See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: dltjdgh0928/test_instruction
bf16: true
chat_template: llama3
dataloader_num_workers: 24
dataset_prepared_path: null
datasets:
- data_files:
  - eb9250d3d2616e69_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/eb9250d3d2616e69_train_data.json
  type:
    field_input: keywords
    field_instruction: captions_background
    field_output: intention
    format: '{instruction} {input}'
    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: 2
eval_max_new_tokens: 128
eval_steps: 500
eval_table_size: null
evals_per_epoch: null
flash_attention: true
fp16: false
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: false
group_by_length: true
hub_model_id: nttx/b68b8a84-63b9-49a6-94ce-c92b2c5ad284
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 50
lora_alpha: 64
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 5000
micro_batch_size: 2
mlflow_experiment_name: /tmp/eb9250d3d2616e69_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 10
optim_args:
  adam_beta1: 0.9
  adam_beta2: 0.999
  adam_epsilon: 1e-8
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: 500
saves_per_epoch: null
sequence_len: 512
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: 4eabefbe-2458-46f4-a26f-2456638d7f39
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 4eabefbe-2458-46f4-a26f-2456638d7f39
warmup_steps: 50
weight_decay: 0.0
xformers_attention: null

b68b8a84-63b9-49a6-94ce-c92b2c5ad284

This model is a fine-tuned version of dltjdgh0928/test_instruction on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.4128

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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-8
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 50
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0004	1	2.7054
6.7657	0.1809	500	1.7205
6.5749	0.3619	1000	1.7143
6.2674	0.5428	1500	1.6468
6.0226	0.7238	2000	1.5607
5.7876	0.9047	2500	1.4811
4.0223	1.0857	3000	1.4806
3.9904	1.2666	3500	1.4531
3.847	1.4476	4000	1.4288
3.8327	1.6285	4500	1.4105
3.6422	1.8095	5000	1.4128

Framework versions

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