See axolotl config

axolotl version: 0.4.1

adapter: lora
auto_resume_from_checkpoints: true
base_model: facebook/opt-125m
bf16: auto
chat_template: llama3
dataset_prepared_path: null
dataset_processes: 6
datasets:
- data_files:
  - 818f8460304dc8f3_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/818f8460304dc8f3_train_data.json
  type:
    field_input: hypothesis
    field_instruction: premise
    field_output: augmented_hypothesis
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 3
eval_max_new_tokens: 128
eval_steps: 200
eval_table_size: null
evals_per_epoch: null
flash_attention: true
fp16: false
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: error577/18092189-46f9-4027-aa36-175d1e605e4a
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: 1
lora_alpha: 64
lora_dropout: 0.1
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: null
micro_batch_size: 24
mlflow_experiment_name: /tmp/818f8460304dc8f3_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
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: 200
sequence_len: 256
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.005
wandb_entity: null
wandb_mode: online
wandb_name: af50a6d4-247f-4e76-817b-1cfd6dd7131d
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: af50a6d4-247f-4e76-817b-1cfd6dd7131d
warmup_steps: 30
weight_decay: 0.0
xformers_attention: null

18092189-46f9-4027-aa36-175d1e605e4a

This model is a fine-tuned version of facebook/opt-125m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1241

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: 24
• eval_batch_size: 24
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 96
• 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: 30
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
12.7141	0.0005	1	3.1690
1.3992	0.0960	200	0.3349
1.0472	0.1921	400	0.2430
0.883	0.2881	600	0.2006
0.7357	0.3842	800	0.1706
0.7221	0.4802	1000	0.1574
0.7825	0.5763	1200	0.1546
0.5444	0.6723	1400	0.1439
0.6198	0.7684	1600	0.1449
0.5838	0.8644	1800	0.1378
0.5814	0.9605	2000	0.1363
0.552	1.0565	2200	0.1346
0.5699	1.1526	2400	0.1337
0.5114	1.2486	2600	0.1359
0.5164	1.3447	2800	0.1310
0.4851	1.4407	3000	0.1314
0.4927	1.5368	3200	0.1320
0.485	1.6328	3400	0.1287
0.5204	1.7289	3600	0.1270
0.5643	1.8249	3800	0.1277
0.525	1.9210	4000	0.1275
0.4682	2.0170	4200	0.1264
0.5078	2.1131	4400	0.1259
0.5088	2.2091	4600	0.1257
0.5096	2.3052	4800	0.1250
0.5158	2.4012	5000	0.1252
0.5353	2.4973	5200	0.1241
0.5068	2.5933	5400	0.1242
0.5512	2.6894	5600	0.1241
0.472	2.7854	5800	0.1242
0.4644	2.8815	6000	0.1242
0.47	2.9775	6200	0.1241

Framework versions

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