axolotl version: 0.6.0
wandb_run_id: 2e2444b4-b741-48af-b32c-b5f44f38688f
wandb_project: llm-training-platform
wandb_name: II-Tulu-8B-SFT
datasets:
- path: allenai/tulu-3-sft-mixture
split: train
type: chat_template
field_messages: messages
message_field_role: role
message_field_content: content
roles:
system:
- system
user:
- user
assistant:
- assistant
chat_template: llama3
sequence_len: 2048
base_model: meta-llama/Llama-3.1-8B
output_dir: checkpoints/deb3448a-60ae-4ad8-bdc2-06cce8c43d02
dataset_prepared_path: checkpoints/deb3448a-60ae-4ad8-bdc2-06cce8c43d02/dataset_prepared
flash_attention: true
train_on_inputs: false
pad_to_sequence_len: true
eval_sample_packing: false
push_to_hub: true
bf16: auto
logging_steps: 10
hub_model_id: phunguyen01/II-8B-SFT
learning_rate: 5.0e-06
micro_batch_size: 2
num_epochs: 2
seed: 42
gradient_accumulation_steps: 2
sample_packing: true
val_set_size: 0
special_tokens:
pad_token: <|end_of_text|>
II-8B-SFT
This model is a fine-tuned version of meta-llama/Llama-3.1-8B on the allenai/tulu-3-sft-mixture dataset.
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: 5e-06
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- total_eval_batch_size: 16
- optimizer: Use adamw_hf 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: 100
- num_epochs: 2
Training results
Framework versions
- Transformers 4.47.0
- Pytorch 2.4.0+cu121
- Datasets 3.1.0
- Tokenizers 0.21.0