mirror of
https://github.com/hanzoai/open-instruct.git
synced 2026-11-02 13:38:22 +00:00
Costa HuangandGitHub
28be6c8402
* use the latest oe-eval-image * use the latest image * rename * do not use slow tokenizer
26 lines
693 B
YAML
26 lines
693 B
YAML
model_name_or_path: allenai/OLMo-2-1124-13B
|
|
use_flash_attn: true
|
|
tokenizer_name: allenai/OLMo-2-1124-13B
|
|
use_slow_tokenizer: false
|
|
dataset_mixer:
|
|
allenai/tulu-3-sft-olmo-2-mixture: 1.0
|
|
|
|
max_seq_length: 4096
|
|
preprocessing_num_workers: 128
|
|
per_device_train_batch_size: 1 # note, this is set up for 8 GPUs
|
|
gradient_accumulation_steps: 4 # effective batch size 128 with 4 nodes
|
|
learning_rate: 6e-06
|
|
lr_scheduler_type: linear
|
|
warmup_ratio: 0.03
|
|
weight_decay: 0.0
|
|
num_train_epochs: 2
|
|
output_dir: /output/
|
|
with_tracking: true
|
|
report_to:
|
|
- wandb
|
|
logging_steps: 1
|
|
reduce_loss: sum
|
|
checkpointing_steps: epoch
|
|
dataset_mix_dir: /output/
|
|
hf_metadata_dataset: allenai/olmo-instruct-evals
|
|
add_bos: true |