD_SIZE}
1212
paths:
1313
checkpoint_saving_path: data/checkpoints
14+
experiments_root_path: ${modalities_env:experiments_root_path}
1415
train_dataset_path: ./data/lorem_ipsum_long.pbin
1516
test_dataset_path: ./data/lorem_ipsum.pbin
1617
intervals:

config_files/training/config_lorem_ipsum_long_fsdp2_warmstart.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,7 @@ settings:
1111
world_size: ${cuda_env:WORLD_SIZE}
1212
paths:
1313
checkpoint_saving_path: data/checkpoints
14+
experiments_root_path: ${modalities_env:experiments_root_path}
1415
train_dataset_path: ./data/lorem_ipsum_long.pbin
1516
test_dataset_path: ./data/lorem_ipsum.pbin
1617
intervals:

0 commit comments

Comments
 (0)