22810c97b7
* use warmup_ratio as a better default than warmup steps since it's data dependent * replace remainder of warmup_steps
52 lines
993 B
YAML
52 lines
993 B
YAML
base_model: cerebras/Cerebras-GPT-1.3B
|
|
# Automatically upload checkpoint and final model to HF
|
|
# hub_model_id: username/custom_model_name
|
|
|
|
load_in_8bit: false
|
|
load_in_4bit: true
|
|
push_dataset_to_hub:
|
|
datasets:
|
|
- path: teknium/GPT4-LLM-Cleaned
|
|
type: alpaca
|
|
dataset_prepared_path:
|
|
val_set_size: 0.05
|
|
adapter: qlora
|
|
lora_model_dir:
|
|
sequence_len: 2048
|
|
lora_r: 16
|
|
lora_alpha: 32
|
|
lora_dropout: 0.05
|
|
lora_target_modules:
|
|
- c_fc
|
|
- c_attn
|
|
- c_proj
|
|
lora_target_linear:
|
|
wandb_project:
|
|
wandb_entity:
|
|
wandb_watch:
|
|
wandb_name:
|
|
wandb_log_model:
|
|
output_dir: ./outputs/qlora-out
|
|
batch_size: 4
|
|
micro_batch_size: 4
|
|
num_epochs: 2
|
|
optimizer: paged_adamw_8bit
|
|
torchdistx_path:
|
|
lr_scheduler: cosine
|
|
learning_rate: 0.0002
|
|
bf16: auto
|
|
tf32: true
|
|
gradient_checkpointing: true
|
|
resume_from_checkpoint:
|
|
logging_steps: 1
|
|
xformers_attention: true
|
|
flash_attention:
|
|
gptq_groupsize:
|
|
gptq_model_v1:
|
|
warmup_ratio: 0.1
|
|
evals_per_epoch: 4
|
|
saves_per_epoch: 1
|
|
weight_decay: 0.1
|
|
special_tokens:
|
|
pad_token: "<|endoftext|>"
|