mirror of
https://github.com/hanzoai/open-instruct.git
synced 2026-11-03 02:13:07 +00:00
56 lines
1.7 KiB
YAML
56 lines
1.7 KiB
YAML
model_name_or_path: meta-llama/Llama-3.1-8B
|
|
model_revision: main
|
|
use_flash_attn: true
|
|
tokenizer_name: meta-llama/Llama-3.1-8B
|
|
use_slow_tokenizer: true
|
|
dataset_mixer:
|
|
# General datasets:
|
|
ai2-adapt-dev/oasst1_converted: 7132
|
|
ai2-adapt-dev/tulu_hard_coded_repeated_10: 240
|
|
ai2-adapt-dev/wildchat_gpt4_converted: 100000
|
|
|
|
# Math datasets:
|
|
ai2-adapt-dev/personahub_math_v5_regen_149960: 149960
|
|
ai2-adapt-dev/personahub_grade_math_v1_49980: 49980
|
|
ai2-adapt-dev/open_math_2_gsm8k_converted: 50000
|
|
ai2-adapt-dev/personahub_math_interm_algebra_50000: 20000
|
|
|
|
# Coding datasets:
|
|
ai2-adapt-dev/personahub_code_v2_34999: 34999
|
|
ai2-adapt-dev/evol_codealpaca_heval_decontaminated: 107276
|
|
|
|
# IF datasets:
|
|
ai2-adapt-dev/personahub_ifdata_manual_seed_v3_29980: 29980
|
|
|
|
# Safety datasets:
|
|
ai2-adapt-dev/coconot_converted: 10983
|
|
ai2-adapt-dev/processed-wildjailbreak: 50000
|
|
ai2-adapt-dev/synthetic-finalresp-wildguarmixtrain: 50000
|
|
|
|
# Specialty datasets:
|
|
ai2-adapt-dev/table_gpt_converted: 5000
|
|
ai2-adapt-dev/aya_dataset_converted: 100000
|
|
|
|
# Datasets removed because of licensing:
|
|
# ai2-adapt-dev/flan_v2_converted: 89982 # ODC-BY
|
|
# ai2-adapt-dev/sciriff_converted: 10000 # ODC-BY
|
|
# ai2-adapt-dev/no_robots_converted: 9500 # NC
|
|
# AI-MO/NuminaMath-TIR: 72441 # NC
|
|
|
|
max_seq_length: 8192
|
|
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: 5.0e-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
|
|
checkpointing_steps: epoch
|
|
dataset_mix_dir: /output/
|