diff --git a/scripts/short_ctx/gemma_qa_short_ctx_exp_default.sh b/scripts/short_ctx/gemma_qa_short_ctx_exp_default.sh index 8acb5ab..aea1a32 100644 --- a/scripts/short_ctx/gemma_qa_short_ctx_exp_default.sh +++ b/scripts/short_ctx/gemma_qa_short_ctx_exp_default.sh @@ -18,9 +18,9 @@ uv run accelerate launch --main_process_port $port \ --gradient_accumulation_steps=2 \ --per_device_eval_batch_size=64 \ --target_modules=down_proj \ ---num_self_attends_per_block=8 \ ---n_cross_attn_layers=8 \ ---num_latent_factor=1 \ +--num_blocks=8 \ +--num_self_attn_per_block=0 \ +--n_latent_queries=208 \ --num_pre_head_layers=1 \ --lora_r=8 \ --eval_steps=1000 \