Rust Search
Find the best content on Rust, curated by the community; a
search engine for Rustaceans.
flodl 0.8.0 supports a second GPU vendor. AMD cards are detected, the code builds and links against ROCm, and the whole test suite runs against it. No AMD card has ever run a training step. Both sentences are true, and the second is the mor...
The last post ended on a promise: the floor is rebuilt, from here it is incremental again. This is seven days later. 42 commits, 141 files, 20,789 lines added. The floor held. 0.6.0 built a cluster. It did not give you a way to look into it...
flodl 0.6.0: multi-host heterogeneous DDP - mismatched GPUs across hosts beat the fastest card alone
flodl 0.6.0 is out, and it is by far the largest release this project has shipped. The distributed layer went from thread-per-GPU in one process to a full multi-host cluster: one Trainer::run() call now drives a single CPU, a single GPU, N...
v0.5.2 made HuggingFace checkpoints load in Rust. v0.5.3 closes the loop: every supported family-and-head combination now round-trips back to the HF ecosystem too, with bit-exact verification at the safetensors layer. Six families, four tas...
BERT in flodl is now from_pretrained("bert-base-uncased")?. So is RoBERTa. So is DistilBERT. So is any sentiment, NER, or SQuAD fine-tune sitting on top of those three families. The new sibling crate flodl-hf ships in 0.5.2 with PyTorch-ver...
Six days ago, v0.3.0 shipped with a confession: the CPU averaging backend was broken. Three policies, all failing to converge. We shipped anyway because the NCCL path worked and an honest label beats a delayed release. That bug is fixed. An...
I’ve been tracking a list. Every time I hit a PyTorch op that didn’t exist in flodl, it went on the list. Every time someone asked “does it have X?”, it went on the list. The list got long. This release empties the list. What was missing fl...
When I published the first benchmark a week ago, the headline was 19% faster than PyTorch. A real model, a real workload, same GPU. People noticed. But one number bugged me: the epoch standard deviation. flodl was 0.10s. PyTorch was 0.85s....
I’m not a systems programmer. I spent twenty years writing PHP and JavaScript. But I had a deep learning project that needed a custom framework, and Python’s overhead was killing the architectures I wanted to explore. So I learned Go and bu...
← Previous
Showing 1-9 of 9 results
Next →
""I'll never!"
"No, never is in the 2024 Edition."
"But never can't be this year, it's never!"
"Well we're trying to make it happen now!"
"But never is..."