configs/oo-7b.yml

base_model: conceptofmind/LLongMA-2-7b-16k
base_model_config: conceptofmind/LLongMA-2-7b-16k
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer
tokenizer_use_fast: true
tokenizer_legacy: true
load_in_8bit: false                    
load_in_4bit: false                                    
strict: false  
push_dataset_to_hub: 
hf_use_auth_token: 
datasets:           
  - path: openaccess-ai-collective/oo-gpt4-filtered
    type: alpaca_w_system.load_open_orca_chatml
    data_files:
      - 1M-GPT4-Augmented-filtered-gt10.parquet
dataset_prepared_path: last_run_prepared
val_set_size: 0.01
adapter:
lora_model_dir:
sequence_len: 16384
max_packed_sequence_len:
sample_packing: true
sample_packing_eff_est: 0.99
sample_packing_seq_len_multiplier: 2
total_num_tokens: 372602546
lora_r:
lora_alpha:
lora_dropout:
lora_target_modules:
lora_target_linear:
lora_fan_in_fan_out:
wandb_project: open-long-orca-7b
wandb_watch:
wandb_run_id:
wandb_log_model:
output_dir: ./open-long-orca-7b
gradient_accumulation_steps: 1
micro_batch_size: 2
num_epochs: 4
optimizer: adamw_torch
adam_beta2: 0.95
adam_eps: 0.00001
max_grad_norm: 1.0
torchdistx_path:
lr_scheduler: cosine
lr_quadratic_warmup: true
learning_rate: 0.000017
train_on_inputs: false
group_by_length: false
bf16: true
fp16: false
tf32: true
gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention: 
flash_attention: true
sdp_attention:
flash_optimum:
gptq_groupsize:
gptq_model_v1:
warmup_steps: 32
eval_steps: 284
save_steps: 
debug:
deepspeed:
weight_decay: 0.1
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"
tokens:
  - "<|im_start|>"
  - "<|im_end|>"
fsdp:
  - full_shard
  - auto_wrap
fsdp_config:
  fsdp_sync_module_states: true
  fsdp_offload_params: true
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer