NLPer

I wrote the following package in hope of helping whoever is seeking for the solutions to TPUs training with TF2.0 from pyT, as I did a week ago and could not find proper solutions. Hence, I built something from scratch shared here #nlp #tpu #transformers: github.com/wangcongcong12…

@AirCtl_rCtlVal @disksing hf上有啊

@ThoughtsNlp @disksing 所以deepseek v3的无审核版你见过吗？

想了想我这儿已经没有deepseek-v4的位置了，论写代码它肯定写不过codex，跑workflow不如kimi便宜，聊天又有审查不如grok。我连测一下的想法都没有。

@Jaciezyt 第四个问题你说一张25元的钱 就没有这个币种 你没必要故意这样愚弄模型

注：均为Flash模式，Expert模式没有问题

Deepseek V4 弱智吧测试 洗车 ✅ ATM ❌ 木棍过门 ❌ 假币找零 ❌

@Jaciezyt 虽然木棍它回答的不是最简单直接的方法 但是他说的也没错吧

@yidabuilds @demonyins 优势是更快吗

@demonyins 和glm，kimi同水平的价格 = 很不经用

就这deepseek api 价格，还要什么coding plan? 来你们帮我算算，100美金换算成deepseek api 能用到什么时候？

@w_yy9179791 @LQniupitang 他的意思是问答式ai不是agentic ai。用来问问题回答问题免费

@LQniupitang 干不了活，免费也没用啊。

chatGPT再牛逼，它收费啊。 豆包，DeepSeek，真免费啊。

@Soranlan @wadezone I hate the company Anthropic

@wadezone 我相信，Claude opus4.7会为了雪耻追回第一

我这边有一个比较客观的大模型智能排名： 1 GPT-5.5 (xhigh) 2 GPT-5.5 (high) 3 Claude Opus 4.7 (max) 4 Gemini 3.1 Pro Preview 5 GPT-5.4 (xhigh)

@arkuy99 so use it for what?

我他妈终于学会了 100 刀 gpt 的正确用法。。🤣 用来写代码真是暴遣天物

@localhost_4173 测了gpt5.5人家回答的也很准，不要尬黑啊

gpt-5.5 / deepseek-v4

@Xuegaogx 别说这么脏的话

操，真特么爽，再也不用提心吊胆订阅被封了 直接中转 + Desktop 妈妈再也不用担心我被封了🤣

@perfect_z39479 @Alexjkman 瞎说完全不是一个档次的东西了

@Alexjkman 其实image2甚至不如banana,外行人的狂欢罢了

我这几天其实很好奇一个事情 这几天gpt-image2 出来以后 现在大学里面设计专业 都在教什么

想问一下市面上有没有这样的工具？

@mao_ge_ge @pangyusio 我咋看不出来

@pangyusio 这张图我的眼睛唯一能稍微看出一点ai痕迹的地方就是爸爸的脸

提示词： 生成一张 2003 年，用数码相机拍摄的，5 岁的我和爸爸妈妈在中国的家，小区院子里的照片。 AI 生图已经基本通过图灵测试了。

@SanejBandgar @sundarpichai @GeminiApp @antigravity This is equivalent to ChatGPT Desktop on Mac it is not like Codex. Antigravity is thought I don’t it.

@sundarpichai @GeminiApp @antigravity looks cool! can we have projects / folders option?

Introducing Gemini on Mac. It’s the first time we’re bringing the @Geminiapp to desktop. The team built this initial release with @Antigravity, and it went from an idea to a native Swift app prototype in a few days. More features on the way!

@pinchetobs @NVIDIAAIDev @karpathy @DellTech @grok how much is it

@NVIDIAAIDev @karpathy @DellTech this is it: 20A. 20 tiny amps.

🙌 Andrej Karpathy’s lab has received the first DGX Station GB300 -- a Dell Pro Max with GB300. 💚 We can't wait to see what you’ll create @karpathy! 🔗 #dgx-station" target="_blank" rel="nofollow noopener">blogs.nvidia.com/blog/gtc-2026-… @DellTech

@mrm8488 @NarrativaAI @huggingface what is the padding id in batches for training?

So at @NarrativaAI we have fine-tuned the first #ByT5 model (small) on the @huggingface model hub. The dataset is: tweets hate speech detection and we got an accuracy of: 97.8 %! Try it out: huggingface.co/mrm8488/byt5-s… #NLP #NLP #Text2Text

@zipfern @me15Third @Geointheworld What is the joke? I did not get it. What does "OP" mean?

Ecco – See what your NLP language model is “thinking” eccox.io Ecstatic to release my first open-source project! Interactive visualizations in jupyter for @huggingface GPT2-based language models. Github: github.com/jalammar/ecco HN: news.ycombinator.com/item?id=256838…

We’ve developed two neural networks which have learned by associating text and images. CLIP maps images into categories described in text, and DALL-E creates new images, like this, from text. A step toward systems with deeper understanding of the world. openai.com/multimodal/

Had a hack on this lib. Loads stuff learnt about #Transformers #NLP.