# google-pytorch/torch_tpu

> TorchTPU is a PyTorch backend for Google TPUs that lets you run PyTorch models and operations on TPU devices, with custom ATen kernels, an XLA compilation cache, and distributed training support.

- Magnitude: 5.4 out of 10 — Early signal
- Stars: 169 total · +60 stars measured 2026-10-20, 17:13–18:55 UTC, ≈ 73 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: C++ · Created: 2026-03-09 · Last push: 2026-10-20
- GitHub: https://github.com/google-pytorch/torch_tpu · Homepage: https://torchtpu.readthedocs.io · Page: https://gitnova.dev/en/r/google-pytorch/torch_tpu

## Useful for

- Run PyTorch models such as Llama, Qwen, ResNet, or UNet on TPU hardware.
- Set up distributed training with data or tensor parallelism on TPU VMs.
- Use torch.compile() graph mode execution with the TPU backend for optimized inference.

## Why it’s here

- Star-counter measurements on 2026-10-20 (UTC), 17:13–18:55: 109 → 169 stars (+60). This is the change over that interval.
- Estimated end-of-day forecast: about +73 stars, using observed gains and the previous day.
- Before this, the repository barely got any stars — about 0 per day.
- Hacker News: “Google open-sources TorchTPU, a native PyTorch back end for TPUs” — 104 points, 2 h ago.
- About 6 forks a day — people are taking the code.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 8
- Issues and pull requests: 5,652
- Watchers: 1
- Average over the last week: 13 per day
- Usual pace: 1 per day
- Stars in the last hour (measured): 35

## Stars per day, last 30 days (oldest → newest, today is partial)

2026-09-21 … 2026-10-20: 1, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 1, 0, 1, 0, 0, 2, 0, 0, 1, 0, 0, 0, 17, 60

## Hacker News

- Google open-sources TorchTPU, a native PyTorch back end for TPUs — 104 points, 28 comments: https://news.ycombinator.com/item?id=50169053

## Spotted in now

- Spotted on Hacker News

## More in this category

1. **deepseek-ai/DeepInfer** — 10.0 · Breakout · Language models · Python · +9,276 stars measured 2026-10-20, 09:29–18:53 UTC, ≈ 11,158 by evening
   Offline inference framework for running DeepSeek models at large scale: optimized kernels, MoE/Attention parallelism, deterministic sampling and structured outputs. Aimed at teams deploying DeepSeek-V4.1-Flash on NVIDIA SM100 or Huawei…
   Full card: https://gitnova.dev/en/r/deepseek-ai/DeepInfer.md
2. **mistralai/timestral** — 5.4 · Early signal · Language models · Python · +90 stars measured 2026-10-20, 00:22–18:54 UTC, ≈ 116 by evening
   A family of pretrained foundation models for multivariate probabilistic time series forecasting with covariate support. Enables zero-shot forecasting on CPU or GPU without training.
   Full card: https://gitnova.dev/en/r/mistralai/timestral.md
3. **headroomlabs-ai/headroom** — 5.4 · Breakout · Language models · Python · +416 stars measured 2026-10-20, 00:23–18:54 UTC, ≈ 520 by evening
   A local library, proxy and MCP server that compress tool outputs, logs, files and RAG chunks before they reach the LLM, cutting token usage without changing answers.
   Full card: https://gitnova.dev/en/r/headroomlabs-ai/headroom.md
4. **MemTensor/MemT-1.0** — 5.2 · Early signal · Language models · Python · +71 stars measured 2026-10-20, 06:37–18:54 UTC, ≈ 90 by evening
   A family of memory foundation models with native long-term memory: dynamic local memory and trainable hyper memory retain history without accessing the original context, cutting storage roughly 66x versus full-context KV cache at 128K.
   Full card: https://gitnova.dev/en/r/MemTensor/MemT-1.0.md
5. **solidsf-inc/solidcv** — 5.1 · Early signal · Language models · Rust · +20 stars measured 2026-10-20, 00:23–18:54 UTC, ≈ 30 by evening
   A pure-Rust OpenCV replacement with zero FFI: classical 2D/3D vision, stereo reconstruction, and mechanistic interpretability for LLMs/VLMs with hardware kernels for Apple Silicon and NVIDIA.
   Full card: https://gitnova.dev/en/r/solidsf-inc/solidcv.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-10-20 19:07 UTC, updated every 30 minutes.