See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen1.5-1.8B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 0d6130b4361297bb_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/0d6130b4361297bb_train_data.json
  type:
    field_instruction: question
    field_output: best
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
device_map:
  ? ''
  : 0,1,2,3,4,5,6,7
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/c044727e-e470-467b-aebb-b93921cd17fd
hub_repo: null
hub_strategy: null
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: 0.1
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
- o_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 3600
micro_batch_size: 4
mlflow_experiment_name: /tmp/0d6130b4361297bb_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: 0.049281476078771515
wandb_entity: null
wandb_mode: online
wandb_name: d4298f6d-583c-4a9a-8925-4c55897e1909
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: d4298f6d-583c-4a9a-8925-4c55897e1909
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

c044727e-e470-467b-aebb-b93921cd17fd

This model is a fine-tuned version of Qwen/Qwen1.5-1.8B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.7053

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: 3600

Training results

Training Loss	Epoch	Step	Validation Loss
3.4815	0.0003	1	3.3099
2.8005	0.0332	100	3.0540
2.5224	0.0663	200	3.0093
2.9589	0.0995	300	2.9760
2.8529	0.1327	400	2.9574
2.7762	0.1659	500	2.9367
2.9447	0.1990	600	2.9140
2.9863	0.2322	700	2.9024
2.5342	0.2654	800	2.8875
3.1029	0.2986	900	2.8697
2.7511	0.3317	1000	2.8566
2.9844	0.3649	1100	2.8466
2.7465	0.3981	1200	2.8402
2.826	0.4313	1300	2.8244
2.809	0.4644	1400	2.8114
2.601	0.4976	1500	2.8005
2.564	0.5308	1600	2.7929
2.8549	0.5640	1700	2.7809
2.7273	0.5971	1800	2.7703
2.6187	0.6303	1900	2.7635
2.4381	0.6635	2000	2.7537
2.8838	0.6967	2100	2.7463
2.6373	0.7298	2200	2.7382
2.6408	0.7630	2300	2.7313
2.578	0.7962	2400	2.7244
2.5839	0.8294	2500	2.7194
2.4858	0.8625	2600	2.7135
2.587	0.8957	2700	2.7080
2.9381	0.9289	2800	2.7040
2.6901	0.9621	2900	2.7006
2.5707	0.9952	3000	2.6978
2.7988	1.0284	3100	2.7044
2.7777	1.0616	3200	2.7053

Framework versions

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