See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: JackFram/llama-68m
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 425c6bf4bb96a710_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/425c6bf4bb96a710_train_data.json
  type:
    field_input: paragraph
    field_instruction: question
    field_output: answer
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 5
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 16
gradient_checkpointing: false
group_by_length: false
hub_model_id: error577/678bdeae-3a40-4e29-b67c-7fbc0c52dd71
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.001
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 8
lora_target_linear: true
lr_scheduler: cosine
max_steps: 1000
micro_batch_size: 2
mlflow_experiment_name: /tmp/425c6bf4bb96a710_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: 4
sequence_len: 512
special_tokens:
  pad_token: </s>
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 832d80d5-251c-46fe-b13b-7cc0427dac5a
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 832d80d5-251c-46fe-b13b-7cc0427dac5a
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

678bdeae-3a40-4e29-b67c-7fbc0c52dd71

This model is a fine-tuned version of JackFram/llama-68m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.0955

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: 0.001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 16
• 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: 387

Training results

Training Loss	Epoch	Step	Validation Loss
5.8379	0.0052	1	5.7354
0.9867	0.2020	39	1.4082
1.0126	0.4040	78	1.2563
0.8883	0.6060	117	1.1930
1.8973	0.8080	156	1.1759
2.993	1.0100	195	1.2300
0.5959	1.2120	234	1.1373
0.7068	1.4140	273	1.1433
0.9381	1.6161	312	1.0941
0.8364	1.8181	351	1.0955

Framework versions

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