See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2.5-3B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - a4d3a9ffc31ac829_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/a4d3a9ffc31ac829_train_data.json
  type:
    field_instruction: question
    field_output: answer
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
ddp_timeout: 1800
debug: null
deepspeed: null
device_map: auto
do_eval: true
early_stopping_patience: 2
eval_batch_size: 4
eval_max_new_tokens: 128
eval_steps: 150
eval_table_size: null
evals_per_epoch: null
flash_attention: true
fp16: false
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 1
gradient_checkpointing: false
group_by_length: false
hub_model_id: Nexspear/3d172c22-133a-45ef-9a25-3274fa701846
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 4.4e-05
load_in_4bit: false
load_in_8bit: false
local_rank: 0
logging_steps: 10
lora_alpha: 64
lora_dropout: 0.3
lora_fan_in_fan_out: null
lora_model_dir: null
lora_modules_to_save:
- lm_head
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine_with_restarts
max_grad_norm: 1.0
max_memory:
  0: 75GB
max_steps: 400
micro_batch_size: 8
mlflow_experiment_name: /tmp/a4d3a9ffc31ac829_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optim_args:
  adam_beta1: 0.9
  adam_beta2: 0.999
  adam_epsilon: 1e-08
optimizer: adamw_torch
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_safetensors: true
save_steps: 150
save_total_limit: 1
saves_per_epoch: null
seed: 840
sequence_len: 512
strict: false
tf32: null
tokenizer_type: AutoTokenizer
torch_compile: false
torch_compile_backend: null
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: acopia-grant
wandb_mode: online
wandb_name: ebe1e44e-fe15-40da-bb22-79c51f30f4ca
wandb_project: Gradients-On-4
wandb_run: your_name
wandb_runid: ebe1e44e-fe15-40da-bb22-79c51f30f4ca
warmup_ratio: 0.1
warmup_steps: 
weight_decay: 0.01
xformers_attention: null

3d172c22-133a-45ef-9a25-3274fa701846

This model is a fine-tuned version of Qwen/Qwen2.5-3B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.6019

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: 4.4e-05
• train_batch_size: 8
• eval_batch_size: 4
• seed: 840
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=adam_beta1=0.9,adam_beta2=0.999,adam_epsilon=1e-08
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_steps: 40
• training_steps: 400

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0007	1	1.8380
1.6014	0.1094	150	1.6096
1.6359	0.2188	300	1.6019

Framework versions

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