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:
  - d54863544af6e081_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/d54863544af6e081_train_data.json
  type:
    field_input: original_version
    field_instruction: title
    field_output: french_version
    format: '{instruction} {input}'
    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/c0aaf4e6-63a1-4a45-9922-241a2b1c7a1a
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: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
- o_proj
- down_proj
- up_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 3060
micro_batch_size: 4
mlflow_experiment_name: /tmp/d54863544af6e081_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
special_tokens:
  pad_token: </s>
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: a9bd013b-3193-4c13-8ce7-c6eff4cc9e40
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: a9bd013b-3193-4c13-8ce7-c6eff4cc9e40
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

c0aaf4e6-63a1-4a45-9922-241a2b1c7a1a

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: 2.5669

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: 3060

Training results

Training Loss	Epoch	Step	Validation Loss
4.8067	0.0003	1	4.6919
3.5023	0.0343	100	3.3711
3.2673	0.0686	200	3.1533
3.0786	0.1029	300	3.0437
2.6755	0.1372	400	2.9643
2.8915	0.1715	500	2.9037
2.881	0.2058	600	2.8452
2.9213	0.2401	700	2.8051
2.7981	0.2744	800	2.7714
3.2448	0.3087	900	2.7443
2.9316	0.3431	1000	2.7185
2.9415	0.3774	1100	2.6960
2.53	0.4117	1200	2.6773
2.6615	0.4460	1300	2.6626
2.6712	0.4803	1400	2.6469
2.8272	0.5146	1500	2.6340
2.981	0.5489	1600	2.6243
2.6223	0.5832	1700	2.6147
2.7597	0.6175	1800	2.6052
2.6652	0.6518	1900	2.5975
2.7457	0.6861	2000	2.5915
2.6991	0.7204	2100	2.5853
2.6705	0.7547	2200	2.5808
2.4935	0.7890	2300	2.5770
2.7034	0.8233	2400	2.5737
2.4642	0.8576	2500	2.5711
2.6038	0.8919	2600	2.5694
2.6465	0.9262	2700	2.5682
2.5621	0.9605	2800	2.5674
2.6497	0.9949	2900	2.5670
2.5665	1.0292	3000	2.5669

Framework versions

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