base_model: kakaocorp/kanana-1.5-2.1b-instruct-2505


load_in_8bit: false
load_in_4bit: false

datasets:
  - path: train.jsonl
    type: chat_template

dataset_prepared_path: preprocess
val_set_size: 0.01
output_dir: ./outputs
dataloader_num_workers: 56

adapter: 
lora_model_dir:

sequence_len: 16384
sample_packing: false
eval_sample_packing: false
pad_to_sequence_len: false

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

wandb_project: fastcampus
wandb_entity: guijinson
wandb_watch:
wandb_name: fc-proj2-reasoning-2.1b
wandb_log_model:
hub_model_id: amphora/fc-reasoning-2.1b

gradient_accumulation_steps: 64
micro_batch_size: 2
num_epochs: 3
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 2e-5

bf16: auto
tf32: false

gradient_checkpointing:
resume_from_checkpoint:
logging_steps: 1
flash_attention: true

warmup_ratio: 0.05
weight_decay: 0.01
evals_per_epoch: 0
saves_per_epoch: 1

fc-reasoning-2.1b

This model is a fine-tuned version of kakaocorp/kanana-1.5-2.1b-instruct-2505 on the train.jsonl 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: 2e-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 64
total_train_batch_size: 128
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: 53
training_steps: 1072

Training results

Framework versions

Transformers 4.55.2
Pytorch 2.6.0+cu126
Datasets 4.0.0
Tokenizers 0.21.4