See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: microsoft/phi-1_5
bf16: false
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 5182ff51221e361c_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/5182ff51221e361c_train_data.json
  type:
    field_input: description
    field_instruction: input persona
    field_output: synthesized text
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deep_speed: true
early_stopping_patience: null
evals_per_epoch: 2
flash_attention: false
fp16: true
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: true
hub_model_id: Racimrl/9597c250-1547-4bee-b260-b6091a042621
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.00015
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 10
lora_alpha: 32
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_steps: 75000
micro_batch_size: 8
mlflow_experiment_name: /tmp/5182ff51221e361c_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
saves_per_epoch: 2
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: 2b2dd179-acaa-4dd7-8e71-b457a04bcc14
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 2b2dd179-acaa-4dd7-8e71-b457a04bcc14
warmup_steps: 100
weight_decay: 0.01
xformers_attention: true

9597c250-1547-4bee-b260-b6091a042621

This model is a fine-tuned version of microsoft/phi-1_5 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9616

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: 0.00015
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 100
• training_steps: 2957
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0007	1	1.4916
1.0315	0.5005	740	1.0100
1.0795	1.0010	1480	0.9775
0.8863	1.5015	2220	0.9616

Framework versions

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