base_model: Qwen/Qwen2-0.5B
batch_size: 32
bf16: true
chat_template: tokenizer_default_fallback_alpaca
datasets:
- data_files:
  - 745d2d05aaed18f4_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/745d2d05aaed18f4_train_data.json
  type:
    field_input: pos
    field_instruction: task
    field_output: query
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
eval_steps: 20
flash_attention: true
gpu_memory_limit: 80GiB
gradient_checkpointing: true
group_by_length: true
hub_model_id: willtensora/459779f2-cbce-4ec0-b11c-1dcdf92498d8
hub_strategy: checkpoint
learning_rate: 0.0002
logging_steps: 10
lr_scheduler: cosine
max_steps: 2500
micro_batch_size: 4
model_type: AutoModelForCausalLM
optimizer: adamw_bnb_8bit
output_dir: /workspace/axolotl/configs
pad_to_sequence_len: true
resize_token_embeddings_to_32x: false
sample_packing: false
save_steps: 40
save_total_limit: 1
sequence_len: 2048
tokenizer_type: Qwen2TokenizerFast
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.1
wandb_entity: ''
wandb_mode: online
wandb_name: Qwen/Qwen2-0.5B-/workspace/input_data/745d2d05aaed18f4_train_data.json
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: default
warmup_ratio: 0.05
xformers_attention: true

459779f2-cbce-4ec0-b11c-1dcdf92498d8

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: 2.4560

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
distributed_type: multi-GPU
num_devices: 8
total_train_batch_size: 32
total_eval_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: 14
training_steps: 291

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0004	1	3.9660
2.8207	0.0086	20	3.1038
3.1247	0.0172	40	3.0989
2.9411	0.0258	60	2.8986
2.9915	0.0344	80	2.8742
2.8038	0.0430	100	2.8405
2.8518	0.0516	120	2.7728
2.7079	0.0602	140	2.6985
2.6076	0.0688	160	2.6416
2.6172	0.0774	180	2.5695
2.552	0.0860	200	2.5151
2.5036	0.0946	220	2.4783
2.4887	0.1032	240	2.4610
2.4008	0.1118	260	2.4569
2.424	0.1204	280	2.4560

Framework versions

Transformers 4.46.0
Pytorch 2.5.0+cu124
Datasets 3.0.1
Tokenizers 0.20.1