# We train the exact same as 02-llama31 but we swap gradient accum and micro bs (should be same but faster?)

base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

# User Liger
plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_fused_linear_cross_entropy: true

chat_template: llama3
datasets:
  - path: augmxnt/ultra-orca-boros-en-ja-v1
    # type: sharegpt deprecated
    type: chat_template
    field_messages: conversations
    message_field_role: from
    message_field_content: value
dataset_prepared_path: last_run_prepared
val_set_size: 0.05
output_dir: ./outputs/ablation-05-bs2ga4-shisa-v2-llama3.1-8b-lr8e6

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

# marginal difference
neftune_noise_alpha: 5

use_wandb: true
wandb_project: shisa-v2
wandb_entity: augmxnt
wandb_name: ablation-04-bs2ga4-shisa-v2-llama3.1-8b.lr-8e6

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 3
optimizer: paged_adamw_8bit
lr_scheduler: linear
learning_rate: 8e-6

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

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 100
evals_per_epoch: 2
eval_table_size:
saves_per_epoch: 0
save_total_limit: 1 # Only store a single checkpoint
debug:
deepspeed: zero3_bf16.json
weight_decay: 0.00
fsdp:
fsdp_config:
special_tokens:
  pad_token: <|end_of_text|>

outputs/ablation-05-bs2ga4-shisa-v2-llama3.1-8b-lr8e6

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the augmxnt/ultra-orca-boros-en-ja-v1 dataset. It achieves the following results on the evaluation set:

Loss: 0.4913

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: 8e-06
train_batch_size: 2
eval_batch_size: 2
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 4
total_train_batch_size: 64
total_eval_batch_size: 16
optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.0952	0.0064	1	0.6669
0.9226	0.4984	78	0.5535
0.8536	0.9968	156	0.5133
0.7586	1.4920	234	0.4989
0.733	1.9904	312	0.4877
0.6535	2.4856	390	0.4931
0.6322	2.9840	468	0.4913

Framework versions

Transformers 4.48.3
Pytorch 2.6.0+cu124
Datasets 3.2.0
Tokenizers 0.21.0

shisa-ai/ablation-05-bs2ga4-shisa-v2-llama3.1-8b-lr8e6

outputs/ablation-05-bs2ga4-shisa-v2-llama3.1-8b-lr8e6

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

shisa-ai/
ablation-05-bs2ga4-shisa-v2-llama3.1-8b-lr8e6