base_model: Qwen/Qwen2.5-7B
hub_model_id: OsakanaTeishoku/Qwen2.5-7B-axolotl-sft-v0.2

load_in_8bit: false
load_in_4bit: true
strict: false

chat_template: qwen_25

datasets:
  # This will be the path used for the data when it is saved to the Volume in the cloud.
  - path: Aratako/Magpie-Tanuki-8B-annotated-96k
    split: train
    type: chat_template
    field_messages: messages
  - path: Aratako/Synthetic-JP-EN-Coding-Dataset-Magpie-69k
    split: train
    type: chat_template
    field_messages: messages
  - path: DataPilot/Zero_SFT_Ja_v2_b3t4
    split: train
    type: chat_template
    field_messages: conversation
    message_property_mappings:
      role: from
      content: value

shuffle_merged_datasets: true

dataset_prepared_path: last_run_prepared
#val_set_size: 0.05
output_dir: ./lora-out

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

adapter: qlora
lora_model_dir:
lora_r: 16
lora_alpha: 32
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_modules_to_save: # required when adding new tokens to LLaMA/Mistral
  - embed_tokens
  - lm_head

wandb_project: modal-axolotl
wandb_name: 20250419-qwen7b-modal

gradient_accumulation_steps: 4
micro_batch_size: 16
#auto_find_batch_size: true
#num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0001

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

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

warmup_ratio: 0.05
save_steps: 50
max_steps: 200
debug:
#deepspeed: /workspace/axolotl/deepspeed_configs/zero2.json
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  eos_token: "<|im_end|>"

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

eval_strategy: "no"
save_strategy: "steps"

Qwen2.5-7B-axolotl-sft-v0.2

This model is a fine-tuned version of Qwen/Qwen2.5-7B on the Aratako/Magpie-Tanuki-8B-annotated-96k, the Aratako/Synthetic-JP-EN-Coding-Dataset-Magpie-69k and the DataPilot/Zero_SFT_Ja_v2_b3t4 datasets.

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: 0.0001
train_batch_size: 16
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 64
optimizer: Use OptimizerNames.ADAMW_BNB 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: 10
training_steps: 200

Training results

Framework versions

PEFT 0.15.1
Transformers 4.51.3
Pytorch 2.6.0+cu124
Datasets 3.5.0
Tokenizers 0.21.1