Downloads · 30 days
38
7% of all-time downloads
codeparrot-ds-sample
Pavithra/codeparrot-ds-sample
codeparrot-ds-sample is a text generation model from Pavithra. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as mit.
should probably proofread and complete it, then remove this comment. --
All-time downloads
517
Public
Repo size
2 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin510 MB · 99%
At a glance
- Task
- Text Generation
- Library
- transformers
- License
- mit
- Model type
- gpt2
- Access
- Public
- Created
- Mar 23, 2022
- Updated
- Mar 24, 2022
- SHA
- bbc16b22
Try a prompt
Model card
From the Hugging Face model README
<!-- 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. -->
codeparrot-ds-sample
This model is a fine-tuned version of gpt2 on an unknown dataset. It achieves the following results on the evaluation set:
- eval_loss: 1.5219
- eval_runtime: 603.3856
- eval_samples_per_second: 154.402
- eval_steps_per_second: 4.826
- epoch: 0.15
- step: 10000
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: 0.0005
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 1000
- num_epochs: 1
- mixed_precision_training: Native AMP
Framework versions
- Transformers 4.17.0
- Pytorch 1.10.0+cu111
- Datasets 2.0.0
- Tokenizers 0.11.6
- Task
- Text Generation
- Library
- transformers
- Type
- gpt2
- License
- mit
- Created
- Mar 23, 2022
- Updated
- Mar 24, 2022