See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2-7B-Instruct
bf16: true
chat_template: llama3
datasets:
- data_files:
  - 694d91dd756eb2cd_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/694d91dd756eb2cd_train_data.json
  type:
    field_instruction: instruction
    field_output: output
    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: false
fp16: false
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 2
gradient_checkpointing: true
group_by_length: false
hub_model_id: sn56z0/a3b37977-b811-4e81-b2f7-b9ff2469a170
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0001
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
lr_scheduler: cosine
max_memory:
  0: 77GiB
max_steps: 100
micro_batch_size: 8
mlflow_experiment_name: /tmp/694d91dd756eb2cd_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_torch
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 25
save_strategy: steps
seed: 3314771512
sequence_len: 1024
shuffle: true
strict: false
tf32: false
tokenizer_type: AutoTokenizer
torch_compile: true
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: sn56-miner
wandb_mode: disabled
wandb_name: null
wandb_project: god
wandb_run: gneu
wandb_runid: null
warmup_steps: 10
weight_decay: 0.01
xformers_attention: false

a3b37977-b811-4e81-b2f7-b9ff2469a170

This model is a fine-tuned version of Qwen/Qwen2-7B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2853

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 3314771512
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• total_eval_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH 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
0.6711	0.0007	1	0.6915
0.5176	0.0059	9	0.4799
0.2893	0.0117	18	0.3014
0.297	0.0176	27	0.2967
0.3104	0.0234	36	0.2923
0.2716	0.0293	45	0.2898
0.2775	0.0351	54	0.2881
0.2736	0.0410	63	0.2871
0.211	0.0468	72	0.2860
0.2256	0.0527	81	0.2856
0.28	0.0585	90	0.2853
0.2611	0.0644	99	0.2853

Framework versions

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