Files
pythia410m-sft-tldr/code/configs/gptrm_sft_pythia7b_4V100_seq640_fp16_3epochs.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

13 lines
325 B
YAML

model_name: /home/toolkit/huggingface/tldr_sft_pythia7b_4V100_seq550
load_in_8bit: False
bf16: False
fp16: True
per_device_train_batch_size: 1
per_device_eval_batch_size: 2
gradient_accumulation_steps: 32
gradient_checkpointing: False
learning_rate: 1e-5
optimizer_type: paged_adamw_32bit
num_train_epochs: 3
seq_length: 640