See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2.5-1.5B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 5f268db15ffc7212_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/5f268db15ffc7212_train_data.json
  type:
    field_input: input
    field_instruction: instruction
    field_output: output
    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: 150
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/12b3db6d-6d65-435d-8a25-2864637b169b
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: 0.3
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: 5040
micro_batch_size: 2
mlflow_experiment_name: /tmp/5f268db15ffc7212_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: 150
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.04469673266884191
wandb_entity: null
wandb_mode: online
wandb_name: e613d36b-a8a6-4953-9f33-ddca7d87fc98
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: e613d36b-a8a6-4953-9f33-ddca7d87fc98
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

12b3db6d-6d65-435d-8a25-2864637b169b

This model is a fine-tuned version of Qwen/Qwen2.5-1.5B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.9663

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 5040

Training results

Training Loss	Epoch	Step	Validation Loss
2.526	0.0001	1	2.4610
2.0932	0.0112	150	2.3241
1.9923	0.0225	300	2.2711
2.5518	0.0337	450	2.2464
2.2405	0.0449	600	2.2287
2.7935	0.0561	750	2.1983
2.1035	0.0674	900	2.1862
2.3212	0.0786	1050	2.1668
1.9817	0.0898	1200	2.1552
2.1509	0.1011	1350	2.1452
2.0084	0.1123	1500	2.1316
1.9503	0.1235	1650	2.1120
1.9088	0.1347	1800	2.0998
2.4338	0.1460	1950	2.0879
1.5782	0.1572	2100	2.0774
2.451	0.1684	2250	2.0680
2.1608	0.1797	2400	2.0558
1.9455	0.1909	2550	2.0448
1.9849	0.2021	2700	2.0360
2.1229	0.2134	2850	2.0262
2.0596	0.2246	3000	2.0166
2.0005	0.2358	3150	2.0094
2.5008	0.2470	3300	2.0030
1.951	0.2583	3450	1.9947
2.2078	0.2695	3600	1.9886
2.0641	0.2807	3750	1.9838
2.1594	0.2920	3900	1.9794
2.405	0.3032	4050	1.9754
1.5995	0.3144	4200	1.9723
1.9719	0.3256	4350	1.9696
1.6564	0.3369	4500	1.9680
1.5252	0.3481	4650	1.9671
1.5916	0.3593	4800	1.9666
1.7533	0.3706	4950	1.9663

Framework versions

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