See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: berkeley-nest/Starling-LM-7B-alpha
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - b20fbfa08217066a_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/b20fbfa08217066a_train_data.json
  type:
    field_instruction: question
    field_output: chosen
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 4
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
gradient_clipping: 1.0
group_by_length: false
hub_model_id: brixeus/515ff868-bccd-4d90-a570-e8a5fdaa7424
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 5.0e-05
load_in_4bit: false
load_in_8bit: false
local_rank: 0
logging_steps: 3
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_steps: 100
micro_batch_size: 8
mlflow_experiment_name: /tmp/b20fbfa08217066a_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
saves_per_epoch: 4
sequence_len: 1024
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: techspear-hub
wandb_mode: online
wandb_name: bb16ccd3-964b-4fd6-975d-73e3b09a8079
wandb_project: Gradients-On-Three
wandb_run: your_name
wandb_runid: bb16ccd3-964b-4fd6-975d-73e3b09a8079
warmup_steps: 10
weight_decay: 0.01
xformers_attention: null

515ff868-bccd-4d90-a570-e8a5fdaa7424

This model is a fine-tuned version of berkeley-nest/Starling-LM-7B-alpha on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8891

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 100

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0027	1	1.1643
4.2326	0.0243	9	1.0253
4.0237	0.0486	18	0.9553
3.6917	0.0729	27	0.9295
3.4829	0.0972	36	0.9152
3.9289	0.1215	45	0.9068
3.6489	0.1458	54	0.9003
3.5293	0.1702	63	0.8952
3.3807	0.1945	72	0.8921
3.4407	0.2188	81	0.8902
3.6309	0.2431	90	0.8892
3.8183	0.2674	99	0.8891

Framework versions

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