| base_model: huggyllama/llama-7b | |
| base_model_config: huggyllama/llama-7b | |
| model_type: LlamaForCausalLM | |
| tokenizer_type: LlamaTokenizer | |
| load_in_8bit: false | |
| load_4bit: false | |
| datasets: | |
| - path: teknium/GPT4-LLM-Cleaned | |
| type: alpaca | |
| dataset_prepared_path: last_run_prepared | |
| val_set_size: 0.02 | |
| adapter: | |
| lora_model_dir: | |
| sequence_len: 2048 | |
| max_packed_sequence_len: | |
| lora_r: 16 | |
| lora_alpha: 16 | |
| lora_dropout: 0.05 | |
| lora_target_modules: | |
| - gate_proj | |
| - down_proj | |
| - up_proj | |
| - q_proj | |
| - v_proj | |
| - k_proj | |
| - o_proj | |
| lora_fan_in_fan_out: | |
| wandb_project: lora-experiment | |
| wandb_watch: | |
| wandb_run_id: | |
| wandb_log_model: | |
| output_dir: ./lora-experiment | |
| batch_size: 8 | |
| micro_batch_size: 4 | |
| num_epochs: 4 | |
| optimizer: adamw_torch | |
| torchdistx_path: | |
| lr_scheduler: cosine | |
| learning_rate: 0.00003 | |
| train_on_inputs: false | |
| group_by_length: true | |
| bf16: true | |
| tf32: true | |
| gradient_checkpointing: | |
| early_stopping_patience: | |
| resume_from_checkpoint: | |
| local_rank: | |
| logging_steps: 1 | |
| xformers_attention: true | |
| flash_attention: | |
| gptq_groupsize: | |
| gptq_model_v1: | |
| warmup_steps: 50 | |
| eval_steps: 134 | |
| save_steps: | |
| debug: | |
| deepspeed: | |
| weight_decay: 0 | |
| fsdp: | |
| - full_shard | |
| - auto_wrap | |
| fsdp_config: | |
| fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer | |
| special_tokens: | |
| bos_token: "<s>" | |
| eos_token: "</s>" | |
| unk_token: "<unk>" | |