ParroT-Hint-7b-lora

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

ParroT-Hint-7b-lora

The LoRA version of ParroT-Hint-7b based on llama-7b.

Inference

Inference based on the LoRA weights looks as below. You may refer to the GitHub project for more details.

# Translation
python3 inference_lora.py --model-name-or-path <your_proj_path>/llama-7b \
    --lora-weights <your_proj_path>/ParroT-Hint-7b-lora/adapter_model \
    -lp 'zh-en' \
    -t 0.1 \
    -sa 'beam' \
    -ins test/instruct_inf.txt \
    -i test/test_rand_50.zh.txt \
    -o test/test_rand_50.zh-en.none-hint.txt
    
# Text generation
python3 inference_lora.py --model-name-or-path <your_proj_path>/llama-7b \
    --lora-weights <your_proj_path>/ParroT-Hint-7b-lora/adapter_model \
    -t 0.7 \
    -sa 'sample' \
    -i test/test_case.txt \
    -o test/test_case.general-task.txt

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 4
• eval_batch_size: 2
• seed: 1
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 1.58

Framework versions

• Transformers 4.28.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.11.0
• Tokenizers 0.13.2