68 lines
1.1 KiB
YAML
68 lines
1.1 KiB
YAML
stage: sft
|
|
do_train: true
|
|
do_eval: true
|
|
|
|
model_name_or_path: slovak-nlp/mistral-sk-7b
|
|
trust_remote_code: true
|
|
|
|
dataset_dir: data
|
|
dataset: aya_slovak_full
|
|
template: alpaca
|
|
cutoff_len: 1024
|
|
packing: false
|
|
train_on_prompt: false
|
|
|
|
finetuning_type: lora
|
|
lora_rank: 16
|
|
lora_alpha: 16
|
|
lora_dropout: 0.0
|
|
lora_target: q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj
|
|
|
|
per_device_train_batch_size: 1
|
|
per_device_eval_batch_size: 1
|
|
gradient_accumulation_steps: 8
|
|
|
|
learning_rate: 0.00005
|
|
lr_scheduler_type: cosine
|
|
warmup_steps: 10000
|
|
max_grad_norm: 0.5
|
|
|
|
eval_strategy: steps
|
|
val_size: 1000
|
|
eval_steps: 25000
|
|
|
|
fp16: true
|
|
optim: adamw_torch
|
|
flash_attn: auto
|
|
plot_loss: true
|
|
|
|
ddp_timeout: 180000000
|
|
ddp_find_unused_parameters: false
|
|
include_num_input_tokens_seen: true
|
|
|
|
enable_thinking: false
|
|
|
|
use_unsloth: true
|
|
use_unsloth_gc: true
|
|
|
|
streaming: true
|
|
buffer_size: 50000
|
|
|
|
max_steps: 225000
|
|
|
|
overwrite_cache: false
|
|
|
|
output_dir: /home/schwarc/diplomovka/aya_streaming_lora_unsloth_2gpu_dispatchfalse_225k
|
|
overwrite_output_dir: false
|
|
|
|
logging_steps: 50
|
|
save_steps: 25000
|
|
save_total_limit: 2
|
|
report_to: none
|
|
|
|
seed: 42
|
|
|
|
accelerator_config:
|
|
dispatch_batches: false
|
|
split_batches: false
|