See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: echarlaix/tiny-random-PhiForCausalLM
bf16: true
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
device_map:
  ? ''
  : 0,1,2,3,4,5,6,7
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: false
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/a20c3179-6bc5-4fb9-a229-a853a34533ba
hub_repo: null
hub_strategy: null
hub_token: null
learning_rate: 0.0002
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 128
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
- o_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 4679
micro_batch_size: 4
mlflow_experiment_name: /tmp/df4145ffec439e54_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 2
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 100
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: Gradients-On-Demand
wandb_run: your_name
wandb_runid: c30de9d5-9d24-44e1-9b62-deb3229d1fce
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

a20c3179-6bc5-4fb9-a229-a853a34533ba

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: 6.6151

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: 0.0002
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• 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: 4679

Training results

Training Loss	Epoch	Step	Validation Loss
6.9446	0.0004	1	6.9443
6.6666	0.0384	100	6.6708
6.6741	0.0768	200	6.6562
6.642	0.1152	300	6.6517
6.6381	0.1536	400	6.6443
6.6413	0.1920	500	6.6408
6.6588	0.2303	600	6.6375
6.6395	0.2687	700	6.6342
6.6493	0.3071	800	6.6328
6.6319	0.3455	900	6.6312
6.6399	0.3839	1000	6.6303
6.6457	0.4223	1100	6.6292
6.6224	0.4607	1200	6.6274
6.6191	0.4991	1300	6.6265
6.6389	0.5375	1400	6.6260
6.6216	0.5759	1500	6.6247
6.6458	0.6143	1600	6.6238
6.6266	0.6527	1700	6.6224
6.647	0.6910	1800	6.6211
6.6328	0.7294	1900	6.6197
6.6377	0.7678	2000	6.6186
6.6258	0.8062	2100	6.6181
6.6202	0.8446	2200	6.6180
6.6182	0.8830	2300	6.6173
6.6185	0.9214	2400	6.6168
6.6255	0.9598	2500	6.6165
6.6301	0.9982	2600	6.6162
7.1902	1.0366	2700	6.6162
5.5572	1.0750	2800	6.6156
6.7101	1.1134	2900	6.6156
6.1659	1.1517	3000	6.6155
6.7201	1.1901	3100	6.6153
6.6188	1.2285	3200	6.6152
5.846	1.2669	3300	6.6151
6.8901	1.3053	3400	6.6148
6.0211	1.3437	3500	6.6149
7.3173	1.3821	3600	6.6151

Framework versions

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