Play with the model on the SantaCoder Space Demo.
This is the same model as SantaCoder but it can be loaded with transformers >=4.28.1 to use the GPTBigCode architecture. We refer the reader to the SantaCoder model page for full documentation about this model
- Repository: bigcode/Megatron-LM
- Project Website: bigcode-project.org
- Paper: 🎅SantaCoder: Don't reach for the stars!🌟
- Point of Contact: email@example.com
There are two versions (branches) of the model:
main: Uses the
gpt_bigcodemodel. Requires the bigcode fork of transformers.
main_custom: Packaged with its modeling code. Requires
transformers>=4.27. Alternatively, it can run on older versions by setting the configuration parameter
activation_function = "gelu_pytorch_tanh".
The model was trained on GitHub code. As such it is not an instruction model and commands like "Write a function that computes the square root." do not work well.
You should phrase commands like they occur in source code such as comments (e.g.
# the following function computes the sqrt) or write a function signature and docstring and let the model complete the function body.
The pretraining dataset of the model was filtered for permissive licenses only. Nevertheless, the model can generate source code verbatim from the dataset. The code's license might require attribution and/or other specific requirements that must be respected. We provide a search index that let's you search through the pretraining data to identify where generated code came from and apply the proper attribution to your code.
- Architecture: GPT-2 model with multi-query attention and Fill-in-the-Middle objective
- Pretraining steps: 600K
- Pretraining tokens: 236 billion
- Precision: float16
- GPUs: 96 Tesla V100
- Training time: 6.2 days
- Total FLOPS: 2.1 x 10e21
The model is licenses under the CodeML Open RAIL-M v0.1 license. You can find the full license here.
- Downloads last month
Inference API has been turned off for this model.
Dataset used to train bigcode/gpt_bigcode-santacoder
Spaces using bigcode/gpt_bigcode-santacoder 8
- pass@1 on MultiPL HumanEval (Python)self-reported0.180
- pass@10 on MultiPL HumanEval (Python)self-reported0.290
- pass@100 on MultiPL HumanEval (Python)self-reported0.490
- pass@1 on MultiPL MBPP (Python)self-reported0.350
- pass@10 on MultiPL MBPP (Python)self-reported0.580
- pass@100 on MultiPL MBPP (Python)self-reported0.770
- pass@1 on MultiPL HumanEval (Java)self-reported0.150
- pass@10 on MultiPL HumanEval (Java)self-reported0.260
- pass@100 on MultiPL HumanEval (Java)self-reported0.410
- pass@1 on MultiPL MBPP (Java)self-reported0.280
- pass@10 on MultiPL MBPP (Java)self-reported0.440
- pass@100 on MultiPL MBPP (Java)self-reported0.590
- single_line on HumanEval FIM (Python)self-reported0.440
- single_line on MultiPL HumanEval FIM (Java)self-reported0.620
- BLEU on CodeXGLUE code-to-text (Python)self-reported18.130