EugeneLam/lihkg-story

model

This model is a fine-tuned version of hon9kon9ize/CantoneseLLM-v0.5 on the lihkg_story_sft_v2 dataset. It achieves the following results on the evaluation set:

• Loss: 2.9276

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: 1e-05
• train_batch_size: 20
• eval_batch_size: 22
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 16
• total_train_batch_size: 1280
• total_eval_batch_size: 88
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1.0

Training results

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cpu
• Datasets 2.19.1
• Tokenizers 0.19.1