shantipriya/hunyuan-ocr-odia

Fix sample table: remove duplicates, correct labels (Partial/Poor)

ckpt-4000 sample 6: CER=3.500

ckpt-4000 sample 5: CER=3.000

ckpt-4000 sample 4: CER=1.000

ckpt-4000 sample 3: CER=1.000

ckpt-4000 sample 2: CER=0.700

ckpt-4000 sample 1: CER=0.636

ckpt-4000 (80% of training): trainer_state.json

ckpt-4000 (80% of training): adapter_model.safetensors

ckpt-4000 (80% of training): README.md

Fix sample table: remove duplicates, correct labels (Partial/Poor)

Add word_samples.json

Add word sample 8

Add word sample 7

Add word sample 6

Add word sample 5

Add word sample 4

Add word sample 3

Add word sample 2

Add word sample 1

Fix sample table: remove duplicates, correct labels (Partial/Poor)

ckpt-3250: adapter_model.safetensors

ckpt-3250: adapter_model.safetensors

ckpt-3250: adapter_config.json

ckpt-3250: README.md

Add benchmark_samples.json

Add benchmark sample 4

Add benchmark sample 3

Add benchmark sample 2

Add benchmark sample 1

Add eval on Iftesha/odia-ocr-benchmark

Fix sample table: remove duplicates, correct labels (Partial/Poor)

Update README: diverse good/mixed/poor samples at ckpt-2500

Add diverse_samples.json

Add diverse sample 6

Add diverse sample 5

Add diverse sample 4

Add diverse sample 3

Add diverse sample 2

Add diverse sample 1

Update README: add inference sample images, loss curve, training progress

Add checkpoint-2000/training_args.bin

Add checkpoint-2000/trainer_state.json

Add checkpoint-2000/scheduler.pt

Add checkpoint-2000/optimizer.pt

Add checkpoint-2000/adapter_model.safetensors

Add checkpoint-2000/adapter_config.json

Update README: add 6 inference samples with images, GT, pred, CER

Add inference sample results JSON

Add inference sample 6