open-llama-Instruct
- This model is a fine-tuned version of meta-llama/Llama-3.2-1B-Instruct on the diabolic6045/OpenHermes-2.5_alpaca_10 dataset. which is 10% of OpenHermes 2.5 Dataset
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
- distributed_type: multi-GPU
- num_devices: 2
- total_train_batch_size: 4
- total_eval_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 10
- num_epochs: 1
- mixed_precision_training: Native AMP
Training results
| Groups | Version | Filter | n-shot | Metric | Value | Stderr | ||
|---|---|---|---|---|---|---|---|---|
| mmlu | 2 | none | acc | ↑ | 0.3632 | ± | 0.0040 | |
| - humanities | 2 | none | acc | ↑ | 0.3411 | ± | 0.0068 | |
| - other | 2 | none | acc | ↑ | 0.4078 | ± | 0.0087 | |
| - social sciences | 2 | none | acc | ↑ | 0.3997 | ± | 0.0087 | |
| - stem | 2 | none | acc | ↑ | 0.3165 | ± | 0.0082 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.1.2
- Datasets 3.0.1
- Tokenizers 0.20.1
axolotl version: 0.4.1
base_model: meta-llama/Llama-3.2-1B-Instruct
load_in_8bit: false
load_in_4bit: false
strict: false
datasets:
- path: diabolic6045/OpenHermes-2.5_alpaca_10
type: alpaca
dataset_prepared_path: last_run_prepared
val_set_size: 0
output_dir: ./outputs/out
hub_model_id: diabolic6045/open-llama-Instruct
hf_use_auth_token: true
sequence_len: 1024
sample_packing: true
pad_to_sequence_len: true
wandb_project: open-llama
wandb_entity:
wandb_watch: all
wandb_name: open-llama
wandb_log_model:
gradient_accumulation_steps: 1
micro_batch_size: 2
num_epochs: 1
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 2e-5
train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false
gradient_checkpointing: true
gradient_checkpointing_kwargs:
use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
logging_steps: 1
xformers_attention:
flash_attention: false
warmup_steps: 10
evals_per_epoch: 2
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
pad_token: <|end_of_text|>