Alade Wealth (@AladeWealth_) header
Alade Wealth (@AladeWealth_) profile picture

17, understanding the universe.

Joined June 2024
Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)Photo by Alade Wealth (@AladeWealth_)
Eventually you discover you can learn anything, don't let anyone hold you back , Keep pushing
1
1
28
4,087
Alade Wealth retweeted
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen
43,871
92,819
19,751
839,428
31,484,430
Προς το άπειρο
2
37
Alade Wealth retweeted
Reinforcement learning is one of the most promising paths to superintelligence. The key is making the training loop turn every experience into useful signal. With prioritised replay, process rewards, and good synthetic rollouts, even failed attempts become lasting progress instead of wasted effort.
Relentless reinforcement learning must feel like Edge of Tomorrow to superintelligence
1
1
75
I built a 70M params model. trained from scratch. Trilingua is a English, Spanish and French language model: own tokenizer, own data pipeline, own training loop. it beats GPT-2, Pythia and SmolLM2 on Spanish and French. English still needs work. trilingua.vercel.app/
8
3
2
20
697
everything is documented in the repo, the architecture, data and licenses, the full compute log, what went wrong, and 5 Kaggle notebooks to reproduce it on free GPUs, from tokenizer to eval. 46 tests, CPU smoke test included. github.com/aladewealth/Trili…
1
6
87
weights are on Hugging Face, the base model at the root, the fine-tuned version in sft/. safetensors, 32k tokenizer included. pick the language with one tag: <|en|>, <|es|> or <|fr|>. runs on CPU. huggingface.co/aladewealth/t…
6
74
Trilingua (70M) achieves 1.060 average BPB across English, Spanish and French, outperforming several larger models. 200 held-out documents per language. In domain evaluation. More testing ahead. 🚀
7
177
I just trained my first ever large language model
1
8
149
The Yoruba voice is trained. v0.1 weights are out 🇳🇬 Called 15–20M parameters. Landed at 19.05M, 77 MB ONNX, runs on CPU. Tone-aware, female and male voices. Early checkpoint, more training coming. huggingface.co/aladewealth/o…
Hello I just created a voice model on Yoruba, French, Spanish and English was trained with Piper, exported to ONNX (runs on CPU). With multi speakers both female and male voices.
2
6
164
Can be downloaded on any PCs
53
Hello I just created a voice model on Yoruba, French, Spanish and English was trained with Piper, exported to ONNX (runs on CPU). With multi speakers both female and male voices.
5
1
9
339
Our Yoruba voice will have roughly 15 to 20 million, in a file of about 60 MB. Kokoro, behind Spanish, French and English, has 82 million. Training on Kaggle is what fills in those numbers
5
42
Parameters are the model's adjustable numbers; weights are their trained values, saved in one file.
6
32
Datasets from OpenSLR 86, Crowdsourced high-quality Yoruba speech by Google, CC BY-SA 4.0. About 3,600 recorded lines from roughly 36 speakers. The dataset notes that its transcripts were not all checked by hand.
6
88
Superintelligence won’t come from feeding models more data forever. The real unlock is data efficiency: extracting more intelligence from every bit of information. If that scales, the bottleneck shifts from how much data we have to how much intelligence we can extract from it.
12
428
Alade Wealth retweeted
Evolution works by mutation and selection. Innovation works by trial and error. Science works by conjecture and criticism. Free markets work by entrepreneurship and risk. All truth-seeking systems work roughly the same way.
199
2,141
263
11,706
It's difficult to say what is impossible, for the dream of yesterday is the hope of today and the reality of tomorrow Nothing is impossible
1
1
10
429