| base_model: mistralai/Mixtral-8x7B-Instruct-v0.1 | |
| model_type: MixtralForCausalLM | |
| tokenizer_type: LlamaTokenizer | |
| is_mistral_derived_model: false | |
| trust_remote_code: true | |
| load_in_8bit: false | |
| load_in_4bit: true | |
| strict: false | |
| datasets: | |
| - path: yentinglin/v1 | |
| type: sharegpt | |
| conversation: chatml | |
| dataset_prepared_path: | |
| val_set_size: 0.005 | |
| output_dir: ./Taiwan-LLM-MoE-pilot-v0.2 | |
| adapter: qlora | |
| lora_model_dir: | |
| sequence_len: 16384 | |
| sample_packing: true | |
| pad_to_sequence_len: true | |
| model_config: | |
| output_router_logits: true | |
| lora_r: 32 | |
| lora_alpha: 16 | |
| lora_dropout: 0.05 | |
| lora_target_modules: | |
| lora_target_linear: true | |
| lora_fan_in_fan_out: | |
| lora_modules_to_save: | |
| - embed_tokens | |
| - lm_head | |
| hub_model_id: yentinglin/Taiwan-LLM-MoE-chat-alpha-3e | |
| hub_strategy: end | |
| wandb_project: Taiwan-LLM-MoE | |
| wandb_entity: | |
| wandb_watch: | |
| wandb_name: | |
| wandb_log_model: | |
| gradient_accumulation_steps: 2 | |
| micro_batch_size: 3 | |
| num_epochs: 3 | |
| optimizer: paged_adamw_8bit | |
| lr_scheduler: cosine | |
| learning_rate: 0.0002 | |
| train_on_inputs: false | |
| group_by_length: false | |
| bf16: true | |
| fp16: false | |
| tf32: false | |
| gradient_checkpointing: true | |
| early_stopping_patience: | |
| resume_from_checkpoint: | |
| local_rank: | |
| logging_steps: 1 | |
| xformers_attention: | |
| flash_attention: true | |
| warmup_steps: 10 | |
| evals_per_epoch: 4 | |
| eval_table_size: | |
| eval_table_max_new_tokens: 128 | |
| saves_per_epoch: 1 | |
| debug: | |
| deepspeed: deepspeed/zero2.json | |
| weight_decay: 0.0 | |
| fsdp: | |
| fsdp_config: | |
| special_tokens: | |
| eos_token: "<|im_end|>" | |
| tokens: | |
| - "<|im_start|>" | |
| trust_remote_code: true | |
| ddp_timeout: 8640000000 | |
| dataset_processes: 16 | |