Code-Llama-3-8B-finetuned-py-to-cpp
This model is a fine-tuned version of ajibawa-2023/Code-Llama-3-8B on the XLCoST (Python-C++) dataset, restricted to code snippets of <= 128 tokens long. It achieves the following results on the evaluation set:
- Loss: 0.4550
Test set:
- BLEU: 38.48
- COMET: 79.28
- CodeBLEU: 64.11
- N-gram match score: 40.15
- Weighted n-gram match score: 77.11
- Syntax match score: 67.02
- Dataflow match score: 72.13
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: 0.0001
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| No log | 0.99 | 87 | 0.5468 |
| No log | 2.0 | 175 | 0.4626 |
| No log | 2.98 | 261 | 0.4550 |
Framework versions
- Transformers 4.33.1
- Pytorch 2.4.0
- Datasets 3.0.1
- Tokenizers 0.13.3
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for hugo-albert/Code-Llama-3-8B-finetuned-py-to-cpp
Base model
ajibawa-2023/Code-Llama-3-8B