See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2.5-0.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - ef537d775e149577_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/ef537d775e149577_train_data.json
  type:
    field_input: document_type
    field_instruction: document_description
    field_output: generated_text
    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: 100
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 6
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/0f21b063-cdeb-49f0-a108-3e731b11cc21
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_best_model_at_end: true
load_in_4bit: true
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: 5376
micro_batch_size: 4
mlflow_experiment_name: /tmp/ef537d775e149577_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: 100
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: 5179b52c-7598-4dc9-8bc6-44c4c1f03590
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 5179b52c-7598-4dc9-8bc6-44c4c1f03590
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

0f21b063-cdeb-49f0-a108-3e731b11cc21

This model is a fine-tuned version of Qwen/Qwen2.5-0.5B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9591

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: 6
• total_train_batch_size: 24
• 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: 5376

Training results

Training Loss	Epoch	Step	Validation Loss
2.0379	0.0005	1	1.9584
1.2714	0.0456	100	1.4167
1.2782	0.0912	200	1.3253
1.1469	0.1368	300	1.2742
1.3074	0.1823	400	1.2352
1.2156	0.2279	500	1.2058
1.1919	0.2735	600	1.1832
1.2372	0.3191	700	1.1655
1.0298	0.3647	800	1.1499
1.0246	0.4103	900	1.1376
1.1558	0.4559	1000	1.1225
1.0212	0.5014	1100	1.1120
1.0905	0.5470	1200	1.1027
1.0495	0.5926	1300	1.0931
1.1128	0.6382	1400	1.0837
1.1127	0.6838	1500	1.0738
1.1927	0.7294	1600	1.0680
1.1662	0.7750	1700	1.0598
1.2109	0.8205	1800	1.0537
0.9344	0.8661	1900	1.0447
1.1805	0.9117	2000	1.0385
0.9501	0.9573	2100	1.0332
1.0264	1.0030	2200	1.0301
1.0236	1.0486	2300	1.0260
0.9103	1.0942	2400	1.0212
0.9896	1.1398	2500	1.0169
0.9263	1.1854	2600	1.0120
0.9147	1.2310	2700	1.0082
1.0366	1.2766	2800	1.0041
0.924	1.3221	2900	0.9999
0.9134	1.3677	3000	0.9955
0.7775	1.4133	3100	0.9903
0.9432	1.4589	3200	0.9867
0.9049	1.5045	3300	0.9829
0.9432	1.5501	3400	0.9793
1.0249	1.5957	3500	0.9760
0.9052	1.6412	3600	0.9732
0.9647	1.6868	3700	0.9700
0.9114	1.7324	3800	0.9675
0.8885	1.7780	3900	0.9649
0.9067	1.8236	4000	0.9630
0.8201	1.8692	4100	0.9608
0.9196	1.9148	4200	0.9593
0.8485	1.9603	4300	0.9572
0.8227	2.0061	4400	0.9571
0.8173	2.0517	4500	0.9599
0.7624	2.0972	4600	0.9600
0.9698	2.1428	4700	0.9593
0.885	2.1884	4800	0.9591

Framework versions

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