base_model: meta-llama/Llama-3.1-8B
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
tokenizer_use_fast: false
resize_token_embeddings_to_32x: false

flash_attention: true
xformers_attention:

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: skymizer/Llama3.1-8B-base-tokenized-fineweb-edu-45B-4096
    train_on_split: train
    type: completion

test_datasets:
  - path: skymizer/Llama3.1-8B-base-tokenized-fineweb-edu-test-4K
    split: test
    type: completion

is_preprocess: true
skip_prepare_dataset: true

dataset_prepared_path: /mnt/home/model-team/datasets/pretokenized/Llama3.1-8B-base-tokenized-fineweb-edu-45B-4096 

hf_use_auth_token: true
output_dir: /mnt/home/model-team/models/Llama3.1-8B-v0.1-relu-stage-1-fineweb-edu-45B-4096
resume_from_checkpoint:
auto_resume_from_checkpoints: true

sequence_len: 4096
sample_packing: true
sample_packing_group_size: 100000
sample_packing_bin_size: 200
pad_to_sequence_len: true

eval_sample_packing: false
# eval_causal_lm_metrics: ["perplexity"]

wandb_project: "sparse-tuning-cpt"
wandb_entity:
wandb_watch:
wandb_name: "Llama3.1-8B-relu-stage-1-fineweb-edu-45B-4096"
wandb_log_model:

# global batch size = 2 * 8 * 8 GPUs * 8 Nodes * 4096 = 4M
gradient_accumulation_steps: 8
micro_batch_size: 2
  # eval_batch_size: 2
max_steps: 10000
optimizer: adamw_torch
learning_rate: 0.000015
lr_scheduler: cosine
cosine_min_lr_ratio: 1.0 
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.95
adam_eps: 0.000001
max_grad_norm: 1.0

train_on_inputs: false
group_by_length: false
bf16: true
fp16:
tf32: false

hub_model_id: "skymizer/Llama3.1-8B-relu-stage-1-fineweb-edu-45B-4096"

save_strategy: "steps"
save_steps: 500

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1

warmup_steps: 1
eval_steps: 500
eval_table_size:
debug:
deepspeed: /root/train/axolotl/deepspeed_configs/zero3_bf16.json
fsdp:
fsdp_config:
seed: 42

special_tokens:
  pad_token: "<|end_of_text|>"

Llama3.1-8B-relu-stage-1-fineweb-edu-45B-4096

This model is a fine-tuned version of meta-llama/Llama-3.1-8B on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 1.9682

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: 1.5e-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
distributed_type: multi-GPU
num_devices: 64
gradient_accumulation_steps: 8
total_train_batch_size: 1024
total_eval_batch_size: 128
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 2
training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss
12.2232	0.0001	1	12.1487
2.2025	0.0424	500	2.2272
2.1454	0.0848	1000	2.1515
2.0991	0.1273	1500	2.1142
2.0604	0.1697	2000	2.0894
2.058	0.2121	2500	2.0711
2.0582	0.2545	3000	2.0561
2.0474	0.2969	3500	2.0442
2.0268	0.3394	4000	2.0347
2.0173	0.3818	4500	2.0256
1.9941	0.4242	5000	2.0178
2.0113	0.4666	5500	2.0106
1.9949	0.5091	6000	2.0040
2.0077	0.5515	6500	1.9984
1.986	0.5939	7000	1.9935
1.9902	0.6363	7500	1.9888
1.9899	0.6787	8000	1.9841
1.9729	0.7212	8500	1.9800
1.971	0.7636	9000	1.9759
1.9784	0.8060	9500	1.9718
1.9553	0.8484	10000	1.9682

Framework versions

Transformers 4.46.3
Pytorch 2.5.1+cu124
Datasets 3.1.0
Tokenizers 0.20.3