这是使用Qwen2.5-14B-Instruct-GPTQ-Int8 为基底，使用 spicy-3.1 数据训练出来的LoRA

qwen2.5-14B-scipy

This model is a fine-tuned version of /root/LLaMA-Factory/models/Qwen2.5-14B-Instruct-GPTQ-Int8 on the airboros-31_en and the airboros-31_zh datasets.

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 4
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
num_epochs: 1.0
mixed_precision_training: Native AMP

Training results

{ "epoch": 0.9997864616698697, "num_input_tokens_seen": 74083488, "total_flos": 4.6864422704480256e+17, "train_loss": 0.692321076499046, "train_runtime": 65496.9949, "train_samples_per_second": 1.144, "train_steps_per_second": 0.036 }

Framework versions

PEFT 0.12.0
Transformers 4.45.2
Pytorch 2.3.0+cu121
Datasets 2.21.0
Tokenizers 0.20.1