Beta
Explore
Marketplace
Neural Labs
Chat
Wallet
Docs
AmharicGPT – AI Model by dagim | AlphaNeural AI
You can deploy this model and start earning money today!
dagim
/
AmharicGPT
like
0
transformers
pytorch
tensorboard
safetensors
gpt2
text-generation
generated_from_trainer
mit
autotrain_compatible
text-generation-inference
endpoints_compatible
us
Views
No views yet
Model card
Files and Versions
Community
API
Deploy
AmharicGPT
This model is a fine-tuned version of
gpt2
on an unknown dataset. It achieves the following results on the evaluation set:
Loss: 5.9221
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: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 64
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
Training results
Training Loss
Epoch
Step
Validation Loss
6.5754
0.71
5000
5.9221
Framework versions
Transformers 4.27.4
Pytorch 1.13.0
Datasets 2.1.0
Tokenizers 0.13.2
