See axolotl config

axolotl version: 0.5.2

adapter: lora
auto_find_batch_size: true
base_model: Qwen/Qwen2.5-7B
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - e67a3bba658eaa0f_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/e67a3bba658eaa0f_train_data.json
  type:
    field_input: input
    field_instruction: system_prompt
    field_output: reference_answer
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: /workspace/axolotl/configs/deepspeed_stage2.json
early_stopping_patience: 5
eval_max_new_tokens: 128
eval_sample_packing: false
eval_steps: 50
eval_table_size: null
evaluation_strategy: steps
flash_attention: true
fp16: false
gpu_memory_limit: 80GiB
gradient_accumulation_steps: 2
gradient_checkpointing: true
greater_is_better: false
group_by_length: true
hub_model_id: PhoenixB/c6460c46-b09a-40f9-8e48-e0bc0d299c2e
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 2e-4
liger_fused_linear_cross_entropy: true
liger_glu_activation: true
liger_layer_norm: true
liger_rms_norm: true
liger_rope: true
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 5
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_steps: 10000
metric_for_best_model: eval_loss
micro_batch_size: 2
mlflow_experiment_name: /tmp/e67a3bba658eaa0f_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_torch_fused
output_dir: miner_id_24
pad_to_sequence_len: true
plugins:
- axolotl.integrations.liger.LigerPlugin
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 50
save_total_limit: 1
sequence_len: 8196
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: c4b7f349-24bc-4e90-8602-b3a61dc8b8ab
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: c4b7f349-24bc-4e90-8602-b3a61dc8b8ab
warmup_steps: 20
weight_decay: 0.0

c6460c46-b09a-40f9-8e48-e0bc0d299c2e

This model is a fine-tuned version of Qwen/Qwen2.5-7B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0151

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
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• total_eval_batch_size: 4
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 20
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0001	1	1.3133
1.3321	0.0056	50	1.1406
1.095	0.0111	100	0.9931
0.7319	0.0167	150	0.8583
0.4488	0.0223	200	0.7062
0.5606	0.0279	250	0.6025
0.369	0.0334	300	0.4909
0.3071	0.0390	350	0.4296
0.1691	0.0446	400	0.3548
0.1887	0.0502	450	0.2855
0.2134	0.0557	500	0.2420
0.0553	0.0613	550	0.2010
0.105	0.0669	600	0.1763
0.0427	0.0724	650	0.1375
0.0575	0.0780	700	0.1137
0.1403	0.0836	750	0.0975
0.0601	0.0892	800	0.0804
0.0619	0.0947	850	0.0610
0.0524	0.1003	900	0.0500
0.1153	0.1059	950	0.0399
0.0138	0.1115	1000	0.0333
0.0136	0.1170	1050	0.0298
0.01	0.1226	1100	0.0271
0.0228	0.1282	1150	0.0201
0.0364	0.1337	1200	0.0175
0.0149	0.1393	1250	0.0167
0.0368	0.1449	1300	0.0184
0.0135	0.1505	1350	0.0151
0.0064	0.1560	1400	0.0133
0.0332	0.1616	1450	0.0144
0.009	0.1672	1500	0.0145
0.0073	0.1728	1550	0.0137
0.0122	0.1783	1600	0.0148
0.0057	0.1839	1650	0.0151

Framework versions

• PEFT 0.13.2
• Transformers 4.46.3
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3