Files
pythia410m-sft-tldr/code/configs/gptrm_sft_pythia1b_traincomp_fp16_4V100.yml
ModelHub XC 4bb3a5553f 初始化项目,由ModelHub XC社区提供模型
Model: mnoukhov/pythia410m-sft-tldr
Source: Original Platform
2026-05-30 14:21:21 +08:00

16 lines
407 B
YAML

model_name: /home/toolkit/trl/results/d31ad1f5a2ea94087cf49d8046228e62/code/results/checkpoint-1000
load_in_8bit: False
bf16: False
fp16: True
use_lora: False
per_device_train_batch_size: 4
per_device_eval_batch_size: 2
gradient_accumulation_steps: 8
gradient_checkpointing: False
learning_rate: 1e-6
optimizer_type: adamw_torch
lr_scheduler_type: linear
num_train_epochs: 3
seq_length: 560
eval_steps: 250