See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: microsoft/Phi-3-mini-128k-instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 1fed254576df142d_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/1fed254576df142d_train_data.json
  type:
    field_instruction: question
    field_output: answer
    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/83085f87-a61b-4e55-9e3a-5c94617602a4
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: 0.3
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: 1734
micro_batch_size: 4
mlflow_experiment_name: /tmp/1fed254576df142d_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
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: 9090c43e-ec93-46ec-8b87-d11083a1aa8d
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 9090c43e-ec93-46ec-8b87-d11083a1aa8d
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

83085f87-a61b-4e55-9e3a-5c94617602a4

This model is a fine-tuned version of microsoft/Phi-3-mini-128k-instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1721

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: 1734

Training results

Training Loss	Epoch	Step	Validation Loss
2.0314	0.0007	1	0.2233
1.4041	0.0682	100	0.1848
1.4015	0.1363	200	0.1847
1.3008	0.2045	300	0.1823
1.4179	0.2727	400	0.1815
1.2804	0.3408	500	0.1807
1.3661	0.4090	600	0.1787
1.7189	0.4772	700	0.1777
1.2251	0.5453	800	0.1764
1.5796	0.6135	900	0.1748
1.5114	0.6817	1000	0.1735
1.4334	0.7498	1100	0.1725
1.2593	0.8180	1200	0.1713
1.14	0.8862	1300	0.1704
1.2217	0.9543	1400	0.1698
0.9364	1.0225	1500	0.1717
1.0701	1.0907	1600	0.1721

Framework versions

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