Rename slime references to vime by aoshen02 · Pull Request #121 · vllm-project/vime

aoshen02 · 2026-06-02T07:16:10Z

Summary

Rename non-acknowledgement slime/SLIME/Slime references to vime/VIME/Vime across docs, scripts, CI, issue templates, and code identifiers
Update repository and docs links from THUDM/slime to vllm-project/vime
Keep the README acknowledgement references to the slime community as requested

Verification

git diff --check
git diff --name-only -z -- '*.py' | xargs -0 ../../.venv/bin/python -m py_compile
git diff --name-only -z -- '*.sh' | xargs -0 bash -n
Confirmed remaining slime matches are only README acknowledgement lines

gemini-code-assist

Code Review

This pull request performs a comprehensive renaming of the framework from 'slime' to 'vime' across documentation, scripts, and codebase comments. The reviewer's feedback correctly points out several instances where renaming 'slime' to 'vime' is problematic, such as creating circular statements (e.g., 'vime is derived from vime'), breaking external links to upstream pull requests, making upstream issue references untraceable, and incorrectly modifying 'upstream slime' references in the backend engine code. All of these comments are valid and should be addressed by reverting those specific renames.

gemini-code-assist · 2026-06-02T07:18:09Z

 ## About vime

-**vime** is a reinforcement learning (RL) post-training framework built on [**vLLM**](https://github.com/vllm-project/vllm) and [**Megatron-LM**](https://github.com/NVIDIA/Megatron-LM) as its core backends. It is derived from [slime](https://github.com/THUDM/slime).
+**vime** is a reinforcement learning (RL) post-training framework built on [**vLLM**](https://github.com/vllm-project/vllm) and [**Megatron-LM**](https://github.com/NVIDIA/Megatron-LM) as its core backends. It is derived from [vime](https://github.com/vllm-project/vime).


The project is derived from the upstream slime framework. Renaming this reference to vime creates a circular statement ('vime is derived from vime'). Please keep the reference to slime here.

Suggested change

**vime** is a reinforcement learning (RL) post-training framework built on [**vLLM**](https://github.com/vllm-project/vllm) and [**Megatron-LM**](https://github.com/NVIDIA/Megatron-LM) as its core backends. It is derived from [vime](https://github.com/vllm-project/vime).

**vime** is a reinforcement learning (RL) post-training framework built on [**vLLM**](https://github.com/vllm-project/vllm) and [**Megatron-LM**](https://github.com/NVIDIA/Megatron-LM) as its core backends. It is derived from [slime](https://github.com/THUDM/slime).

gemini-code-assist · 2026-06-02T07:18:10Z

 ## 关于vime

-**vime**是以[**vLLM**](https://github.com/vllm-project/vllm)与[**Megatron-LM**](https://github.com/NVIDIA/Megatron-LM)为核心后端的RL后训练框架，源自[slime](https://github.com/THUDM/slime)。
+**vime**是以[**vLLM**](https://github.com/vllm-project/vllm)与[**Megatron-LM**](https://github.com/NVIDIA/Megatron-LM)为核心后端的RL后训练框架，源自[vime](https://github.com/vllm-project/vime)。


The project is derived from the upstream slime framework. Renaming this reference to vime in the Chinese documentation creates a circular statement. Please keep the reference to slime here.

Suggested change

**vime**是以[**vLLM**](https://github.com/vllm-project/vllm)与[**Megatron-LM**](https://github.com/NVIDIA/Megatron-LM)为核心后端的RL后训练框架，源自[vime](https://github.com/vllm-project/vime)。

**vime**是以[**vLLM**](https://github.com/vllm-project/vllm)与[**Megatron-LM**](https://github.com/NVIDIA/Megatron-LM)为核心后端的RL后训练框架，源自[slime](https://github.com/THUDM/slime)。

gemini-code-assist · 2026-06-02T07:18:10Z

 ```

-For screen shots of the wandb, please refer to [pull#370](https://github.com/THUDM/slime/pull/370).
+For screen shots of the wandb, please refer to [pull#370](https://github.com/vllm-project/vime/pull/370).


The pull request #370 refers to the original PR in the upstream THUDM/slime repository where the wandb screenshots are located. Changing this link to vllm-project/vime will break the reference as PR #370 does not exist on the new repository. Please keep the original link to the upstream repository.

Suggested change

For screen shots of the wandb, please refer to [pull#370](https://github.com/vllm-project/vime/pull/370).

For screen shots of the wandb, please refer to [pull#370](https://github.com/THUDM/slime/pull/370).

gemini-code-assist · 2026-06-02T07:18:10Z

 ```

-这个 PR 中记录了 wandb 的截图 [pull#370](https://github.com/THUDM/slime/pull/370)。
+这个 PR 中记录了 wandb 的截图 [pull#370](https://github.com/vllm-project/vime/pull/370)。


The pull request #370 refers to the original PR in the upstream THUDM/slime repository where the wandb screenshots are located. Changing this link to vllm-project/vime will break the reference as PR #370 does not exist on the new repository. Please keep the original link to the upstream repository.

Suggested change

这个 PR 中记录了 wandb 的截图 [pull#370](https://github.com/vllm-project/vime/pull/370)。

这个 PR 中记录了 wandb 的截图 [pull#370](https://github.com/THUDM/slime/pull/370)。

gemini-code-assist · 2026-06-02T07:18:10Z

+# torch_memory_saver pinned to a193d9dd (upstream vime #1916). The newer commit
 # ships a multi-CUDA wheel and requires TMS_CUDA_MAJOR at build time; default it
-# to the running torch's CUDA major (slime #1924).
+# to the running torch's CUDA major (vime #1924).


These comments refer to specific issues/PRs (#1916 and #1924) in the upstream THUDM/slime repository. Changing them to vime makes them untraceable. Please keep the references to slime here.

# torch_memory_saver pinned to a193d9dd (upstream slime #1916). The newer commit\n# ships a multi-CUDA wheel and requires TMS_CUDA_MAJOR at build time; default it\n# to the running torch's CUDA major (slime #1924).

gemini-code-assist · 2026-06-02T07:18:10Z

    # heterogeneous per-group engines (different num_gpus_per_engine). TP is resolved per engine
    # in ``vllm_engine._resolve_vllm_parallel_sizes`` (tp = gpus_per_engine // pp), mirroring
-    # upstream slime's sglang_engine. (pp divisibility is validated there, per engine.)
+    # upstream vime's sglang_engine. (pp divisibility is validated there, per engine.)


This comment refers to the upstream slime project's sglang_engine. Since slime is the upstream project, the term 'upstream slime' is correct. Please revert this change to keep the reference accurate.

Suggested change

# upstream vime's sglang_engine. (pp divisibility is validated there, per engine.)

# upstream slime's sglang_engine. (pp divisibility is validated there, per engine.)

gemini-code-assist · 2026-06-02T07:18:10Z


 def _resolve_vllm_parallel_sizes(args, *, gpus_per_engine: int) -> tuple[int, int]:
-    # Derive TP per-engine from THIS engine's GPU count (matches upstream slime's
+    # Derive TP per-engine from THIS engine's GPU count (matches upstream vime's


This comment refers to the upstream slime project's sglang_engine. Since slime is the upstream project, the term 'upstream slime' is correct. Please revert this change to keep the reference accurate.

Suggested change

# Derive TP per-engine from THIS engine's GPU count (matches upstream vime's

# Derive TP per-engine from THIS engine's GPU count (matches upstream slime's

CalvinXKY · 2026-06-04T02:19:12Z

gemini-code-assist Bot reviewed Jun 2, 2026

View reviewed changes

aoshen02 force-pushed the chore/rename-slime-to-vime-clean branch from aa7e58c to 96a947e Compare June 2, 2026 13:59

aoshen02 added 5 commits June 3, 2026 01:58

Rename slime references to vime

d9b10e3

Tighten vime wording in issue templates

6f35f22

Preserve upstream slime references

a900466

Remove OPD docs and examples

7fb4701

Remove low precision docs and examples

85f7354

aoshen02 force-pushed the chore/rename-slime-to-vime-clean branch from 96a947e to 85f7354 Compare June 3, 2026 01:59

aoshen02 added 11 commits June 3, 2026 08:04

Restore MTP examples

76ce97a

Remove English GLM-4.7 MTP examples

4a01936

Remove MiMo MTP script

5404cb2

Remove Chinese GLM-4.7 MTP examples

5d26377

Restore upstream acknowledgements

309fed3

Resolve merge conflicts and restore provenance text

a44d28a

Restore upstream slime code reference

0fa39f2

Clean up remaining slime references

f19fea9

Restore vLLM GLM rollout scripts

f701371

Remove selected SGLang references

2f1cdd4

Point geo3k guide to vime implementation

3a1f04f

Remove fp8 SGLang note

4c86268

CalvinXKY approved these changes Jun 4, 2026

View reviewed changes

CalvinXKY merged commit 9524dd5 into main Jun 4, 2026
10 of 13 checks passed

aoshen02 mentioned this pull request Jun 4, 2026

fix(test): update vllm_generate_endpoint to post-#68 launch_server_process API #149

Merged

aoshen02 deleted the chore/rename-slime-to-vime-clean branch June 8, 2026 14:17

Navigation Menu

Uh oh!

Rename slime references to vime - #121

Rename slime references to vime#121
CalvinXKY merged 17 commits into
mainfrom
chore/rename-slime-to-vime-clean

aoshen02 commented Jun 2, 2026

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Uh oh!

CalvinXKY commented Jun 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	vime is a reinforcement learning (RL) post-training framework built on [vLLM](https://github.com/vllm-project/vllm) and [Megatron-LM](https://github.com/NVIDIA/Megatron-LM) as its core backends. It is derived from [vime](https://github.com/vllm-project/vime).
	vime is a reinforcement learning (RL) post-training framework built on [vLLM](https://github.com/vllm-project/vllm) and [Megatron-LM](https://github.com/NVIDIA/Megatron-LM) as its core backends. It is derived from [slime](https://github.com/THUDM/slime).

	vime是以[vLLM](https://github.com/vllm-project/vllm)与[Megatron-LM](https://github.com/NVIDIA/Megatron-LM)为核心后端的RL后训练框架，源自[vime](https://github.com/vllm-project/vime)。
	vime是以[vLLM](https://github.com/vllm-project/vllm)与[Megatron-LM](https://github.com/NVIDIA/Megatron-LM)为核心后端的RL后训练框架，源自[slime](https://github.com/THUDM/slime)。

	For screen shots of the wandb, please refer to [pull#370](https://github.com/vllm-project/vime/pull/370).
	For screen shots of the wandb, please refer to [pull#370](https://github.com/THUDM/slime/pull/370).

	这个 PR 中记录了 wandb 的截图 [pull#370](https://github.com/vllm-project/vime/pull/370)。
	这个 PR 中记录了 wandb 的截图 [pull#370](https://github.com/THUDM/slime/pull/370)。

	# upstream vime's sglang_engine. (pp divisibility is validated there, per engine.)
	# upstream slime's sglang_engine. (pp divisibility is validated there, per engine.)

	# Derive TP per-engine from THIS engine's GPU count (matches upstream vime's
	# Derive TP per-engine from THIS engine's GPU count (matches upstream slime's

Uh oh!

Conversation

aoshen02 commented Jun 2, 2026

Summary

Verification

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

CalvinXKY commented Jun 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants