Skip to content

fix: correct enable_activation_offload config parameter name#4692

Merged
vermouth1992 merged 1 commit intoverl-project:mainfrom
yurekami:fix/issue-4672-activation-offload-typo
Dec 28, 2025
Merged

fix: correct enable_activation_offload config parameter name#4692
vermouth1992 merged 1 commit intoverl-project:mainfrom
yurekami:fix/issue-4672-activation-offload-typo

Conversation

@yurekami
Copy link
Contributor

@yurekami yurekami commented Dec 27, 2025

Summary

Fixes #4672 - The shell scripts were using enable_activation_offloading but the HFModelConfig field is named enable_activation_offload (without the "ing" suffix).

This mismatch caused OmegaConf to fail with:

omegaconf.errors.ConfigKeyError: Key 'enable_activation_offloading' not in 'HFModelConfig'

Changes

Fixed the parameter name in two shell scripts:

  • examples/sglang_multiturn/run_qwen2.5-0.5b_gsm8k_multiturn_w_interaction.sh
  • tests/special_e2e/run_gsm8k_fsdp_sgl_multiturn_sf_tool.sh

Note

The function that implements the feature is correctly named enable_activation_offloading() (with "ing"), but the config field that controls it is enable_activation_offload (without "ing"). The worker code correctly uses the config field name.

…oject#4672)

The shell scripts were using `enable_activation_offloading` but the
HFModelConfig field is named `enable_activation_offload` (without "ing").

This mismatch caused OmegaConf to fail with:
  Key 'enable_activation_offloading' not in 'HFModelConfig'

Fixed files:
- examples/sglang_multiturn/run_qwen2.5-0.5b_gsm8k_multiturn_w_interaction.sh
- tests/special_e2e/run_gsm8k_fsdp_sgl_multiturn_sf_tool.sh

Fixes verl-project#4672

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request correctly fixes a typo in the configuration parameter name for enabling activation offloading. The parameter was changed from enable_activation_offloading to enable_activation_offload in two shell scripts, which aligns with the expected field name in HFModelConfig and resolves a ConfigKeyError. The change is accurate and well-targeted. It's worth noting the underlying inconsistency between the function name enable_activation_offloading() and the configuration key enable_activation_offload, which was the root cause of this bug. Aligning these names in a future refactoring would improve clarity and prevent similar issues. The current fix is correct and I have no further comments.

@vermouth1992 vermouth1992 merged commit a93be2e into verl-project:main Dec 28, 2025
3 of 4 checks passed
boren-ms pushed a commit to boren-ms/verl that referenced this pull request Dec 30, 2025
…oject#4692)

## Summary

Fixes verl-project#4672 - The shell scripts were using
`enable_activation_offloading` but the `HFModelConfig` field is named
`enable_activation_offload` (without the "ing" suffix).

This mismatch caused OmegaConf to fail with:
```
omegaconf.errors.ConfigKeyError: Key 'enable_activation_offloading' not in 'HFModelConfig'
```

### Changes

Fixed the parameter name in two shell scripts:
-
`examples/sglang_multiturn/run_qwen2.5-0.5b_gsm8k_multiturn_w_interaction.sh`
- `tests/special_e2e/run_gsm8k_fsdp_sgl_multiturn_sf_tool.sh`

### Note

The function that implements the feature is correctly named
`enable_activation_offloading()` (with "ing"), but the config field that
controls it is `enable_activation_offload` (without "ing"). The worker
code correctly uses the config field name.

## Test plan

- [x] Verified HFModelConfig has `enable_activation_offload` field (line
73 of `verl/workers/config/model.py`)
- [ ] Run the example script to verify it works

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: yurekami <yurekami@users.noreply.github.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
jsfanfanfan pushed a commit to meituan-search/verl that referenced this pull request Jan 9, 2026
…oject#4692)

## Summary

Fixes verl-project#4672 - The shell scripts were using
`enable_activation_offloading` but the `HFModelConfig` field is named
`enable_activation_offload` (without the "ing" suffix).

This mismatch caused OmegaConf to fail with:
```
omegaconf.errors.ConfigKeyError: Key 'enable_activation_offloading' not in 'HFModelConfig'
```

### Changes

Fixed the parameter name in two shell scripts:
-
`examples/sglang_multiturn/run_qwen2.5-0.5b_gsm8k_multiturn_w_interaction.sh`
- `tests/special_e2e/run_gsm8k_fsdp_sgl_multiturn_sf_tool.sh`

### Note

The function that implements the feature is correctly named
`enable_activation_offloading()` (with "ing"), but the config field that
controls it is `enable_activation_offload` (without "ing"). The worker
code correctly uses the config field name.

## Test plan

- [x] Verified HFModelConfig has `enable_activation_offload` field (line
73 of `verl/workers/config/model.py`)
- [ ] Run the example script to verify it works

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: yurekami <yurekami@users.noreply.github.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
vyomakesh0728 added a commit to vyomakesh0728/verl that referenced this pull request Jan 22, 2026
…oject#4692)

## Summary

Fixes verl-project#4672 - The shell scripts were using
`enable_activation_offloading` but the `HFModelConfig` field is named
`enable_activation_offload` (without the "ing" suffix).

This mismatch caused OmegaConf to fail with:
```
omegaconf.errors.ConfigKeyError: Key 'enable_activation_offloading' not in 'HFModelConfig'
```

### Changes

Fixed the parameter name in two shell scripts:
-
`examples/sglang_multiturn/run_qwen2.5-0.5b_gsm8k_multiturn_w_interaction.sh`
- `tests/special_e2e/run_gsm8k_fsdp_sgl_multiturn_sf_tool.sh`

### Note

The function that implements the feature is correctly named
`enable_activation_offloading()` (with "ing"), but the config field that
controls it is `enable_activation_offload` (without "ing"). The worker
code correctly uses the config field name.

## Test plan

- [x] Verified HFModelConfig has `enable_activation_offload` field (line
73 of `verl/workers/config/model.py`)
- [ ] Run the example script to verify it works

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: yurekami <yurekami@users.noreply.github.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
sophiayyya pushed a commit to sophiayyya/verl that referenced this pull request Jan 25, 2026
…oject#4692)

## Summary

Fixes verl-project#4672 - The shell scripts were using
`enable_activation_offloading` but the `HFModelConfig` field is named
`enable_activation_offload` (without the "ing" suffix).

This mismatch caused OmegaConf to fail with:
```
omegaconf.errors.ConfigKeyError: Key 'enable_activation_offloading' not in 'HFModelConfig'
```

### Changes

Fixed the parameter name in two shell scripts:
-
`examples/sglang_multiturn/run_qwen2.5-0.5b_gsm8k_multiturn_w_interaction.sh`
- `tests/special_e2e/run_gsm8k_fsdp_sgl_multiturn_sf_tool.sh`

### Note

The function that implements the feature is correctly named
`enable_activation_offloading()` (with "ing"), but the config field that
controls it is `enable_activation_offload` (without "ing"). The worker
code correctly uses the config field name.

## Test plan

- [x] Verified HFModelConfig has `enable_activation_offload` field (line
73 of `verl/workers/config/model.py`)
- [ ] Run the example script to verify it works

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: yurekami <yurekami@users.noreply.github.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Key 'enable_activation_offloading' not in 'HFModelConfig'

2 participants