See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: openlm-research/open_llama_3b
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 7e19f93aea14929b_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/7e19f93aea14929b_train_data.json
  type:
    field_instruction: input persona
    field_output: synthesized text
    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: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/eddad3ba-9804-4dbb-bf90-9f5b5a6fed4c
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: 1734
micro_batch_size: 4
mlflow_experiment_name: /tmp/7e19f93aea14929b_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: 2048
special_tokens:
  pad_token: </s>
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: 16fab843-6012-4bc5-8500-e22db5acad8c
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 16fab843-6012-4bc5-8500-e22db5acad8c
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

eddad3ba-9804-4dbb-bf90-9f5b5a6fed4c

This model is a fine-tuned version of openlm-research/open_llama_3b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8444

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
1.4332	0.0007	1	1.4838
1.0237	0.0678	100	0.9930
0.9257	0.1355	200	0.9565
0.9386	0.2033	300	0.9361
0.9052	0.2710	400	0.9206
0.8509	0.3388	500	0.9081
0.8631	0.4065	600	0.8954
0.8586	0.4743	700	0.8861
0.8717	0.5420	800	0.8763
0.8122	0.6098	900	0.8680
0.906	0.6775	1000	0.8607
0.8449	0.7453	1100	0.8537
0.8222	0.8130	1200	0.8480
0.8817	0.8808	1300	0.8437
0.8114	0.9485	1400	0.8405
0.7496	1.0163	1500	0.8429
0.6948	1.0840	1600	0.8444

Framework versions

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