angkyw

[Weights & Biases @wandb]

Fine-tuning just got a whole lot easier.

Serverless SFT is now in public preview on W&B!

Managed infrastructure (powered by @CoreWeave) that auto-scales to your training workloads. No cluster setup. No idle GPU costs.

[angkyw @angkywilliam]

@Yuchenj_UW The EO is targeting H1B recipient from outside the States, which mostly is IT consultant.

It doesn’t affect and may even boost H1B chance of US college grad.

[angkyw @angkywilliam]

@EgeErdil2 Well, it does say without thinking

[Jack Morris @jxmnop]

happy birthday to the USA, the greatest country, and the origin of the following innovations:

- Transformers
- Pre-training (web-scale next-token prediction)
- RLHF
- RLVR
- RL
- GPUs
- TPUs
- PyTorch
- word2vec
- reasoning models
- GANs
- diffusion models
- VLMs
- self-driving cars

🇺🇸

[angkyw @angkywilliam]

@jxmnop Interesting take. What's the tradeoff between SGLang and vLLM?

[angkyw @angkywilliam]

@LBacaj Lol, should start a small bet as a composer.

[angkyw @angkywilliam]

@dvassallo Amazon likely build their internal cursor, dogfood internally and release it to public to compete with cursor.

[angkyw @angkywilliam]

The steam engine moment for intelligence is coming fast.

[angkyw @angkywilliam]

Starting to see why top AI labs believe ASI is inevitable.
Blending imitation learning (SFT) at different checkpoints with exploration learning (RL) can uncovers new solutions to existing problems and also tackle entirely new unsolved problems.

[angkyw @angkywilliam]

@goldstein_aa @jxmnop @AlexIrpan @sea_snell Training RL from scratch is hard. DeepSeek's approach builds on a strong base model.

Similar to how college helps build one knowledge base before applying it to solve real-world problems.

[Ross Taylor @rosstaylor90]

“Wait that can’t be right” in the wild

Thank you internet anons for your service to LLM reasoning. We found you through RL eventually 🫡

[angkyw @angkywilliam]

Agent for workflow is a deterministic task similar to math and coding with verifiable output.

[angkyw @angkywilliam]

"Reasoning" model trade compute for data efficiency

[angkyw @angkywilliam]

A naive approach in training "reasoning" model
1. FineTune instruct model with chain of thought
2. Use best of N to find chain of thought that give the right answer
3. Re fineTune the model with chain of thought that yields the right answer

[angkyw @angkywilliam]

@ironcarbs Seattle 😎

[angkyw @angkywilliam]

@volokuleshov @brandondamos The FIM part help me understand how cursor is being trained under the hood.

[angkyw @angkywilliam]

@volokuleshov @brandondamos I wish I could attend the lecture!
I am working on creating customize chat template serializer for Llama, Mistral and Qwen model.

[angkyw @angkywilliam]

@HaramiParindey I know this place, Roosevelt island!

[angkyw @angkywilliam]

Have not found class on dataset impact to model performance

[angkyw @angkywilliam]

DNN: Learning Algorithm + Dataset + System/Scale

Learning Algorithm: Architecture + Loss function + Optimizer (Gradient Descent)

Core classes:
- Architecture: https://t.co/pCZpDJNjS9
- Loss function: https://t.co/YbyHjadNWI
- System: https://t.co/VPK7hfjznv