[Docs] Add PyTorch Blog Link

[yubofredwang]

Summary

• refresh the README structure and intro wording for a clearer project overview
• update README links for the MiniMax example and add the PyTorch TorchSpec blog post to the Blogs section
• keep the change scoped to documentation in README.md

Test plan

• Review the rendered README link updates
• Confirm the branch only changes README.md

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cb2f0dd725

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Copilot]

Pull request overview

Updates README.md to provide a clearer high-level introduction to TorchSpec, refresh navigation/structure, and update external references (blogs + examples).

Changes:

• Reworks the README intro to include a short “what it is / what it includes” summary plus an Architecture Overview section.
• Adds a Blogs section (including a new PyTorch blog link) and introduces a Table of Contents for easier navigation.
• Updates Setup/Quick Start and the Examples table content/wording.

Comments suppressed due to low confidence (3)

README.md:107

• This Examples table links to examples/minimax-m25-5node-h200/, but that directory does not exist in the repository. Either add the missing example (if intended) or update/remove the link so the README doesn’t point to a broken path.

| [minimax-m25-5node-h200](examples/minimax-m25-5node-h200/) | Inference engine | MiniMax-M2.5 |

README.md:183

• Minor wording: “for every Ray actor on both training and inference” is grammatically awkward. Consider “for every Ray actor in both training and inference” or “for both training and inference actors”.

Set `TORCHSPEC_LOG_DIR` to an absolute path on a shared filesystem (NFS) to enable per-rank log files for every Ray actor on both training and inference:

README.md:88

• The vLLM example command uses ./examples/qwen3-8b-single-node/run.sh --config ..., but run.sh treats its first positional argument as the config file path and will pass any extra args through to torchspec.train_entry (which already receives a --config). Using --config here will likely result in a duplicate/invalid --config argument. Update the README to pass the config file as the first positional arg (e.g., ./examples/qwen3-8b-single-node/run.sh configs/vllm_qwen3_8b.yaml).

**vLLM**

```bash
./examples/qwen3-8b-single-node/run.sh --config configs/vllm_qwen3_8b.yaml

</details>



---

💡 <a href="/torchspec-project/TorchSpec/new/main?filename=.github/instructions/*.instructions.md" class="Link--inTextBlock" target="_blank" rel="noopener noreferrer">Add Copilot custom instructions</a> for smarter, more guided reviews. <a href="https://docs.github.com/en/copilot/customizing-copilot/adding-repository-custom-instructions-for-github-copilot" class="Link--inTextBlock" target="_blank" rel="noopener noreferrer">Learn how to get started</a>.