AlexHung29629/FormoMouse

See axolotl config

axolotl version: 0.9.2

base_model: AlexHung29629/FormoMouse123

plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: false
liger_rms_norm: true
liger_glu_activation: true
liger_fused_linear_cross_entropy: true

unfrozen_parameters:
  - model.lm_head.weight
  - model.embed_tokens.weight

pretraining_dataset:
  - path: AlexHung29629/para_pat_text
    split: train
    text_column: text
    type: pretrain
  - path: AlexHung29629/urban_dictionary
    split: train
    text_column: text
    type: pretrain

dataset_prepared_path: data_prep_1
val_set_size: 0.0
output_dir: ./out_1

shuffle_merged_datasets: true

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

use_tensorboard: true

gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 2
max_steps: 10000
save_steps: 1000
save_total_limit: 1
save_only_model: true
optimizer: ao_adamw_fp8
lr_scheduler: cosine
learning_rate: 2e-4
max_grad_norm: 1.0

bfloat16: true

gradient_checkpointing: true

logging_steps: 1
torch_compile: false
sdp_attention: true


warmup_ratio: 0.1
weight_decay: 0.1

out_1

This model is a fine-tuned version of AlexHung29629/FormoMouse123 on an unknown dataset.

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: 0.0002
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 1000
• training_steps: 10000

Training results

Framework versions

• Transformers 4.51.3
• Pytorch 2.7.0+cu128
• Datasets 3.5.1
• Tokenizers 0.21.1