See axolotl config

axolotl version: 0.4.1

adapter: lora
auto_find_batch_size: true
base_model: Qwen/Qwen2.5-1.5B
bf16: auto
chat_template: llama3
dataloader_num_workers: 12
dataset_prepared_path: null
datasets:
- data_files:
  - 30467321b0218499_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/30467321b0218499_train_data.json
  type:
    field_input: Japanese
    field_instruction: ENName
    field_output: English
    format: '{instruction} {input}'
    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/ce1a2005-ee52-45ee-8f32-03027c38b14f
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/30467321b0218499_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: f01f50dd-806e-4a8f-9c25-300f4646bd3d
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: f01f50dd-806e-4a8f-9c25-300f4646bd3d
warmup_ratio: 0.05
weight_decay: 0.0
xformers_attention: null

ce1a2005-ee52-45ee-8f32-03027c38b14f

This model is a fine-tuned version of Qwen/Qwen2.5-1.5B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.4814

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: 1299
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0010	1	3.1418
No log	0.0385	40	3.0617
No log	0.0770	80	2.3578
2.7405	0.1154	120	1.9472
2.7405	0.1539	160	1.8134
1.8729	0.1924	200	1.7430
1.8729	0.2309	240	1.6926
1.8729	0.2694	280	1.6552
1.6979	0.3078	320	1.6294
1.6979	0.3463	360	1.5977
1.6043	0.3848	400	1.5806
1.6043	0.4233	440	1.5645
1.6043	0.4618	480	1.5476
1.5427	0.5002	520	1.5395
1.5427	0.5387	560	1.5332
1.5321	0.5772	600	1.5244
1.5321	0.6157	640	1.5119
1.5321	0.6542	680	1.5102
1.5327	0.6926	720	1.5054
1.5327	0.7311	760	1.5002
1.5065	0.7696	800	1.5024
1.5065	0.8081	840	1.4940
1.5065	0.8466	880	1.4840
1.5024	0.8850	920	1.4788
1.5024	0.9235	960	1.4758
1.4772	0.9620	1000	1.4725
1.4772	1.0005	1040	1.4676
1.4772	1.0390	1080	1.4773
1.4016	1.0774	1120	1.4743
1.4016	1.1159	1160	1.4814

Framework versions

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