Spaces:

Gabriel
/

diffu-htr-speed

Running on Zero

42.8 MB

Ctrl+K

1 contributor

Jev in one accordion titled with its verdict; each reading shows its errors and Jev's score in percent; picks marked with the reader's name and ✓ or ✗

55c179d verified about 19 hours ago

examples
Examples drawn at random from test pages (TrOCR trained on our dev volumes), timed again; Jev's section: the region's verdict, CER bars and one row per line about 20 hours ago
.gitattributes

2.78 kB
Examples drawn at random from test pages (TrOCR trained on our dev volumes), timed again; Jev's section: the region's verdict, CER bars and one row per line about 20 hours ago
DejaVuSans.ttf

760 kB
xet

Swedish HTR speed and Jev: TrOCR vs MiniCPM-OCR alone and with its DSpark draft (PyTorch on CUDA graphs), Jev per line about 23 hours ago
HOW_IT_WORKS.md

8.83 kB
Jev in one fold titled with its verdict (picked the reading with fewer errors on K of M lines); each reading's errors and Jev's certainty in percent about 19 hours ago
README.md

1.94 kB
One stream each: TrOCR reads its lines one by one (its one-batch time beside it), the replay starts all three together, our server's vLLM timings under the race, one region box, Jev's rows folded with marked errors about 20 hours ago
app.py

32.3 kB
Jev in one fold titled with its verdict (picked the reading with fewer errors on K of M lines); each reading's errors and Jev's certainty in percent about 19 hours ago
check_decode.py

4.23 kB
The race as a custom component: each reader's reading replayed at its measured pace on the manuscript (a highlighter along the lines' outlines) and as text, a block per reader pass, lines linked on hover; results streamed as each reader finishes; Jev scores both readers' reading of each line; a 4,096-token cache; no fla about 22 hours ago
dspark_hf.py

29.3 kB
The race as a custom component: each reader's reading replayed at its measured pace on the manuscript (a highlighter along the lines' outlines) and as text, a block per reader pass, lines linked on hover; results streamed as each reader finishes; Jev scores both readers' reading of each line; a 4,096-token cache; no fla about 22 hours ago
jev_prompt.py

2.11 kB
Swedish HTR speed and Jev: TrOCR vs MiniCPM-OCR alone and with its DSpark draft (PyTorch on CUDA graphs), Jev per line about 23 hours ago
jev_render.py

2.38 kB
Swedish HTR speed and Jev: TrOCR vs MiniCPM-OCR alone and with its DSpark draft (PyTorch on CUDA graphs), Jev per line about 23 hours ago
measure_reference.py

6.96 kB
One stream each: TrOCR reads its lines one by one (its one-batch time beside it), the replay starts all three together, our server's vLLM timings under the race, one region box, Jev's rows folded with marked errors about 20 hours ago
race_view.css

9.82 kB
Jev in one accordion titled with its verdict; each reading shows its errors and Jev's score in percent; picks marked with the reader's name and ✓ or ✗ about 19 hours ago
race_view.js

23 kB
Jev in one fold titled with its verdict (picked the reading with fewer errors on K of M lines); each reading's errors and Jev's certainty in percent about 19 hours ago
race_view.py

1.54 kB
One stream each: TrOCR reads its lines one by one (its one-batch time beside it), the replay starts all three together, our server's vLLM timings under the race, one region box, Jev's rows folded with marked errors about 20 hours ago
reference.json

1.04 kB
Examples drawn at random from test pages (TrOCR trained on our dev volumes), timed again; Jev's section: the region's verdict, CER bars and one row per line about 20 hours ago
reference.py

757 Bytes
One stream each: TrOCR reads its lines one by one (its one-batch time beside it), the replay starts all three together, our server's vLLM timings under the race, one region box, Jev's rows folded with marked errors about 20 hours ago
requirements.txt

101 Bytes
The race as a custom component: each reader's reading replayed at its measured pace on the manuscript (a highlighter along the lines' outlines) and as text, a block per reader pass, lines linked on hover; results streamed as each reader finishes; Jev scores both readers' reading of each line; a 4,096-token cache; no fla about 22 hours ago