Not-so-bright-AGI-3

This model is a fine-tuned version of meta-llama/Meta-Llama-3-70B-Instruct on an unknown 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: 2e-5
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• bf16: True
• num_train_epochs: 5
• per_device_train_batch_size: 8
• evaluation_strategy: "steps"
• save_strategy: "steps"
• learning_rate: 1e-5
• warmup_ratio: 0.03
• lr_scheduler_type: "linear"
• max_grad_norm: 0.01
• logging_steps: 100
• do_train
• do_eval
• use_habana
• use_lazy_mode
• throughput_warmup_steps: 5
• lora_rank: 8
• lora_alpha: 32
• lora_dropout: 0.1
• lora_target_modules: "q_proj" "v_proj"
• dataset_concatenation
• report_to: none
• max_seq_length: 512
• low_cpu_mem_usage: True
• validation_split_percentage: 15
• adam_epsilon: 1e-08

Framework versions

• PEFT 0.6.2
• Transformers 4.38.2
• Pytorch @ file:///tmp/tmp.5raBIUJfCK/torch-2.2.0a0%2Bgit8964477-cp310-cp310-linux_x86_64.whl#sha256=fe3b24f994c5e69f45942fb7def7f7e1b3617c20471f3fcadd16e4c7c85fb697
• Datasets 2.20.0
• Tokenizers 0.15.2