Hugging Face's logo

cyttic
/
trocr-bigram5

Image-Text-to-Text
Transformers
Safetensors
vision-encoder-decoder
Generated from Trainer
Model card Files
xet
 Community
How to use from
SGLang
Install from pip and serve model 
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "cyttic/trocr-bigram5" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "cyttic/trocr-bigram5",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'
Use Docker images 
docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "cyttic/trocr-bigram5" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "cyttic/trocr-bigram5",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'
Quick Links

trocr-bigram5

This model is a fine-tuned version of cyttic/exp2-frozen-benyehuda-cont on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4484
  • Cer: 0.0245
  • Wer: 0.0695

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: 2e-05
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 4650
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss Cer Wer
5.1007 0.1290 2000 2.3098 0.2357 0.4602
3.7860 0.2581 4000 1.7228 0.1555 0.3295
2.9171 0.3871 6000 1.3482 0.1044 0.2393
2.3965 0.5161 8000 1.1463 0.0872 0.2014
2.2010 0.6452 10000 0.9637 0.0678 0.1679
1.9024 0.7742 12000 0.8735 0.0596 0.1500
1.6664 0.9032 14000 0.7796 0.0526 0.1344
1.3597 1.0323 16000 0.7268 0.0460 0.1207
1.1604 1.1613 18000 0.6860 0.0420 0.1127
1.2101 1.2903 20000 0.6513 0.0396 0.1051
1.0679 1.4194 22000 0.6164 0.0372 0.0994
1.0820 1.5484 24000 0.5783 0.0321 0.0892
1.0818 1.6774 26000 0.5519 0.0301 0.0856
1.0749 1.8065 28000 0.5298 0.0300 0.0841
0.9703 1.9355 30000 0.5098 0.0280 0.0792
0.7494 2.0645 32000 0.4943 0.0283 0.0785
0.7928 2.1935 34000 0.4828 0.0266 0.0742
0.6909 2.3226 36000 0.4806 0.0256 0.0727
0.7412 2.4516 38000 0.4683 0.0263 0.0726
0.7338 2.5806 40000 0.4625 0.0255 0.0711
0.7880 2.7097 42000 0.4548 0.0252 0.0700
0.7299 2.8387 44000 0.4489 0.0246 0.0693
0.7580 2.9677 46000 0.4480 0.0241 0.0689
0.7460 3.0 46500 0.4484 0.0245 0.0695

Framework versions

  • Transformers 5.15.0
  • Pytorch 2.11.0+cu128
  • Datasets 5.0.1
  • Tokenizers 0.22.2
Downloads last month
37
Safetensors
Model size
0.3B params
Tensor type
F32
·
Inference Providers NEW
Image-Text-to-Text
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for cyttic/trocr-bigram5

Base model

cyttic/exp2-frozen-benyehuda-cont
Finetuned
(48)
this model