# This works!

base_model: meta-llama/Llama-3.1-405B-Instruct
hub_model_id: jplhughes2/1a_meta-llama-Llama-3.1-405B-Instruct-fsdp-lr1e-5
load_in_8bit: false
load_in_4bit: true
adapter: qlora
wandb_name: 1a_meta-llama-Llama-3.1-405B-Instruct-fsdp-lr1e-5
output_dir: ./outputs/out/1a_meta-llama-Llama-3.1-405B-Instruct-fsdp-lr1e-5
# base_model:
# hub_model_id:
# load_in_8bit:
# load_in_4bit:
# adapter:
# wandb_name:
# output_dir:

tokenizer_type: AutoTokenizer
push_dataset_to_hub:
strict: false

datasets:
  - path: jplhughes2/docs_only_30k_filtered
    type: completion
    field: text
    split: train
dataset_prepared_path: last_run_prepared
# val_set_size: 0.05
test_datasets:
  - path: jplhughes2/docs_only_val_5k_filtered
    type: completion
    field: text
    split: train
save_safetensors: true

sequence_len: 1024
sample_packing: true
pad_to_sequence_len: true

lora_r: 64
lora_alpha: 128
lora_dropout: 0.05
lora_target_modules:
lora_target_linear: true

wandb_mode:
wandb_project: alignment-faking
wandb_entity: academicsnyuperez
wandb_watch:
wandb_run_id:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 1
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 0.00001

train_on_inputs: false
group_by_length: false
bf16: true
tf32: true

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: true
logging_steps: 1
flash_attention: true

warmup_steps: 10
evals_per_epoch: 3
saves_per_epoch: 1
weight_decay: 0.01
fsdp:
  - full_shard
  - auto_wrap
fsdp_config:
  fsdp_limit_all_gathers: true
  fsdp_sync_module_states: true
  fsdp_offload_params: false
  fsdp_use_orig_params: false
  fsdp_cpu_ram_efficient_loading: true
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_sharding_strategy: FULL_SHARD
special_tokens:
  pad_token: <|finetune_right_pad_id|>

1a_meta-llama-Llama-3.1-405B-Instruct-fsdp-lr1e-5

This model is a fine-tuned version of meta-llama/Llama-3.1-405B-Instruct on the jplhughes2/docs_only_30k_filtered dataset. It achieves the following results on the evaluation set:

Loss: 0.6041

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 8
total_train_batch_size: 32
total_eval_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 10
num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.323	0.0016	1	1.3262
0.648	0.3344	204	0.6514
0.6137	0.6689	408	0.6041

Framework versions

PEFT 0.14.0
Transformers 4.48.3
Pytorch 2.4.1+cu124
Datasets 3.2.0
Tokenizers 0.21.0