See axolotl config

axolotl version: 0.4.1

adapter: lora
auto_find_batch_size: false
base_model: unsloth/Qwen2.5-3B
bf16: auto
chat_template: llama3
dataloader_num_workers: 12
dataset_prepared_path: null
datasets:
- data_files:
  - cbfc164a7ba5dfd4_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/cbfc164a7ba5dfd4_train_data.json
  type:
    field_instruction: input persona
    field_output: synthesized text
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 1000
early_stopping_threshold: 1.0e-07
eval_max_new_tokens: 128
eval_steps: 220
eval_strategy: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 2
gradient_checkpointing: false
group_by_length: false
hub_model_id: mrferr3t/e9d760b4-d8b5-4635-8832-ae9ca64a1eb9
hub_repo: null
hub_strategy: all_checkpoints
hub_token: null
learning_rate: 0.00015
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 176
lora_alpha: 64
lora_dropout: 0.15
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1
max_steps: 17480
micro_batch_size: 4
mlflow_experiment_name: /tmp/cbfc164a7ba5dfd4_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 200
optimizer: adamw_torch_fused
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 220
save_total_limit: 10
saves_per_epoch: 0
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
use_rslora: true
val_set_size: .01886792
wandb_entity: null
wandb_mode: disabled
wandb_name: a18f4630-bcbf-426d-863a-da31c2f7c188
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: a18f4630-bcbf-426d-863a-da31c2f7c188
warmup_steps: 1748
weight_decay: 0
xformers_attention: null

e9d760b4-d8b5-4635-8832-ae9ca64a1eb9

This model is a fine-tuned version of unsloth/Qwen2.5-3B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7828

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: 0.00015
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• 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: 1748
• training_steps: 17480

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0002	1	1.0805
0.9316	0.0363	220	0.7945
0.7715	0.0726	440	0.7721
0.7584	0.1089	660	0.7638
0.7335	0.1453	880	0.7561
0.7493	0.1816	1100	0.7526
0.7347	0.2179	1320	0.7502
0.7387	0.2542	1540	0.7501
0.7484	0.2905	1760	0.7490
0.7402	0.3268	1980	0.7442
0.7434	0.3631	2200	0.7446
0.7295	0.3994	2420	0.7426
0.7302	0.4358	2640	0.7421
0.7196	0.4721	2860	0.7412
0.7154	0.5084	3080	0.7393
0.731	0.5447	3300	0.7378
0.7342	0.5810	3520	0.7371
0.7227	0.6173	3740	0.7346
0.7258	0.6536	3960	0.7341
0.7292	0.6899	4180	0.7310
0.7278	0.7263	4400	0.7326
0.7174	0.7626	4620	0.7306
0.7239	0.7989	4840	0.7291
0.7149	0.8352	5060	0.7270
0.7171	0.8715	5280	0.7275
0.7135	0.9078	5500	0.7261
0.711	0.9441	5720	0.7245
0.7224	0.9804	5940	0.7241
0.6513	1.0168	6160	0.7333
0.6055	1.0531	6380	0.7333
0.6099	1.0894	6600	0.7345
0.5935	1.1257	6820	0.7376
0.6052	1.1620	7040	0.7367
0.5997	1.1983	7260	0.7332
0.5992	1.2346	7480	0.7328
0.5981	1.2709	7700	0.7314
0.5967	1.3073	7920	0.7336
0.6002	1.3436	8140	0.7294
0.599	1.3799	8360	0.7316
0.5949	1.4162	8580	0.7308
0.5966	1.4525	8800	0.7291
0.6066	1.4888	9020	0.7267
0.6041	1.5251	9240	0.7257
0.5982	1.5614	9460	0.7287
0.591	1.5978	9680	0.7301
0.5931	1.6341	9900	0.7236
0.6101	1.6704	10120	0.7237
0.6068	1.7067	10340	0.7257
0.6055	1.7430	10560	0.7213
0.6023	1.7793	10780	0.7212
0.5991	1.8156	11000	0.7205
0.5993	1.8519	11220	0.7205
0.5858	1.8883	11440	0.7191
0.5958	1.9246	11660	0.7172
0.5869	1.9609	11880	0.7187
0.5867	1.9972	12100	0.7163
0.4452	2.0335	12320	0.7713
0.4373	2.0698	12540	0.7774
0.447	2.1061	12760	0.7740
0.4382	2.1424	12980	0.7827
0.442	2.1788	13200	0.7790
0.4346	2.2151	13420	0.7822
0.4335	2.2514	13640	0.7789
0.4393	2.2877	13860	0.7840
0.4315	2.3240	14080	0.7841
0.4459	2.3603	14300	0.7816
0.4324	2.3966	14520	0.7805
0.4364	2.4329	14740	0.7841
0.4356	2.4693	14960	0.7814
0.442	2.5056	15180	0.7814
0.4323	2.5419	15400	0.7836
0.4393	2.5782	15620	0.7828
0.4259	2.6145	15840	0.7836
0.4455	2.6508	16060	0.7831
0.4318	2.6871	16280	0.7823
0.4318	2.7234	16500	0.7825
0.4376	2.7598	16720	0.7831
0.4377	2.7961	16940	0.7831
0.433	2.8324	17160	0.7828
0.4449	2.8687	17380	0.7828

Framework versions

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