See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2-0.5B
bf16: true
chat_template: llama3
cosine_min_lr_ratio: 0.3
dataset_prepared_path: null
datasets:
- data_files:
  - ce58c478eddedd80_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/ce58c478eddedd80_train_data.json
  type:
    field_input: document_description
    field_instruction: document_type
    field_output: generated_text
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 4
eval_max_new_tokens: 128
eval_steps: 200
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/aa0dee0e-5c38-46e4-bb36-c4f4412de45f
hub_repo: null
hub_strategy: checkpoint
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: 128
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 9439
micro_batch_size: 4
mlflow_experiment_name: /tmp/ce58c478eddedd80_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: 200
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: 591dc9f4-5db4-495e-9e98-7714b16fa4a5
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 591dc9f4-5db4-495e-9e98-7714b16fa4a5
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

aa0dee0e-5c38-46e4-bb36-c4f4412de45f

This model is a fine-tuned version of Qwen/Qwen2-0.5B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9554

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: 4
• total_train_batch_size: 16
• 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: 9439

Training results

Training Loss	Epoch	Step	Validation Loss
1.8083	0.0003	1	1.7581
1.2581	0.0605	200	1.3239
1.4061	0.1209	400	1.2469
1.2636	0.1814	600	1.2025
1.066	0.2418	800	1.1714
1.2745	0.3023	1000	1.1459
0.9485	0.3627	1200	1.1273
1.1605	0.4232	1400	1.1103
1.1213	0.4836	1600	1.0942
0.9332	0.5441	1800	1.0831
1.1496	0.6045	2000	1.0725
1.2126	0.6650	2200	1.0613
1.0792	0.7254	2400	1.0537
1.0194	0.7859	2600	1.0442
0.85	0.8463	2800	1.0372
1.0499	0.9068	3000	1.0277
0.9282	0.9672	3200	1.0204
0.7726	1.0277	3400	1.0192
0.9746	1.0881	3600	1.0161
0.8028	1.1486	3800	1.0106
0.7984	1.2090	4000	1.0061
0.9615	1.2695	4200	0.9987
0.8726	1.3299	4400	0.9946
1.0579	1.3904	4600	0.9887
0.8278	1.4508	4800	0.9836
0.9859	1.5113	5000	0.9799
0.9021	1.5717	5200	0.9751
0.8576	1.6322	5400	0.9701
0.766	1.6926	5600	0.9660
0.7905	1.7531	5800	0.9620
0.7745	1.8135	6000	0.9571
0.8148	1.8740	6200	0.9534
0.7729	1.9344	6400	0.9491
0.8221	1.9949	6600	0.9450
0.7334	2.0553	6800	0.9620
0.7732	2.1158	7000	0.9587
0.8123	2.1762	7200	0.9566
0.7535	2.2367	7400	0.9554

Framework versions

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