See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: bigscience/bloom-560m
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 90fa039c0864e3ca_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/90fa039c0864e3ca_train_data.json
  type:
    field_input: context
    field_instruction: instruction
    field_output: response
    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/2c03de99-9002-4ac9-8c88-35c7725fb9d3
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:
- query_key_value
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 8832
micro_batch_size: 4
mlflow_experiment_name: /tmp/90fa039c0864e3ca_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: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.01833395668786072
wandb_entity: null
wandb_mode: online
wandb_name: d7df8b7c-3cdd-464a-8a34-49efeb3f7525
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: d7df8b7c-3cdd-464a-8a34-49efeb3f7525
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

2c03de99-9002-4ac9-8c88-35c7725fb9d3

This model is a fine-tuned version of bigscience/bloom-560m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.0593

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: 8832

Training results

Training Loss	Epoch	Step	Validation Loss
20.3231	0.0001	1	2.4185
17.2285	0.0120	100	2.0986
16.5696	0.0239	200	2.0902
16.1132	0.0359	300	2.0848
17.11	0.0478	400	2.0771
18.0987	0.0598	500	2.0744
16.3503	0.0717	600	2.0742
16.5062	0.0837	700	2.0718
15.716	0.0956	800	2.0683
15.7055	0.1076	900	2.0659
14.5481	0.1195	1000	2.0690
17.5521	0.1315	1100	2.0632
15.8487	0.1434	1200	2.0599
16.4771	0.1554	1300	2.0610
17.0074	0.1673	1400	2.0564
16.241	0.1793	1500	2.0572
15.6472	0.1912	1600	2.0593

Framework versions

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