File size: 2,443 Bytes
01bc9e4 ee47987 01bc9e4 ee47987 01bc9e4 ee47987 01bc9e4 fbd2075 01bc9e4 fbd2075 01bc9e4 fbd2075 01bc9e4 fbd2075 01bc9e4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 |
---
license: bigcode-openrail-m
library_name: peft
tags:
- trl
- sft
- generated_from_trainer
base_model: bigcode/starcoder2-3b
model-index:
- name: finetune_starcoder2_with_R_data
results: []
datasets:
- bigcode/the-stack
language:
- en
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# finetune_starcoder2_with_R_data
This model is a fine-tuned version of [bigcode/starcoder2-3b](https://huggingface.co/bigcode/starcoder2-3b), adapted and fine-tuned specifically for generating R programming code.
## Model description
This model is a specialized version of the bigcode/starcoder2-3b architecture fine-tuned on a subset of the Stack dataset, focusing solely on R programming language data. The fine-tuning process utilized the PEFT (Parameter Efficient Fine Tuning) method and included loading the model with 4-bit quantization using the LoRA library. It's tailored for generating R programming code, offering optimized performance for tasks within this domain.
## Intended uses & limitations
Tailored for R programming tasks, this model is optimized for generating code snippets, functions, or scripts in the R language. Its limitations may include its applicability solely within the domain of R programming and potential constraints related to the size and diversity of the training data.
## Training and evaluation data
The model was trained and evaluated on a subset of the bigcode/Stack dataset containing R programming language data.
## Training procedure
Fine-tuning was performed using the PEFT (Parameter Efficient Fine Tuning) method over 1000 epochs on the R dataset. Additionally, the model was loaded with 4-bit quantization using the LoRA library to optimize memory usage and inference speed, enhancing its efficiency for generating R code.
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 2
- eval_batch_size: 16
- seed: 0
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
- mixed_precision_training: Native AMP
=
### Framework versions
- PEFT 0.8.2
- Transformers 4.40.0.dev0
- Pytorch 2.1.2
- Datasets 2.16.1
- Tokenizers 0.15.2 |