Mixtral_Rio_oasst2_v1

This model is a fine-tuned version of mistralai/Mixtral-8x7B-v0.1 on the dataset oasst2 (OpenAssistant). It achieves the following results on the evaluation set:

• eval_loss: 1.0570
• eval_runtime: 2.8042
• eval_samples_per_second: 3.566
• eval_steps_per_second: 0.713
• epoch: 30.0
• step: 150

Model description

This is a LoRA trained on OpenAssistant data. The settings for the base model should be: Model loader: Transformers Compute_dtype: bfloat16 quant_type: nf4 cpu: enabled load-in-4bit: enabled use_double_quant: enabled set GPU memory as high as possible unless running locally to give some space for your desktop environment tweak CPU usage until it loads successfully

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: 2.5e-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.03
• training_steps: 175

Framework versions

• PEFT 0.8.2
• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1