See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.2-3B-Instruct
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - dialogsum_train_data.json
  ds_type: json
  path: /workspace/input_data/dialogsum_train_data.json
  type:
    field_input: dialogue
    field_instruction: topic
    field_output: summary
    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: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: masatochi/tuning-3150dc33-fd7b-4e78-8439-914c8ea061dd
hub_strategy: checkpoint
hub_token: null
learning_rate: 2.0e-05
load_in_4bit: false
load_in_8bit: true
local_rank: null
logging_steps: 1
lora_alpha: 32
lora_dropout: 0.06
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_steps: 10
micro_batch_size: 2
mlflow_experiment_name: /tmp/dialogsum_train_data.json
model_type: LlamaForCausalLM
num_epochs: 20
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: 5
save_strategy: steps
sequence_len: 4096
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
val_set_size: 0.05
wandb_entity: lkotbimehdi
wandb_mode: online
wandb_project: lko
wandb_run: miner_id_24
wandb_runid: 3150dc33-fd7b-4e78-8439-914c8ea061dd
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

tuning-3150dc33-fd7b-4e78-8439-914c8ea061dd

This model is a fine-tuned version of unsloth/Llama-3.2-3B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.2505

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: 2e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• training_steps: 10

Training results

Training Loss	Epoch	Step	Validation Loss
2.6406	0.0006	1	2.2871
2.4788	0.0013	2	2.2882
2.1601	0.0019	3	2.2878
2.4547	0.0026	4	2.2855
2.1888	0.0032	5	2.2844
2.3581	0.0039	6	2.2819
2.1773	0.0045	7	2.2760
2.5797	0.0052	8	2.2711
2.486	0.0058	9	2.2616
2.4162	0.0065	10	2.2505

Framework versions

• PEFT 0.13.2
• Transformers 4.45.2
• Pytorch 2.4.1+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1