this is why i use paddlepaddle ocr vl 1.6 for my doujin
November 9, 2026 at 9:52 AM
this is why i use paddlepaddle ocr vl 1.6 for my doujin
it really hurt my soul but now i just have the command to run for it so. i don't remember why, i think that was one of the guides that paddlepaddle had bc chinese devs hate the western man
November 6, 2026 at 8:07 PM
it really hurt my soul but now i just have the command to run for it so. i don't remember why, i think that was one of the guides that paddlepaddle had bc chinese devs hate the western man
Serving LLMs in production gets painful when you're stuck on one GPU vendor or rewriting clients for a new inference engine. FastDeploy from is an open source toolkit for LLMs and vision-language models on PaddlePaddle, with an OpenAI-compatible server, speculative decoding, prefix caching and...
November 7, 2026 at 12:44 AM
Serving LLMs in production gets painful when you're stuck on one GPU vendor or rewriting clients for a new inference engine. FastDeploy from is an open source toolkit for LLMs and vision-language models on PaddlePaddle, with an OpenAI-compatible server, speculative decoding, prefix caching and...
www.paddleocr.ai/main/en/vers... its sho good at least for me
PaddleOCR-VL-1.6 Introduction - PaddleOCR Documentation
Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and de...
www.paddleocr.ai
November 9, 2026 at 9:53 AM
www.paddleocr.ai/main/en/vers... its sho good at least for me
Baidu's camp leans on ERNIE and real manufacturing deployments, more than most vendor training sessions can claim. But a camp is marketing until attendees ship something measurable. Anyone who has sent engineers through one will know if the follow-through is real.
Baidu AI Technology Empowerment: Industrial-Grade AI Implementation Intensive Camp Opens
Baidu's PaddlePaddle has opened a five-week online intensive camp for partners in manufacturing, energy, education and healthcare, offering courses and expert support on deploying PaddlePaddle and ERNIE-based AI products.
aichina.news
November 1, 2026 at 5:03 PM
Baidu's camp leans on ERNIE and real manufacturing deployments, more than most vendor training sessions can claim. But a camp is marketing until attendees ship something measurable. Anyone who has sent engineers through one will know if the follow-through is real.
Baidu's PaddlePaddle inference benchmarks are unusually well documented — but they're still Baidu's numbers on Baidu's hardware, with no independent reproduction. Hard to see teams already deep in vLLM or TensorRT switching on this evidence.
PaddlePaddle high-performance inference upgrade accelerates LLM and multimodal model deployment
PaddlePaddle Framework 3.0 upgrades its inference engine with PIR and pass-based optimisation, new quantisation and dynamic-to-static features, and broader hardware support for deploying large language and multimodal models.
aichina.news
November 1, 2026 at 5:31 PM
Baidu's PaddlePaddle inference benchmarks are unusually well documented — but they're still Baidu's numbers on Baidu's hardware, with no independent reproduction. Hard to see teams already deep in vLLM or TensorRT switching on this evidence.
A single-hotel, vendor-measured uplift with no baseline is marketing, not evidence — though LLM agents muscling into pricing turf long held by RMS vendors like IDeaS is notable. Hard to see a revenue manager switching until chain-wide numbers with seasonality controls land.
AI “Revenue Genie” one-click pricing becomes a new hotel revenue tool as single-hotel monthly revenue rises more than 22%
A vendor case study from PaddlePaddle describes Shenyuhei’s Revenue Genie, an ERNIE-based hotel pricing assistant that analyses more than 700 factors to recommend room rates. The company claims one month of use raised a partner hotel’s occupancy from 70% to 91% and monthly revenue by 22.6%, though the results are self-reported promotional claims.
aichina.news
November 1, 2026 at 2:44 PM
A single-hotel, vendor-measured uplift with no baseline is marketing, not evidence — though LLM agents muscling into pricing turf long held by RMS vendors like IDeaS is notable. Hard to see a revenue manager switching until chain-wide numbers with seasonality controls land.
The ease-of-use pitch is Baidu grading its own homework, and a slim op set was never Paddle's bottleneck — ecosystem gravity was. Still, one code path for dynamic and static graphs is right; torch.compile proved the demand. Hard to see migrations before third-party numbers.
Unified Dynamic–Static Composition Mechanism: Drastically Fewer Operators, Developer-Friendly
PaddlePaddle 3.0's new operator composition mechanism splits operators into a strictly limited primitive set and composite operators built from them, cutting the cost of compiler integration, higher-order differentiation and new-hardware adaptation.
aichina.news
November 1, 2026 at 5:12 PM
The ease-of-use pitch is Baidu grading its own homework, and a slim op set was never Paddle's bottleneck — ecosystem gravity was. Still, one code path for dynamic and static graphs is right; torch.compile proved the demand. Hard to see migrations before third-party numbers.
Softest metric going: no retention, no outcomes, and every Chinese job board bundles the same writer. Still, LLM career tools clearly reach first-time jobseekers at scale. My read: distribution, not demand — recruiters still screening on these CVs will say otherwise.
New users double in H1: AI CV tool helps job-hunting novices rebuild their application competitiveness
A Baidu PaddlePaddle industry case study reports that Shanghai Caibei’s Zhitu CV tool doubled new users in H1 2024 after integrating the ERNIE large model for CV assessment, optimisation and interview-question generation.
aichina.news
November 1, 2026 at 4:51 PM
Softest metric going: no retention, no outcomes, and every Chinese job board bundles the same writer. Still, LLM career tools clearly reach first-time jobseekers at scale. My read: distribution, not demand — recruiters still screening on these CVs will say otherwise.
Open source and built for edge inference, PaddleX suits camera traps with no connectivity. But no accuracy figures or false-positive rates in the release — and rangers chase false alarms. Hard to see this displacing MegaDetector until Baidu publishes numbers on real trap data.
PaddleX image classification empowers wildlife protection: building an AI guardian for endangered species
Baidu PaddlePaddle and IFAW explain how PaddleX image classification and a no-code pipeline power the Endangered Species AI Guardian, which screens online listings for illegal wildlife products. The case study covers model selection, semi-supervised learning, deployment and reported results.
aichina.news
November 1, 2026 at 3:21 PM
Open source and built for edge inference, PaddleX suits camera traps with no connectivity. But no accuracy figures or false-positive rates in the release — and rangers chase false alarms. Hard to see this displacing MegaDetector until Baidu publishes numbers on real trap data.
Open weights plus vertical-domain fine-tuning is a real edge over closed document APIs. The accuracy claims, though, are Baidu grading its own homework — hard to see this displacing a tuned commercial extractor until third-party numbers on messy scans land.
PP-ChatOCRv3: Higher accuracy in general scenarios, stronger vertical-domain fine-tuning, and a new leap in intelligent text-image analysis
Baidu's PaddlePaddle team has released PP-ChatOCRv3, a text-image intelligent analysis pipeline in PaddleX that pairs improved layout parsing with ERNIE Bot for information extraction. The update adds image rectification and seal-text detection, and strengthens fine-tuning for text recognition and layout-region localisation in vertical domains.
aichina.news
November 1, 2026 at 2:36 PM
Open weights plus vertical-domain fine-tuning is a real edge over closed document APIs. The accuracy claims, though, are Baidu grading its own homework — hard to see this displacing a tuned commercial extractor until third-party numbers on messy scans land.
🚨 EUVD-2026-98276
📊 5.3/10
🏢 PaddlePaddle
📝 A flaw has been found in PaddlePaddle PaddleX up to 3.7.2. Affected by this vulnerability is the function NormalizeImage.__init__ of the file paddlex...
🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-98276
#cybersecurity #infosec #cve #euvd
📊 5.3/10
🏢 PaddlePaddle
📝 A flaw has been found in PaddlePaddle PaddleX up to 3.7.2. Affected by this vulnerability is the function NormalizeImage.__init__ of the file paddlex...
🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-98276
#cybersecurity #infosec #cve #euvd
October 13, 2026 at 5:00 AM
🚨 EUVD-2026-98276
📊 5.3/10
🏢 PaddlePaddle
📝 A flaw has been found in PaddlePaddle PaddleX up to 3.7.2. Affected by this vulnerability is the function NormalizeImage.__init__ of the file paddlex...
🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-98276
#cybersecurity #infosec #cve #euvd
📊 5.3/10
🏢 PaddlePaddle
📝 A flaw has been found in PaddlePaddle PaddleX up to 3.7.2. Affected by this vulnerability is the function NormalizeImage.__init__ of the file paddlex...
🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-98276
#cybersecurity #infosec #cve #euvd
PaddleOCR 3.7.0 업데이트에서 PP-OCRv6가 공개되었네요. 34.5M 파라미터의 미디엄 티어 모델로 대형 VLM을 능가하는 정확도를 보여주며, CPU와 애플 M4에서도 빠르게 추론할 수 있어 온프레미스 문서 파싱에 유용해 보입니다.
#AI #개발
https://github.com/PaddlePaddle/PaddleOCR/blob/main/readme/README_ko.md
#AI #개발
https://github.com/PaddlePaddle/PaddleOCR/blob/main/readme/README_ko.md
October 6, 2026 at 1:13 PM
PaddleOCR 3.7.0 업데이트에서 PP-OCRv6가 공개되었네요. 34.5M 파라미터의 미디엄 티어 모델로 대형 VLM을 능가하는 정확도를 보여주며, CPU와 애플 M4에서도 빠르게 추론할 수 있어 온프레미스 문서 파싱에 유용해 보입니다.
#AI #개발
https://github.com/PaddlePaddle/PaddleOCR/blob/main/readme/README_ko.md
#AI #개발
https://github.com/PaddlePaddle/PaddleOCR/blob/main/readme/README_ko.md
Baidu’s PaddlePaddle Team Releases PaddleOCR-VL (0.9B): a NaViT-style + ERNIE-4.5-0.3B VLM Targeting End-to-End Multilingual Document Parsing
How do you convert complex, multilingual documents—dense layouts, small scripts, formulas, charts, and handwriting—into faithful structured Markdown/JSON…
How do you convert complex, multilingual documents—dense layouts, small scripts, formulas, charts, and handwriting—into faithful structured Markdown/JSON…
Baidu’s PaddlePaddle Team Releases PaddleOCR-VL (0.9B): a NaViT-style + ERNIE-4.5-0.3B VLM Targeting End-to-End Multilingual Document Parsing
How do you convert complex, multilingual documents—dense layouts, small scripts, formulas, charts, and handwriting—into faithful structured Markdown/JSON with state-of-the-art accuracy while keeping inference latency and memory low enough for real deployments?Baidu’s PaddlePaddle group has released PaddleOCR-VL, a 0.9B-parameter vision-language model designed for end-to-end document parsing across text, tables, formulas, charts, and handwriting. The core model combines a NaViT-style…
nexttech-news.com
October 17, 2025 at 9:07 AM
Baidu’s PaddlePaddle Team Releases PaddleOCR-VL (0.9B): a NaViT-style + ERNIE-4.5-0.3B VLM Targeting End-to-End Multilingual Document Parsing
How do you convert complex, multilingual documents—dense layouts, small scripts, formulas, charts, and handwriting—into faithful structured Markdown/JSON…
How do you convert complex, multilingual documents—dense layouts, small scripts, formulas, charts, and handwriting—into faithful structured Markdown/JSON…
Join us for our annual Kew Bridge Paddlesports Open Day 🏄🛶🚣♂️
There will be activities on the day including a River clean up Paddle to the Pub, Kayak sessions and Low tide rentals for club members/experienced paddlers.
#paddlesports #openday #kewbridge #standuppaddleuk #activelifestyle
There will be activities on the day including a River clean up Paddle to the Pub, Kayak sessions and Low tide rentals for club members/experienced paddlers.
#paddlesports #openday #kewbridge #standuppaddleuk #activelifestyle
May 21, 2025 at 8:29 PM
Join us for our annual Kew Bridge Paddlesports Open Day 🏄🛶🚣♂️
There will be activities on the day including a River clean up Paddle to the Pub, Kayak sessions and Low tide rentals for club members/experienced paddlers.
#paddlesports #openday #kewbridge #standuppaddleuk #activelifestyle
There will be activities on the day including a River clean up Paddle to the Pub, Kayak sessions and Low tide rentals for club members/experienced paddlers.
#paddlesports #openday #kewbridge #standuppaddleuk #activelifestyle
Intel oneDNN 3.7 Begins Tuning For Xe3 Graphics, Adds More Granite Rapids Optimizations Intel sof...
https://www.phoronix.com/news/Intel-oneDNN-3.7
Event Attributes
https://www.phoronix.com/news/Intel-oneDNN-3.7
Event Attributes
Intel oneDNN 3.7 Begins Tuning For Xe3 Graphics, Adds More Granite Rapids Optimizations
Intel software engineers on Tuesday released oneDNN 3.7 as the newest version of this oneAPI Deep Neural Network Library that is in turn used by software like ONNX, MATLAB, PaddlePaddle, Apache MXNet, and others as part of the building blocks for deep learning software...
www.phoronix.com
February 19, 2025 at 2:35 PM
Intel oneDNN 3.7 Begins Tuning For Xe3 Graphics, Adds More Granite Rapids Optimizations Intel sof...
https://www.phoronix.com/news/Intel-oneDNN-3.7
Event Attributes
https://www.phoronix.com/news/Intel-oneDNN-3.7
Event Attributes
✨Big update from Baidu
The PaddleOCR now supports Transformers as an inference backend 🔥 Really cool to see it becoming easier to use within the @hf.co ecosystem!
huggingface.co/blog/PaddleP...
The PaddleOCR now supports Transformers as an inference backend 🔥 Really cool to see it becoming easier to use within the @hf.co ecosystem!
huggingface.co/blog/PaddleP...
PaddleOCR 3.5: Running OCR and Document Parsing Tasks with a Transformers Backend
A Blog post by PaddlePaddle on Hugging Face
huggingface.co
May 19, 2026 at 8:00 AM
✨Big update from Baidu
The PaddleOCR now supports Transformers as an inference backend 🔥 Really cool to see it becoming easier to use within the @hf.co ecosystem!
huggingface.co/blog/PaddleP...
The PaddleOCR now supports Transformers as an inference backend 🔥 Really cool to see it becoming easier to use within the @hf.co ecosystem!
huggingface.co/blog/PaddleP...
Baidu just released a new VLM : PaddleOCR-VL-1.5🔥
huggingface.co/PaddlePaddle...
✨ 0.9B - Apache 2.0
✨ 94.5% OmniDocBench v1.5
✨ Multilingual OCR: strong on rare characters & ancient texts
huggingface.co/PaddlePaddle...
✨ 0.9B - Apache 2.0
✨ 94.5% OmniDocBench v1.5
✨ Multilingual OCR: strong on rare characters & ancient texts
January 29, 2026 at 10:32 AM
Baidu just released a new VLM : PaddleOCR-VL-1.5🔥
huggingface.co/PaddlePaddle...
✨ 0.9B - Apache 2.0
✨ 94.5% OmniDocBench v1.5
✨ Multilingual OCR: strong on rare characters & ancient texts
huggingface.co/PaddlePaddle...
✨ 0.9B - Apache 2.0
✨ 94.5% OmniDocBench v1.5
✨ Multilingual OCR: strong on rare characters & ancient texts
Baidu - which committed to open source its LLM by June, has just released a demo of ERNIE 4.5 on @hf.co today👀
huggingface.co/spaces/Paddl...
huggingface.co/spaces/Paddl...
ERNIE 4.5 - a Hugging Face Space by PaddlePaddle
ERNIE 4.5 (BAIDU's LLM), https://yiyan.baidu.com/
huggingface.co
April 22, 2025 at 12:50 PM
Baidu - which committed to open source its LLM by June, has just released a demo of ERNIE 4.5 on @hf.co today👀
huggingface.co/spaces/Paddl...
huggingface.co/spaces/Paddl...
Did you catch our latest changelog update? Updates include:
⚡️ Faster Koyeb Sandboxes boot times
🧑💻 New tutorial: Generate code using @OpenAI Codex in a Koyeb Sandbox
🚀 New one-click: PaddlePaddle FastDeploy
& more!
Check out the full changelog update below 👇🧵
⚡️ Faster Koyeb Sandboxes boot times
🧑💻 New tutorial: Generate code using @OpenAI Codex in a Koyeb Sandbox
🚀 New one-click: PaddlePaddle FastDeploy
& more!
Check out the full changelog update below 👇🧵
December 9, 2025 at 9:23 AM
Did you catch our latest changelog update? Updates include:
⚡️ Faster Koyeb Sandboxes boot times
🧑💻 New tutorial: Generate code using @OpenAI Codex in a Koyeb Sandbox
🚀 New one-click: PaddlePaddle FastDeploy
& more!
Check out the full changelog update below 👇🧵
⚡️ Faster Koyeb Sandboxes boot times
🧑💻 New tutorial: Generate code using @OpenAI Codex in a Koyeb Sandbox
🚀 New one-click: PaddlePaddle FastDeploy
& more!
Check out the full changelog update below 👇🧵
[2/3] has completed Level III compatibility testing with Baidu's deep learning platform Baidu PaddlePaddle. Both parties have excellent performance in terms of compatibility, and the overall system operation is stable and reliable. At present, the Paddle-MUSA operator adaptation rate
July 3, 2026 at 4:45 AM
[2/3] has completed Level III compatibility testing with Baidu's deep learning platform Baidu PaddlePaddle. Both parties have excellent performance in terms of compatibility, and the overall system operation is stable and reliable. At present, the Paddle-MUSA operator adaptation rate
[1/3] 2026-07-03 12:33:45 - [Moore Thread's full-featured GPU obtains Baidu PaddlePaddle Level III compatibility certification] Reporters learned from domestic GPU manufacturer Moore Thread that the company's flagship AI training and promotion all-in-one smart computing card MTT S5000
July 3, 2026 at 4:45 AM
[1/3] 2026-07-03 12:33:45 - [Moore Thread's full-featured GPU obtains Baidu PaddlePaddle Level III compatibility certification] Reporters learned from domestic GPU manufacturer Moore Thread that the company's flagship AI training and promotion all-in-one smart computing card MTT S5000
[2/3] has completed Level III compatibility testing with Baidu's deep learning platform Baidu PaddlePaddle. Both parties have excellent performance in terms of compatibility, and the overall system operation is stable and reliable. At present, the Paddle-MUSA operator adaptation rate
July 3, 2026 at 4:45 AM
[2/3] has completed Level III compatibility testing with Baidu's deep learning platform Baidu PaddlePaddle. Both parties have excellent performance in terms of compatibility, and the overall system operation is stable and reliable. At present, the Paddle-MUSA operator adaptation rate
[1/3] 2026-07-03 12:33:45 - [Moore Thread's full-featured GPU obtains Baidu PaddlePaddle Level III compatibility certification] Reporters learned from domestic GPU manufacturer Moore Thread that the company's flagship AI training and promotion all-in-one smart computing card MTT S5000
July 3, 2026 at 4:45 AM
[1/3] 2026-07-03 12:33:45 - [Moore Thread's full-featured GPU obtains Baidu PaddlePaddle Level III compatibility certification] Reporters learned from domestic GPU manufacturer Moore Thread that the company's flagship AI training and promotion all-in-one smart computing card MTT S5000