See axolotl config

axolotl version: 0.6.0

base_model: unsloth_phi-4
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

  #hub_model_id: NewEden/Phi4-pretrain
  #hub_strategy: "all_checkpoints"
  #push_dataset_to_hub:
  #hf_use_auth_token: true

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

  #plugins:
  #  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin

  #cut_cross_entropy: true

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: Mielikki/Erebus-87k
    type: completion
    field: body
  - path: NewEden/Orion-Asstr-Stories-16K
    type: completion
    field: content
shuffle_merged_datasets: true
dataset_prepared_path: prepared_data
val_set_size: 0.0
output_dir: ./phi4-pt-out-r1

sequence_len: 16384
sample_packing: true
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 128
lora_alpha: 16 
lora_dropout: 0.05
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

lora_modules_to_save:
 - embed_tokens
 - lm_head


wandb_project: mag-phi
wandb_entity:
wandb_watch:
wandb_name: attempt-01
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 1
optimizer: paged_ademamix_8bit
lr_scheduler: constant_with_warmup
learning_rate: 0.0001

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

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

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 4
debug:
deepspeed: /workspace/axolotl/deepspeed_configs/zero3_bf16_cpuoffload_params.json
weight_decay: 0.01
fsdp:
fsdp_config:

phi4-pt-out-r1

This model was trained from scratch on the Mielikki/Erebus-87k and the NewEden/Orion-Asstr-Stories-16K 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: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• total_eval_batch_size: 8
• optimizer: Use OptimizerNames.PAGED_ADEMAMIX_8BIT and the args are: No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 10
• num_epochs: 1.0

Training results

Framework versions

• PEFT 0.14.0
• Transformers 4.48.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0