See axolotl config

axolotl version: 0.4.1

adapter: lora
auto_find_batch_size: true
base_model: defog/llama-3-sqlcoder-8b
bf16: auto
chat_template: llama3
dataloader_num_workers: 12
dataset_prepared_path: null
datasets:
- data_files:
  - 4d3ccfd148a48951_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/4d3ccfd148a48951_train_data.json
  type:
    field_instruction: question
    field_output: positive_answer
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 3
early_stopping_threshold: 0.001
eval_max_new_tokens: 128
eval_steps: 40
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 2
gradient_checkpointing: false
group_by_length: false
hub_model_id: mrferr3t/08668496-5e8f-473f-8969-e5e7f7f05c61
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0003
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 100
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 8
lora_target_linear: true
lr_scheduler: cosine
micro_batch_size: 32
mlflow_experiment_name: /tmp/4d3ccfd148a48951_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 50
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
s2_attention: null
sample_packing: false
save_steps: 40
saves_per_epoch: 0
sequence_len: 512
special_tokens:
  pad_token: <|eot_id|>
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.02
wandb_entity: null
wandb_mode: online
wandb_name: fbae069e-fb6e-483d-98a1-0b88f4497a85
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: fbae069e-fb6e-483d-98a1-0b88f4497a85
warmup_ratio: 0.05
weight_decay: 0.0
xformers_attention: null

08668496-5e8f-473f-8969-e5e7f7f05c61

This model is a fine-tuned version of defog/llama-3-sqlcoder-8b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6502

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: 0.0003
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use adamw_bnb_8bit 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: 449
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0007	1	1.2115
No log	0.0278	40	1.0597
No log	0.0556	80	0.9304
1.0735	0.0833	120	0.8850
1.0735	0.1111	160	0.8702
0.8708	0.1389	200	0.8522
0.8708	0.1667	240	0.8398
0.8708	0.1944	280	0.8326
0.8352	0.2222	320	0.8218
0.8352	0.25	360	0.8126
0.8303	0.2778	400	0.8182
0.8303	0.3056	440	0.8048
0.8303	0.3333	480	0.8074
0.831	0.3611	520	0.7934
0.831	0.3889	560	0.7838
0.7971	0.4167	600	0.7793
0.7971	0.4444	640	0.7775
0.7971	0.4722	680	0.7648
0.7748	0.5	720	0.7571
0.7748	0.5278	760	0.7516
0.7664	0.5556	800	0.7503
0.7664	0.5833	840	0.7475
0.7664	0.6111	880	0.7394
0.7595	0.6389	920	0.7341
0.7595	0.6667	960	0.7318
0.7373	0.6944	1000	0.7329
0.7373	0.7222	1040	0.7273
0.7373	0.75	1080	0.7231
0.7192	0.7778	1120	0.7199
0.7192	0.8056	1160	0.7076
0.727	0.8333	1200	0.7006
0.727	0.8611	1240	0.6935
0.727	0.8889	1280	0.6962
0.7096	0.9167	1320	0.6902
0.7096	0.9444	1360	0.6879
0.6914	0.9722	1400	0.6850
0.6914	1.0	1440	0.6824
0.6914	1.0278	1480	0.6757
0.6123	1.0556	1520	0.6694
0.6123	1.0833	1560	0.6713
0.5842	1.1111	1600	0.6688
0.5842	1.1389	1640	0.6696
0.5842	1.1667	1680	0.6716
0.6055	1.1944	1720	0.6644
0.6055	1.2222	1760	0.6659
0.5986	1.25	1800	0.6627
0.5986	1.2778	1840	0.6571
0.5986	1.3056	1880	0.6509
0.5943	1.3333	1920	0.6492
0.5943	1.3611	1960	0.6528
0.5999	1.3889	2000	0.6494
0.5999	1.4167	2040	0.6502

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.3.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1