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:
  - b78d911fe67c1c29_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/b78d911fe67c1c29_train_data.json
  type:
    field_input: context
    field_instruction: prompt_serial
    field_output: hypothesis
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 150
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/02fd5a8b-55d2-49c6-b7e6-b94405441095
hub_repo: null
hub_strategy: checkpoint
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: 0.3
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
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 5040
micro_batch_size: 2
mlflow_experiment_name: /tmp/b78d911fe67c1c29_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: 150
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: 18ac5d29-7685-46f6-ac86-d879293e3c5f
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 18ac5d29-7685-46f6-ac86-d879293e3c5f
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

02fd5a8b-55d2-49c6-b7e6-b94405441095

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: 0.0112

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_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: 5040

Training results

Training Loss	Epoch	Step	Validation Loss
3.1266	0.0002	1	3.6308
0.0981	0.0315	150	0.0880
0.2966	0.0630	300	0.0530
0.0123	0.0945	450	0.0245
0.0024	0.1259	600	0.0220
0.1027	0.1574	750	0.0175
0.0128	0.1889	900	0.0274
0.0113	0.2204	1050	0.0150
0.1088	0.2519	1200	0.0182
0.0377	0.2834	1350	0.0129
0.002	0.3149	1500	0.0131
0.0387	0.3463	1650	0.0109
0.0039	0.3778	1800	0.0124
0.0009	0.4093	1950	0.0112

Framework versions

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