See axolotl config

axolotl version: 0.13.0.dev0

base_model: Qwen/Qwen2.5-72B
# Alpaca finetuning configuration for Qwen2.5-72B
# hub_model_id: username/custom_model_name

load_in_8bit: false
load_in_4bit: false
strict: false

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
dataset_prepared_path: /home/ubuntu/cluster-axolotl/alpaca_prepared_qwen72b
seed: 42
chat_template: qwen_25
datasets:
  - path: tatsu-lab/alpaca
    type: alpaca

output_dir: ./outputs/qat_out_qwen72b/

sequence_len: 8096
sample_packing: true
flash_attention: true

qat:
  activation_dtype: nvfp4
  weight_dtype: nvfp4
  group_size: 16 # only group_size of 16 is supported with nvfp4

wandb_project: qat_b200
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 16

num_epochs: 1
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 2e-6

bf16: true
tf32: true

resume_from_checkpoint:
logging_steps: 1
save_strategy: "no"

# evals_per_epoch: 1
# saves_per_epoch: 1

warmup_ratio: 0.1
weight_decay: 0.0
fsdp_version: 2

fsdp_config:
  offload_params: false
  cpu_ram_efficient_loading: true
  auto_wrap_policy: TRANSFORMER_BASED_WRAP
  transformer_layer_cls_to_wrap: Qwen2DecoderLayer
  state_dict_type: FULL_STATE_DICT
  sharding_strategy: FULL_SHARD
  reshard_after_forward: true
  activation_checkpointing: true

special_tokens:

# save_first_step: true  # uncomment this to validate checkpoint saving works with your config

home/ubuntu/cluster-axolotl/logs/Qwen72BAlpaca2e-6_qat

This model was trained from scratch on the tatsu-lab/alpaca 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-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 16
• total_train_batch_size: 256
• total_eval_batch_size: 256
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 1.0

Training results

Framework versions

• Transformers 4.56.1
• Pytorch 2.8.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.1