See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: peft-internal-testing/tiny-dummy-qwen2
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 4740323202236304_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/4740323202236304_train_data.json
  type:
    field_instruction: init_prompt
    field_output: init_response
    format: '{instruction}'
    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: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/645cc86e-e44f-411b-8f52-1b8b15d69691
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:
- q_proj
- k_proj
- v_proj
- o_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 4140
micro_batch_size: 4
mlflow_experiment_name: /tmp/4740323202236304_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: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: db8257b3-78c7-45a9-8135-fd36a7a19299
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: db8257b3-78c7-45a9-8135-fd36a7a19299
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

645cc86e-e44f-411b-8f52-1b8b15d69691

This model is a fine-tuned version of peft-internal-testing/tiny-dummy-qwen2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 11.9043

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: 2625

Training results

Training Loss	Epoch	Step	Validation Loss
11.9297	0.0008	1	11.9307
11.9166	0.0762	100	11.9180
11.9182	0.1524	200	11.9147
11.914	0.2286	300	11.9121
11.9129	0.3048	400	11.9108
11.9134	0.3810	500	11.9097
11.908	0.4571	600	11.9087
11.912	0.5333	700	11.9078
11.9108	0.6095	800	11.9071
11.9089	0.6857	900	11.9066
11.905	0.7619	1000	11.9062
11.91	0.8381	1100	11.9058
11.9062	0.9143	1200	11.9055
11.911	0.9905	1300	11.9054
9.6202	1.0667	1400	11.9051
11.9829	1.1429	1500	11.9049
11.86	1.2190	1600	11.9048
11.5186	1.2952	1700	11.9046
12.3236	1.3714	1800	11.9045
13.1264	1.4476	1900	11.9045
13.5479	1.5238	2000	11.9044
10.3651	1.6	2100	11.9043
11.4862	1.6762	2200	11.9043
11.5082	1.7524	2300	11.9043
12.481	1.8286	2400	11.9043
10.881	1.9048	2500	11.9043
10.0359	1.9810	2600	11.9043

Framework versions

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