See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: DeepMount00/Llama-3-8b-Ita
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - c27ad72042f3def8_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/c27ad72042f3def8_train_data.json
  type:
    field_instruction: inputs
    field_output: targets
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 4
flash_attention: false
fp16: true
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 6
gradient_checkpointing: true
group_by_length: false
hub_model_id: dimasik2987/085cc89c-8663-43ff-ae30-3e0b4ab741e3
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 32
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_memory:
  0: 70GiB
max_steps: 50
micro_batch_size: 4
mlflow_experiment_name: /tmp/c27ad72042f3def8_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_torch
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 25
save_strategy: steps
sequence_len: 2028
special_tokens:
  pad_token: <|eot_id|>
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 085cc89c-8663-43ff-ae30-3e0b4ab741e3
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 085cc89c-8663-43ff-ae30-3e0b4ab741e3
warmup_steps: 10
weight_decay: 0.01
xformers_attention: null

085cc89c-8663-43ff-ae30-3e0b4ab741e3

This model is a fine-tuned version of DeepMount00/Llama-3-8b-Ita on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7518

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_TORCH 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss
1.2567	0.0316	1	1.2329
1.0618	0.1579	5	1.0954
0.8714	0.3158	10	0.8887
0.9889	0.4737	15	0.8109
0.7376	0.6316	20	0.7796
0.8079	0.7895	25	0.7661
0.7532	0.9474	30	0.7561
0.7877	1.1053	35	0.7507
0.7139	1.2632	40	0.7521
0.7842	1.4211	45	0.7522
0.8305	1.5789	50	0.7518

Framework versions

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