See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: openlm-research/open_llama_3b
bf16: auto
chat_template: llama3
dataset_exact_deduplication: true
datasets:
- data_files:
  - afd945a66e170177_train_data.json
  ds_type: json
  format: custom
  path: /runs/taopanda_645a42bb-5a50-4196-be61-1e184897d93d/afd945a66e170177_train_data.json
  preprocessing:
  - shuffle: true
  type:
    field: null
    field_input: rejected_response
    field_instruction: instruction
    field_output: chosen_response
    field_system: null
    format: null
    no_input_format: null
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
device_map: auto
early_stopping_patience: 6
eval_batch_size: 12
eval_steps: 10
eval_strategy: steps
fp16: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: taopanda/c16f3786-8e5d-4ae9-b571-f7da96e254a8
label_smoothing: 0.08
learning_rate: 0.00014016703367409407
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: 0.033
lora_fan_in_fan_out: true
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 623
micro_batch_size: 12
model_type: AutoModelForCausalLM
num_epochs: 5
optimizer: adamw_torch
output_dir: ./outputs/lora-out/taopanda_645a42bb-5a50-4196-be61-1e184897d93d
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
save_safetensors: true
save_steps: 10
save_total_limit: 7
seed: 94259
sequence_len: 512
special_tokens:
  pad_token: </s>
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.04
wandb_entity: fatcat87-taopanda
wandb_mode: online
wandb_name: taopanda_645a42bb-5a50-4196-be61-1e184897d93d
wandb_project: subnet56
wandb_runid: taopanda_645a42bb-5a50-4196-be61-1e184897d93d
warmup_ratio: 0.05
weight_decay: 0.16
xformers_attention: null

c16f3786-8e5d-4ae9-b571-f7da96e254a8

This model is a fine-tuned version of openlm-research/open_llama_3b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.1916

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: 0.00014016703367409407
• train_batch_size: 12
• eval_batch_size: 12
• seed: 94259
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 192
• total_eval_batch_size: 48
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• training_steps: 4

Training results

Training Loss	Epoch	Step	Validation Loss
1.6895	1.0	1	2.1916

Framework versions

• PEFT 0.11.1
• Transformers 4.43.4
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1