See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: fxmarty/really-tiny-falcon-testing
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 3bf2354067fe4af7_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/3bf2354067fe4af7_train_data.json
  type:
    field_input: prompt
    field_instruction: question
    field_output: chosen
    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: false
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/ce770bbd-2877-4d64-8331-884386dbaf3b
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: 0.1
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
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 3060
micro_batch_size: 4
mlflow_experiment_name: /tmp/3bf2354067fe4af7_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: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: 5c60d90b-d7f4-4de5-bf75-084c7f7cf079
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 5c60d90b-d7f4-4de5-bf75-084c7f7cf079
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

ce770bbd-2877-4d64-8331-884386dbaf3b

This model is a fine-tuned version of fxmarty/really-tiny-falcon-testing on the None dataset. It achieves the following results on the evaluation set:

• Loss: 10.9689

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: 2589

Training results

Training Loss	Epoch	Step	Validation Loss
88.6826	0.0008	1	11.0838
88.0259	0.0773	100	11.0001
87.9623	0.1545	200	10.9915
87.91	0.2318	300	10.9877
87.9079	0.3090	400	10.9837
87.8231	0.3863	500	10.9807
87.9996	0.4635	600	10.9787
87.834	0.5408	700	10.9776
87.9084	0.6181	800	10.9758
87.69	0.6953	900	10.9747
87.7913	0.7726	1000	10.9739
87.8704	0.8498	1100	10.9729
87.8838	0.9271	1200	10.9722
87.8234	1.0043	1300	10.9717
87.8447	1.0816	1400	10.9710
87.8522	1.1589	1500	10.9707
87.7542	1.2361	1600	10.9702
87.7928	1.3134	1700	10.9699
87.9173	1.3906	1800	10.9696
87.8323	1.4679	1900	10.9695
87.7511	1.5451	2000	10.9692
87.765	1.6224	2100	10.9691
87.8693	1.6997	2200	10.9690
87.8853	1.7769	2300	10.9689
87.7936	1.8542	2400	10.9689
87.7994	1.9314	2500	10.9689

Framework versions

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