See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: elyza/Llama-3-ELYZA-JP-8B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - c6b743fba31532f9_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/c6b743fba31532f9_train_data.json
  type:
    field_instruction: sentence1
    field_output: sentence2
    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/9e3d9f59-6ca7-41b9-95ec-28f8483957fa
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: 1344
micro_batch_size: 4
mlflow_experiment_name: /tmp/c6b743fba31532f9_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: 2048
special_tokens:
  pad_token: <|eot_id|>
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: a596d9c0-8671-4d5a-bc86-1654e251ff18
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: a596d9c0-8671-4d5a-bc86-1654e251ff18
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

9e3d9f59-6ca7-41b9-95ec-28f8483957fa

This model is a fine-tuned version of elyza/Llama-3-ELYZA-JP-8B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7716

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: 1344

Training results

Training Loss	Epoch	Step	Validation Loss
3.5501	0.0004	1	3.5651
1.0198	0.0383	100	0.9737
1.0114	0.0767	200	0.9385
1.0125	0.1150	300	0.9165
0.8194	0.1533	400	0.8836
1.0592	0.1917	500	0.8626
0.5757	0.2300	600	0.8414
0.8539	0.2683	700	0.8244
0.6803	0.3066	800	0.8099
0.6308	0.3450	900	0.7954
0.754	0.3833	1000	0.7857
0.6467	0.4216	1100	0.7770
0.8339	0.4600	1200	0.7727
0.8036	0.4983	1300	0.7716

Framework versions

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