See axolotl config

axolotl version: 0.4.0

base_model: EnumaInc/ko-TinyLlama-1.1B-intermediate-step-1431k-3Tb-vocab-extend-45000-untrained-v1
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: /root/axolotl/datasets/mix_corpus_extended_validated.json
    type: completion
    field: text
dataset_prepared_path:

val_set_size: 0.01
output_dir: ./out

sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true
eval_sample_packing: false

wandb_project: language-transfer-eeve-v2
wandb_entity:
wandb_watch:
wandb_name: eeve-v2-stage1
wandb_log_model:

gradient_accumulation_steps: 8
micro_batch_size: 32
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.00015

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 500
evals_per_epoch: 1
eval_table_size:
eval_max_new_tokens: 128

save_strategy: steps
save_steps: 100
save_total_limit: 5

#saves_per_epoch: 1

debug:
deepspeed:
weight_decay: 0.1
fsdp:
fsdp_config:
special_tokens:

# for curriculum learning
shuffle_merged_datasets: false

unfrozen_parameters:
  - ^model.embed_tokens.weight$[32000:]
# - model.layers.2[0-9]+.block_sparse_moe.gate
#  - model.layers.2[0-9]+.block_sparse_moe.experts
#  - model.layers.3[0-9]+.block_sparse_moe.gate
#  - model.layers.3[0-9]+.block_sparse_moe.experts

out

This model is a fine-tuned version of EnumaInc/ko-TinyLlama-1.1B-intermediate-step-1431k-3Tb-vocab-extend-45000-untrained-v1 on the None dataset. It achieves the following results on the evaluation set:

Loss: 1.6306

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: 0.00015
train_batch_size: 32
eval_batch_size: 32
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 8
total_train_batch_size: 1024
total_eval_batch_size: 128
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 500
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.651	1.0	1918	1.6306

Framework versions

Transformers 4.40.0.dev0
Pytorch 2.1.2+cu118
Datasets 2.18.0
Tokenizers 0.15.0