See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: trl-internal-testing/tiny-random-LlamaForCausalLM
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - fb3f92080c25ece3_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/fb3f92080c25ece3_train_data.json
  type:
    field_instruction: source_text
    field_output: target_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/32983a73-7290-46dc-9e4a-06e649f1a344
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: 0.3
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
- o_proj
- gate_proj
- down_proj
- up_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 3600
micro_batch_size: 4
mlflow_experiment_name: /tmp/fb3f92080c25ece3_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
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.023962312075567548
wandb_entity: null
wandb_mode: online
wandb_name: 9c23a19f-2769-4bed-9616-2c50a201f19a
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 9c23a19f-2769-4bed-9616-2c50a201f19a
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

32983a73-7290-46dc-9e4a-06e649f1a344

This model is a fine-tuned version of trl-internal-testing/tiny-random-LlamaForCausalLM on the None dataset. It achieves the following results on the evaluation set:

• Loss: 10.3037

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: 3600

Training results

Training Loss	Epoch	Step	Validation Loss
10.3727	0.0002	1	10.3746
10.3418	0.0157	100	10.3386
10.3276	0.0314	200	10.3224
10.3376	0.0471	300	10.3171
10.3224	0.0628	400	10.3141
10.3309	0.0786	500	10.3121
10.3239	0.0943	600	10.3109
10.3207	0.1100	700	10.3099
10.3188	0.1257	800	10.3088
10.3228	0.1414	900	10.3087
10.3154	0.1571	1000	10.3077
10.3189	0.1728	1100	10.3072
10.3194	0.1885	1200	10.3068
10.3218	0.2043	1300	10.3065
10.3236	0.2200	1400	10.3061
10.3205	0.2357	1500	10.3055
10.321	0.2514	1600	10.3054
10.3203	0.2671	1700	10.3054
10.321	0.2828	1800	10.3053
10.3179	0.2985	1900	10.3048
10.3228	0.3142	2000	10.3046
10.3145	0.3300	2100	10.3045
10.3242	0.3457	2200	10.3044
10.318	0.3614	2300	10.3042
10.3167	0.3771	2400	10.3042
10.3135	0.3928	2500	10.3042
10.3163	0.4085	2600	10.3042
10.3239	0.4242	2700	10.3038
10.317	0.4399	2800	10.3038
10.3154	0.4557	2900	10.3039
10.3187	0.4714	3000	10.3037
10.3217	0.4871	3100	10.3037
10.317	0.5028	3200	10.3037
10.3137	0.5185	3300	10.3037
10.3193	0.5342	3400	10.3037

Framework versions

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