See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2-0.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 793d7a595eea5026_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/793d7a595eea5026_train_data.json
  type:
    field_instruction: inputs
    field_output: targets
    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: 100
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: romainnn/ec82bdf0-1c38-4767-90a4-49691d3946fb
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_grad_norm: 1.0
max_steps: 7680
micro_batch_size: 4
mlflow_experiment_name: /tmp/793d7a595eea5026_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 2
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: 100
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.024532412223029067
wandb_entity: null
wandb_mode: online
wandb_name: 3ac4752d-41eb-417c-8944-fd4d6582c896
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 3ac4752d-41eb-417c-8944-fd4d6582c896
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

ec82bdf0-1c38-4767-90a4-49691d3946fb

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

• Loss: 1.7272

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: 8
• total_train_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: 10
• training_steps: 7680

Training results

Training Loss	Epoch	Step	Validation Loss
3.1051	0.0002	1	2.4445
2.6218	0.0161	100	2.2324
1.903	0.0322	200	2.1711
1.8631	0.0483	300	2.1264
2.4634	0.0644	400	2.0912
2.1012	0.0805	500	2.0665
2.3333	0.0966	600	2.0488
1.9606	0.1127	700	2.0289
1.9752	0.1288	800	2.0109
1.5932	0.1449	900	1.9970
2.1825	0.1610	1000	1.9857
1.9413	0.1771	1100	1.9693
2.2342	0.1931	1200	1.9583
1.676	0.2092	1300	1.9447
1.8041	0.2253	1400	1.9370
1.5181	0.2414	1500	1.9274
1.8414	0.2575	1600	1.9187
1.523	0.2736	1700	1.9130
2.1404	0.2897	1800	1.9028
2.13	0.3058	1900	1.8954
2.3082	0.3219	2000	1.8864
1.878	0.3380	2100	1.8819
1.8234	0.3541	2200	1.8729
1.5926	0.3702	2300	1.8680
1.8501	0.3863	2400	1.8604
1.8567	0.4024	2500	1.8563
1.6729	0.4185	2600	1.8532
2.2024	0.4346	2700	1.8461
1.9052	0.4507	2800	1.8366
1.9378	0.4668	2900	1.8349
1.878	0.4829	3000	1.8296
1.9213	0.4990	3100	1.8231
1.4924	0.5151	3200	1.8204
2.0513	0.5312	3300	1.8152
2.0806	0.5473	3400	1.8115
1.5045	0.5633	3500	1.8048
1.5841	0.5794	3600	1.8029
1.7065	0.5955	3700	1.8000
1.4571	0.6116	3800	1.7957
2.0094	0.6277	3900	1.7914
1.6805	0.6438	4000	1.7880
2.3777	0.6599	4100	1.7844
1.8218	0.6760	4200	1.7801
1.7424	0.6921	4300	1.7753
1.4086	0.7082	4400	1.7714
1.4387	0.7243	4500	1.7692
1.518	0.7404	4600	1.7648
1.3727	0.7565	4700	1.7627
2.0031	0.7726	4800	1.7599
1.5169	0.7887	4900	1.7564
1.3817	0.8048	5000	1.7543
1.5441	0.8209	5100	1.7521
1.5501	0.8370	5200	1.7496
1.3207	0.8531	5300	1.7469
1.8306	0.8692	5400	1.7454
2.0758	0.8853	5500	1.7427
1.7154	0.9014	5600	1.7415
1.4177	0.9174	5700	1.7395
1.551	0.9335	5800	1.7378
1.5811	0.9496	5900	1.7364
1.4597	0.9657	6000	1.7347
1.3535	0.9818	6100	1.7339
1.4687	0.9979	6200	1.7321
1.6851	1.0141	6300	1.7316
1.419	1.0302	6400	1.7312
1.974	1.0463	6500	1.7305
1.6064	1.0624	6600	1.7301
1.5348	1.0785	6700	1.7296
1.6983	1.0946	6800	1.7288
2.0455	1.1107	6900	1.7282
1.5807	1.1268	7000	1.7278
1.5712	1.1428	7100	1.7275
1.7076	1.1589	7200	1.7274
1.3702	1.1750	7300	1.7272
1.5782	1.1911	7400	1.7271
1.5359	1.2072	7500	1.7271
1.6834	1.2233	7600	1.7272

Framework versions

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