See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: tiiuae/falcon-rw-1b
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 0e47a112d62defbf_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/0e47a112d62defbf_train_data.json
  type:
    field_instruction: instruction
    field_output: response
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 2
early_stopping_threshold: 0.0001
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: romainnn/16e13f52-31b2-405c-ab63-d00182566cf6
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: 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
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 2652
micro_batch_size: 4
mlflow_experiment_name: /tmp/0e47a112d62defbf_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: <|endoftext|>
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: 11e0e99d-9c45-4295-ba54-7587a246e547
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 11e0e99d-9c45-4295-ba54-7587a246e547
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

16e13f52-31b2-405c-ab63-d00182566cf6

This model is a fine-tuned version of tiiuae/falcon-rw-1b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4799

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: 2652

Training results

Training Loss	Epoch	Step	Validation Loss
13.366	0.0007	1	1.7099
4.9212	0.0746	100	0.6743
4.9522	0.1492	200	0.6196
4.7239	0.2238	300	0.5920
4.62	0.2984	400	0.5753
4.5609	0.3730	500	0.5612
4.3202	0.4476	600	0.5496
3.9549	0.5222	700	0.5417
4.3027	0.5968	800	0.5335
3.9765	0.6714	900	0.5280
4.077	0.7460	1000	0.5210
4.1917	0.8206	1100	0.5157
4.4773	0.8952	1200	0.5104
4.3226	0.9698	1300	0.5063
3.7242	1.0444	1400	0.5027
3.6256	1.1190	1500	0.4998
3.7289	1.1936	1600	0.4964
3.8566	1.2682	1700	0.4937
3.6595	1.3428	1800	0.4911
3.5007	1.4174	1900	0.4877
3.7398	1.4920	2000	0.4856
3.9167	1.5666	2100	0.4839
3.8562	1.6412	2200	0.4826
3.4086	1.7158	2300	0.4812
3.5444	1.7904	2400	0.4805
3.4341	1.8650	2500	0.4800
3.8782	1.9396	2600	0.4799

Framework versions

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