metadata
base_model: meta-llama/Llama-2-7b-hf
tags:
- generated_from_trainer
datasets:
- glue
metrics:
- accuracy
- f1
model-index:
- name: Llama-2-7b-hf-finetuned-mrpc-v5
results: []
library_name: peft
Llama-2-7b-hf-finetuned-mrpc-v5
This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the glue dataset. It achieves the following results on the evaluation set:
- Loss: 0.5514
- Accuracy: 0.8284
- F1: 0.8776
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
The following bitsandbytes quantization config was used during training:
- load_in_8bit: True
- load_in_4bit: False
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: fp4
- bnb_4bit_use_double_quant: False
- bnb_4bit_compute_dtype: float32
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 9
Training results
| Training Loss | Epoch | Step | Accuracy | F1 | Validation Loss |
|---|---|---|---|---|---|
| 0.733 | 1.0 | 917 | 0.6912 | 0.7974 | 0.6016 |
| 0.6103 | 2.0 | 1834 | 0.7402 | 0.8339 | 0.5650 |
| 0.508 | 3.0 | 2751 | 0.7525 | 0.8358 | 0.5246 |
| 0.5354 | 4.0 | 3668 | 0.7794 | 0.8529 | 0.5318 |
| 0.4246 | 5.0 | 4585 | 0.7843 | 0.8508 | 0.5279 |
| 0.4295 | 6.0 | 5502 | 0.7966 | 0.8591 | 0.5248 |
| 0.4473 | 7.0 | 6419 | 0.5169 | 0.8162 | 0.8696 |
| 0.419 | 8.0 | 7336 | 0.5552 | 0.8260 | 0.8778 |
| 0.3876 | 9.0 | 8253 | 0.5514 | 0.8284 | 0.8776 |
Framework versions
- PEFT 0.4.0
- Transformers 4.31.0
- Pytorch 2.0.1+cu118
- Datasets 2.14.5
- Tokenizers 0.13.3