See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: facebook/opt-350m
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - f09a068a17c73549_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/f09a068a17c73549_train_data.json
  type:
    field_instruction: prompt
    field_output: gold_standard_solution
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
device_map: auto
do_eval: true
early_stopping_patience: 3
eval_batch_size: 4
eval_max_new_tokens: 128
eval_steps: 150
eval_table_size: null
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: Nexspear/5f6a803a-1722-4933-8f51-74662c492225
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_in_4bit: false
load_in_8bit: false
local_rank: 0
logging_steps: 10
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: constant
max_grad_norm: 1.0
max_memory:
  0: 140GB
max_steps: 1500
micro_batch_size: 16
mlflow_experiment_name: /tmp/f09a068a17c73549_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 10
optim_args:
  adam_beta1: 0.9
  adam_beta2: 0.999
  adam_epsilon: 1e-08
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: 150
saves_per_epoch: null
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: acopia-grant
wandb_mode: online
wandb_name: fb6671a1-3169-4044-b339-5b07d92c1691
wandb_project: Gradients-On-44
wandb_run: your_name
wandb_runid: fb6671a1-3169-4044-b339-5b07d92c1691
warmup_steps: 50
weight_decay: 0.0
xformers_attention: null

5f6a803a-1722-4933-8f51-74662c492225

This model is a fine-tuned version of facebook/opt-350m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.3589

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: 16
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=adam_beta1=0.9,adam_beta2=0.999,adam_epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 50
• training_steps: 1500

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0023	1	4.9117
21.2245	0.3390	150	2.5232
18.3485	0.6780	300	2.7024
29.4897	1.0169	450	2.5901
26.4556	1.3559	600	2.5042
26.7033	1.6949	750	2.5247
28.5978	2.0339	900	2.4551
28.4341	2.3729	1050	2.4488
29.5822	2.7119	1200	2.4180
20.835	3.0508	1350	2.3425
21.1484	3.3898	1500	2.3589

Framework versions

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