github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 08:30:05 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[PR #5657] [CLOSED] Parallelize Tokenization in api/embed #58569

New Issue
Closed
opened 2026-04-29 13:27:28 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-29 13:27:28 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/5657
Author: @royjhan
Created: 7/12/2024
Status: ❌ Closed

Base: main ← Head: royh-embed-parallel

📝 Commits (10+)

📊 Changes

8 files changed (+486 additions, -32 deletions)

View changed files

📝 api/client.go (+10 -1)
📝 api/types.go (+24 -0)
➕ integration/embed_test.go (+152 -0)
📝 llm/ext_server/server.cpp (+23 -16)
📝 llm/server.go (+8 -8)
📝 server/routes.go (+162 -3)
📝 server/routes_test.go (+103 -0)
📝 server/sched_test.go (+4 -4)

📄 Description

Example: batch embedding of 250 2500 token inputs with nomic-embed-text
Numbers: tokenizing + detokenizing one input takes 1.3ms. Comparatively, the batch embedding of 250 inputs takes 21.76s out of 22.64s total. Clear bottle neck on the embedding rather than tokenizing

TLDR parallelizing tokenization has no benefit for currently relevant workloads

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/ollama/ollama/pull/5657 **Author:** [@royjhan](https://github.com/royjhan) **Created:** 7/12/2024 **Status:** ❌ Closed **Base:** `main` ← **Head:** `royh-embed-parallel` --- ### 📝 Commits (10+) - [`c22d548`](https://github.com/ollama/ollama/commit/c22d54895a280b54c727279d85a5fc94defb5a29) Initial Batch Embedding - [`0f87628`](https://github.com/ollama/ollama/commit/0f87628b6dcbd29badc98868ea8a07e5c3ba80b1) Revert "Initial Batch Embedding" - [`ff191d7`](https://github.com/ollama/ollama/commit/ff191d7cba60caebe24201b1b7890a612efc956c) Initial Draft - [`22458c5`](https://github.com/ollama/ollama/commit/22458c573acd7c34d67b7b7b367bb98f8f9c2748) mock up notes - [`c406fa7`](https://github.com/ollama/ollama/commit/c406fa7a4c1f989a4be1a2ed995db6994ebb5816) api/embed draft - [`49e3411`](https://github.com/ollama/ollama/commit/49e341147db0fefb1a11f03b81af6f0e85eb7d45) add server function - [`b9c74df`](https://github.com/ollama/ollama/commit/b9c74df37b0ab3f63f2af37d8f443c1ff822dfb0) check normalization - [`5213c12`](https://github.com/ollama/ollama/commit/5213c1235436d99a59b144f87e43cbfc83d94d31) clean up - [`c111d8b`](https://github.com/ollama/ollama/commit/c111d8bb514f4615d69368278af4ef24551b4f65) normalization - [`80c1a3f`](https://github.com/ollama/ollama/commit/80c1a3f812cacff19e8857cc77dfd2fd4502b1e7) playing around with truncate stuff ### 📊 Changes **8 files changed** (+486 additions, -32 deletions) <details> <summary>View changed files</summary> 📝 `api/client.go` (+10 -1) 📝 `api/types.go` (+24 -0) ➕ `integration/embed_test.go` (+152 -0) 📝 `llm/ext_server/server.cpp` (+23 -16) 📝 `llm/server.go` (+8 -8) 📝 `server/routes.go` (+162 -3) 📝 `server/routes_test.go` (+103 -0) 📝 `server/sched_test.go` (+4 -4) </details> ### 📄 Description Example: batch embedding of 250 2500 token inputs with nomic-embed-text Numbers: tokenizing + detokenizing one input takes 1.3ms. Comparatively, the batch embedding of 250 inputs takes 21.76s out of 22.64s total. Clear bottle neck on the embedding rather than tokenizing TLDR parallelizing tokenization has no benefit for currently relevant workloads --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-04-29 13:27:28 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#58569
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?