docs: document multi-GPU / multi-node tensor parallelism across all markdown docs - #96
Merged
Merged
Conversation
…arkdown docs The Chinese docs had no tensor-parallelism coverage at all, and both languages were missing the cluster tuning/diagnostics knobs that ship in TensorSharp.Distributed and the CUDA backend. - README.md: rewrite the multi-GPU quick start (fix typos, add the multi-node peer-list example, link the full reference); note the hierarchical AllReduce and host-staging fallback in Current Status. - README_zh-cn.md: add the TP highlight bullet, the multi-GPU/multi-node quick start, and the Current Status row. - FEATURES.md / FEATURES_zh-cn.md: describe the P2P self-test and the host-staging fallback; port the whole TP chapter to Chinese. - USAGE.md / USAGE_zh-cn.md: new "Cluster tuning & diagnostics" subsection; port the full TP + Redis shared-state chapter, the --tp* CLI flags, and the TENSORSHARP_TP_* env vars to Chinese. - docs/env_var_feature_matrix*.md: add TENSORSHARP_TP_DISABLE_P2P, _HOST_ALLREDUCE, _CONNECT_TIMEOUT_SECONDS and _RECV_TIMEOUT_SECONDS; add the missing TP and Redis tables to the Chinese matrix. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Umnq2SKaLVk2uytwTezvh2
…lism site-wide The wiki had zero coverage of tensor parallelism in either language, even though it is a headline feature in the README. - New distributed.html / distributed_zh-cn.html: when to use TP, how the column/row-parallel split and AllReduce work, local --tp N, multi-node --tp-node-id / --tp-peers, the per-architecture strategy table, requirements and constraints, the P2P self-test plus the TENSORSHARP_TP_* tuning switches, Redis-backed shared state, and a troubleshooting table. - Registered the page in the Deep dive nav group and wired it into the advanced -> distributed -> benchmarks footer chain. - cli / api-reference: --tp, --tp-node-id, --tp-peers and the TENSORSHARP_TP_* / TS_*_REDIS_* environment variables. - server: TENSORSHARP_TP_DEGREE (the server's only way to enable TP) and the multi-node + Redis variables. - backends: the direct cuda backend is the only one that shards a model. - features / overview / index / glossary: highlight card, status row, home-page entry points, and TP / AllReduce / node-rank definitions. - Both search indexes gained nine entries for the new page. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Umnq2SKaLVk2uytwTezvh2
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
Engine comparison — TensorSharp vs llama.cpp (PR smoke)No report artifact was produced — the benchmark failed before generating results (see the workflow logs). |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The Chinese docs had no tensor-parallelism coverage at all, and both
languages were missing the cluster tuning/diagnostics knobs that ship in
TensorSharp.Distributed and the CUDA backend.
multi-node peer-list example, link the full reference); note the
hierarchical AllReduce and host-staging fallback in Current Status.
quick start, and the Current Status row.
host-staging fallback; port the whole TP chapter to Chinese.
subsection; port the full TP + Redis shared-state chapter, the --tp*
CLI flags, and the TENSORSHARP_TP_* env vars to Chinese.
_HOST_ALLREDUCE, _CONNECT_TIMEOUT_SECONDS and _RECV_TIMEOUT_SECONDS;
add the missing TP and Redis tables to the Chinese matrix.
Co-Authored-By: Claude Opus 5 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_01Umnq2SKaLVk2uytwTezvh2