See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2-0.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - aa33efdcea3f7395_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/aa33efdcea3f7395_train_data.json
  type:
    field_instruction: question
    field_output: answer
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 150
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 16
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/299af3f6-7367-4acd-8d81-81b0719cb2e4
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: 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
lr_scheduler: cosine
max_steps: 2384
micro_batch_size: 4
mlflow_experiment_name: /tmp/aa33efdcea3f7395_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
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: 300
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 694298ba-c6cc-4345-b0b4-84c5983d0048
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 694298ba-c6cc-4345-b0b4-84c5983d0048
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

299af3f6-7367-4acd-8d81-81b0719cb2e4

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

• Loss: 1.8383

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: 16
• total_train_batch_size: 64
• 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: 2384

Training results

Training Loss	Epoch	Step	Validation Loss
2.3285	0.0004	1	2.2895
1.8466	0.0569	150	1.9426
1.8839	0.1139	300	1.9219
1.9803	0.1708	450	1.9068
1.8772	0.2277	600	1.8959
1.9223	0.2847	750	1.8859
1.8185	0.3416	900	1.8769
1.8538	0.3985	1050	1.8689
1.8921	0.4554	1200	1.8628
1.6638	0.5124	1350	1.8562
1.8098	0.5693	1500	1.8506
1.8962	0.6262	1650	1.8466
1.8684	0.6832	1800	1.8429
1.8269	0.7401	1950	1.8403
1.9265	0.7970	2100	1.8389
1.7941	0.8540	2250	1.8383

Framework versions

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