See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: echarlaix/tiny-random-PhiForCausalLM
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - df4145ffec439e54_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/df4145ffec439e54_train_data.json
  type:
    field_instruction: text
    field_output: text_ja
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
do_eval: true
early_stopping_patience: 3
eval_batch_size: 4
eval_max_new_tokens: 128
eval_steps: 500
evals_per_epoch: null
flash_attention: true
fp16: false
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: true
hub_model_id: lesso03/c06ad496-016e-4dba-912f-d4694b1326e5
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.000203
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 50
lora_alpha: 128
lora_dropout: 0.15
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 50000
micro_batch_size: 4
mlflow_experiment_name: /tmp/df4145ffec439e54_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 10
optimizer: adamw_torch_fused
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 500
saves_per_epoch: null
seed: 30
sequence_len: 1024
special_tokens:
  pad_token: <|endoftext|>
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: c30de9d5-9d24-44e1-9b62-deb3229d1fce
wandb_project: 03a
wandb_run: your_name
wandb_runid: c30de9d5-9d24-44e1-9b62-deb3229d1fce
warmup_steps: 100
weight_decay: 0.0
xformers_attention: null

c06ad496-016e-4dba-912f-d4694b1326e5

This model is a fine-tuned version of echarlaix/tiny-random-PhiForCausalLM on the None dataset. It achieves the following results on the evaluation set:

• Loss: 6.5884

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: 0.000203
• train_batch_size: 4
• eval_batch_size: 4
• seed: 30
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• 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: 100
• training_steps: 26048

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0004	1	6.9452
6.6474	0.1920	500	6.6445
6.6353	0.3839	1000	6.6336
6.6273	0.5759	1500	6.6230
6.6234	0.7678	2000	6.6201
6.6209	0.9598	2500	6.6149
6.6169	1.1517	3000	6.6078
6.618	1.3437	3500	6.6060
6.6122	1.5357	4000	6.6016
6.6069	1.7276	4500	6.5994
6.6064	1.9196	5000	6.5984
6.6088	2.1115	5500	6.5970
6.6035	2.3035	6000	6.5960
6.6125	2.4954	6500	6.5956
6.6003	2.6874	7000	6.5950
6.6008	2.8794	7500	6.5937
6.6031	3.0713	8000	6.5941
6.6004	3.2633	8500	6.5936
6.6042	3.4552	9000	6.5931
6.6021	3.6472	9500	6.5927
6.6011	3.8391	10000	6.5925
6.5981	4.0311	10500	6.5920
6.5998	4.2231	11000	6.5916
6.5987	4.4150	11500	6.5916
6.6003	4.6070	12000	6.5915
6.5987	4.7989	12500	6.5908
6.6007	4.9909	13000	6.5910
6.5971	5.1828	13500	6.5904
6.6008	5.3748	14000	6.5898
6.6001	5.5668	14500	6.5896
6.5955	5.7587	15000	6.5894
6.6	5.9507	15500	6.5894
6.6004	6.1426	16000	6.5891
6.6042	6.3346	16500	6.5895
6.5999	6.5265	17000	6.5891
6.5999	6.7185	17500	6.5886
6.5996	6.9105	18000	6.5882
6.5971	7.1024	18500	6.5881
6.5959	7.2944	19000	6.5885
6.5999	7.4863	19500	6.5887
6.5977	7.6783	20000	6.5884

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1