2023-04-07 21:53:07 +00:00
|
|
|
# model/tokenizer
|
2023-04-08 20:33:02 +00:00
|
|
|
model_name: "EleutherAI/gpt-j-6b"
|
|
|
|
|
tokenizer_name: "EleutherAI/gpt-j-6b"
|
2023-04-07 21:53:07 +00:00
|
|
|
gradient_checkpointing: false
|
2023-04-12 03:50:54 +00:00
|
|
|
save_name: # CHANGE
|
2023-04-07 21:53:07 +00:00
|
|
|
|
|
|
|
|
# dataset
|
|
|
|
|
streaming: false
|
|
|
|
|
num_proc: 64
|
2023-04-12 03:50:54 +00:00
|
|
|
dataset_path: # CHANGE
|
2023-04-07 21:53:07 +00:00
|
|
|
max_length: 1024
|
2023-04-08 21:47:07 +00:00
|
|
|
batch_size: 1
|
2023-04-07 21:53:07 +00:00
|
|
|
|
|
|
|
|
# train dynamics
|
|
|
|
|
lr: 2.0e-5
|
|
|
|
|
min_lr: 0
|
|
|
|
|
weight_decay: 0.0
|
|
|
|
|
eval_every: 500
|
|
|
|
|
eval_steps: 105
|
|
|
|
|
save_every: 500
|
|
|
|
|
log_grads_every: 500
|
2023-04-12 03:50:54 +00:00
|
|
|
output_dir: # CHANGE
|
2023-04-07 21:53:07 +00:00
|
|
|
checkpoint: null
|
|
|
|
|
lora: true
|
|
|
|
|
warmup_steps: 500
|
|
|
|
|
num_epochs: 2
|
|
|
|
|
|
|
|
|
|
# logging
|
|
|
|
|
wandb: true
|
2023-04-12 03:50:54 +00:00
|
|
|
wandb_entity: # CHANGE
|
|
|
|
|
wandb_project_name: # CHANGE
|
2023-04-07 21:53:07 +00:00
|
|
|
seed: 42
|
|
|
|
|
|
