7b68dfafd7
* jagged lr restart scheudler var name fix make sure to create scheduler first * wire things together * more fixes * fix for nesting scheduler and first anneal phase * no need for relora trainer anymore since we've generalized the relora scheduler * remove redundant relora scheduler and lint * update relora e2e test for updated params * need restart steps for relora test * update quarto docs for dropped relora trainer * update example yaml * drop verbose arg * min lr scale support for jagged lr * don't let min_lr be nonetype * cleanup args
66 lines
1.1 KiB
YAML
66 lines
1.1 KiB
YAML
base_model: NousResearch/Llama-2-7b-hf
|
|
model_type: LlamaForCausalLM
|
|
tokenizer_type: LlamaTokenizer
|
|
|
|
|
|
load_in_8bit: false
|
|
load_in_4bit: true
|
|
|
|
datasets:
|
|
- path: teknium/GPT4-LLM-Cleaned
|
|
type: alpaca
|
|
dataset_prepared_path:
|
|
val_set_size: 0.05
|
|
output_dir: ./outputs/relora-out
|
|
|
|
adapter: qlora
|
|
lora_model_dir:
|
|
|
|
sequence_len: 4096
|
|
sample_packing: true
|
|
|
|
|
|
lora_r: 8
|
|
lora_alpha: 16
|
|
lora_dropout: 0.05
|
|
lora_target_linear: true
|
|
|
|
relora: true
|
|
relora_prune_ratio: 0.9
|
|
relora_cpu_offload: false
|
|
jagged_restart_steps: 150
|
|
jagged_restart_warmup_steps: 10
|
|
jagged_restart_anneal_steps: false
|
|
|
|
wandb_project:
|
|
wandb_entity:
|
|
wandb_watch:
|
|
wandb_name:
|
|
wandb_log_model:
|
|
|
|
gradient_accumulation_steps: 4
|
|
micro_batch_size: 4
|
|
num_epochs: 4
|
|
optimizer: adamw_bnb_8bit
|
|
lr_scheduler: cosine
|
|
learning_rate: 0.0002
|
|
|
|
bf16: auto
|
|
tf32: false
|
|
|
|
gradient_checkpointing: true
|
|
resume_from_checkpoint:
|
|
logging_steps: 1
|
|
flash_attention: true
|
|
|
|
warmup_ratio: 0.1
|
|
evals_per_epoch: 4
|
|
saves_per_epoch: 1
|
|
weight_decay: 0.0
|
|
special_tokens:
|
|
bos_token: "<s>"
|
|
eos_token: "</s>"
|
|
unk_token: "<unk>"
|
|
|
|
# save_first_step: true # uncomment this to validate checkpoint saving works with your config
|