See axolotl config

axolotl version: 0.4.1

adapter: lora
auto_find_batch_size: true
base_model: 01-ai/Yi-1.5-9B-Chat-16K
bf16: auto
chat_template: llama3
dataloader_num_workers: 12
dataset_prepared_path: null
datasets:
- data_files:
  - 7a9b7e93517dd03f_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/7a9b7e93517dd03f_train_data.json
  type:
    field_instruction: prompt
    field_output: chosen
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 3
early_stopping_threshold: 0.001
eval_max_new_tokens: 128
eval_steps: 40
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 2
gradient_checkpointing: false
group_by_length: false
hub_model_id: mrferr3t/ef4f4295-87ed-4267-902b-1f09fd90db29
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0003
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 100
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 8
lora_target_linear: true
lr_scheduler: cosine
micro_batch_size: 32
mlflow_experiment_name: /tmp/7a9b7e93517dd03f_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 50
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
s2_attention: null
sample_packing: false
save_steps: 40
saves_per_epoch: 0
sequence_len: 512
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.02
wandb_entity: null
wandb_mode: online
wandb_name: 1600e4aa-9898-4b90-be27-589afaed7e49
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 1600e4aa-9898-4b90-be27-589afaed7e49
warmup_ratio: 0.05
weight_decay: 0.0
xformers_attention: null

ef4f4295-87ed-4267-902b-1f09fd90db29

This model is a fine-tuned version of 01-ai/Yi-1.5-9B-Chat-16K on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2497

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: 0.0003
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use adamw_bnb_8bit 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: 1570
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0002	1	1.1487
No log	0.0080	40	1.1022
No log	0.0159	80	0.6008
0.9502	0.0239	120	0.3673
0.9502	0.0318	160	0.3115
0.345	0.0398	200	0.2932
0.345	0.0478	240	0.2880
0.345	0.0557	280	0.2814
0.297	0.0637	320	0.2760
0.297	0.0716	360	0.2711
0.294	0.0796	400	0.2678
0.294	0.0876	440	0.2693
0.294	0.0955	480	0.2652
0.2745	0.1035	520	0.2626
0.2745	0.1115	560	0.2638
0.2658	0.1194	600	0.2579
0.2658	0.1274	640	0.2578
0.2658	0.1353	680	0.2589
0.2691	0.1433	720	0.2555
0.2691	0.1513	760	0.2574
0.2623	0.1592	800	0.2528
0.2623	0.1672	840	0.2527
0.2623	0.1751	880	0.2508
0.2641	0.1831	920	0.2538
0.2641	0.1911	960	0.2561
0.2563	0.1990	1000	0.2477
0.2563	0.2070	1040	0.2579
0.2563	0.2149	1080	0.2607
0.2574	0.2229	1120	0.2497

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.3.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1