diff --git a/config/llama_1b_with_olmo_config.yaml b/config/llama_1b_with_olmo_config.yaml index cfe0893af..fe315de71 100644 --- a/config/llama_1b_with_olmo_config.yaml +++ b/config/llama_1b_with_olmo_config.yaml @@ -9,7 +9,6 @@ model: # 1B class model num_kv_heads: 16 use_flash_attention: True flash_attention_block_size: 1024 - attn_backend: splash trainer: tracker: type: wandb