Nightly #3664

danielhanchen · 2025-12-01T15:13:09Z

No description provided.

* Enable FP8 + RL training for bf16 models **Summary:** Enable FP8 + RL training using TorchAO for 1.33x faster training and 42% less model memory usage: - We quantize the frozen LoRA weights into fp8 and keep the LoRA adapters in bf16 - We leverage TorchAO's `Float8Tensor`, which calls into fbgemm's fp8 x fp8 rowwise matmul kernel - For now, we need to do an offline quantization first, because vllm doesn't support on-the-fly quantization for torchao yet (this is in progress: vllm-project/vllm#26327) **Example usage:** ``` model, tokenizer = FastLanguageModel.from_pretrained( model_name = "unsloth/Qwen3-8B-Base", max_seq_length = 2048, load_in_4bit = False, fast_inference = True, max_lora_rank = 32, load_in_fp8 = True, # set this to True ) \# the rest is the same as before model = FastLanguageModel.get_peft_model(...) ``` **Initial results:** ``` \# fp8 {'train_runtime': 1725.4337, 'train_samples_per_second': 0.232, 'train_steps_per_second': 0.058, 'train_loss': 0.00015715716748673002, 'epoch': 0.01} \# bf16 {'train_runtime': 2297.8145, 'train_samples_per_second': 0.174, 'train_steps_per_second': 0.044, 'train_loss': 0.00016081033063528594, 'epoch': 0.01} ``` <img width="1199" height="448" alt="Screenshot 2025-11-11 at 4 10 50 PM" src="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/user-attachments/assets/b6304afd-89e9-42b1-8064-775807e17b23" /> Test script: https://gist.github.com/andrewor14/5b85119fae46845d07b608d420907423 **Requires:** - pytorch/ao#3158 (torchao nightly or 0.15.0+) - unslothai/unsloth-zoo#351 * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update utils.py * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * _get_inference_mode_context_manager * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update utils.py * Update utils.py * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* make loading gpt-oss-BF16 faster. Linked to unsloth-zoo PR #314 * fix model loading and clean merged model directory * revert default quant * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert mapper.py --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Add 128x128 PerBlock FP8 + RL **Summary:** Following #3440, this PR extends torchao FP8 + RL support to also handle 128x128 PerBlock granularity (in addition to PerRow). **Example usage:** ``` model, tokenizer = FastLanguageModel.from_pretrained( model_name = "unsloth/Qwen3-8B-Base", max_seq_length = 2048, load_in_4bit = False, fast_inference = True, max_lora_rank = 32, load_in_fp8 = "block", # or "row" or True ) ``` **Initial results:** TBD **Note:** - Requires pytorch/ao#3370 * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

for more information, see https://pre-commit.ci

@danielhanchen

* vllm sampling params fix * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * do not patch base_trainer * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * seperate vllm fixes * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Apply suggestion from @danielhanchen * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks" This reverts commit fbb98c5. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks" This reverts commit c64d5b4. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks" This reverts commit c156545. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Daniel Han <danielhanchen@gmail.com>

This reverts commit 6c47dc5.

* vllm sampling params fix * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * do not patch base_trainer * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * seperate vllm fixes * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fixup deletion * Fix indentation * revert to old style --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

gemini-code-assist · 2025-12-01T15:13:33Z

Summary of Changes

Hello @danielhanchen, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request encompasses a series of maintenance and compatibility updates across the codebase. It ensures that the project remains aligned with external library changes, particularly concerning vLLM and trl, by introducing specific patches. Additionally, it includes minor code refinements, such as typo corrections and improved debugging output, to enhance overall stability and developer experience. The project's internal version has also been incremented to reflect these changes.

Highlights

Dependency Update: The unsloth_zoo dependency has been updated to version 2025.11.6 in pyproject.toml for both huggingface and colab-new configurations.
vLLM Compatibility Fix: A new fix, fix_vllm_guided_decoding_params, has been introduced to address a breaking change in vLLM where GuidedDecodingParams was renamed to StructuredOutputsParams. This ensures compatibility with trl which still expects the old name.
Version Increment: The internal __version__ string in unsloth/models/_utils.py has been updated to 2025.11.5.
Typo Correction: A minor typo from 'Inferene' to 'Inference' has been corrected in comments across multiple model files (Falcon, Gemma, Llama, Mistral, Qwen2, Qwen3, Qwen3-MoE) related to KV Cache and CUDAGraphing.
Qwen3-MoE Model Enhancements: Modifications were made to Qwen3MoeSparseMoeBlock_fast_forward to explicitly cast routing_weights and final_X to torch.float32 for improved precision, and residual additions in Qwen3MoeDecoderLayer_fast_forward were simplified from hidden_states = residual + hidden_states to hidden_states += hidden_states.
TRL Trainer Patching Improvements: Enhanced debugging messages have been added to unsloth/models/rl.py for trl trainer patching failures. The dtype retrieval for input embeddings was refined, patch_module's overwrite parameter was set to False, and base_trainer is now explicitly excluded from patching.
Console Output Formatting: The statistics printout in unsloth/models/vision.py now includes improved formatting for model architecture names, ensuring consistent capitalization (e.g., _VL_, _MoE).

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

gemini-code-assist · 2025-12-01T15:15:09Z

Summary of Changes

Hello @danielhanchen, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request delivers a set of maintenance and compatibility enhancements across the codebase. Key updates include bumping a core dependency version, implementing a crucial fix for vLLM integration with trl to ensure proper guided decoding, and refining the RL trainer patching mechanism with better error reporting and more precise type handling. Additionally, minor textual corrections and internal version updates contribute to overall code quality and stability.

Highlights

Dependency Updates: The unsloth_zoo dependency has been updated to >=2025.11.6 in pyproject.toml for both huggingface and colab-new configurations.
vLLM Compatibility Fix: A new fix, fix_vllm_guided_decoding_params, was introduced to address a renaming issue in vLLM where GuidedDecodingParams became StructuredOutputsParams, ensuring compatibility with trl.
Version Bump: The internal version of the library has been incremented from 2025.11.4 to 2025.11.5 in unsloth/models/_utils.py.
Typo Corrections: A consistent typo, 'Inferene', was corrected to 'Inference' in comments across several model files (falcon_h1, gemma, gemma2, llama, mistral, qwen2, qwen3, qwen3_moe).
Qwen3 MoE Enhancements: Explicit torch.float32 dtype casting was added for torch_nn_functional_softmax and final_X initialization, and in-place addition hidden_states += hidden_states was adopted for efficiency in unsloth/models/qwen3_moe.py.
RL Trainer Patching Improvements: Enhanced debugging output was added for trl.trainer imports and class lookups, dtype inference was refined to use model.get_input_embeddings().weight.dtype, overwrite was set to False for patch_code, and base_trainer was explicitly excluded from patching in unsloth/models/rl.py.
Statistics Output Formatting: The display of arch_name in the statistics output was improved by correctly capitalizing 'VL' and 'MoE' in unsloth/models/vision.py.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

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".

chatgpt-codex-connector · 2025-12-01T15:15:32Z

unsloth/models/qwen3_moe.py

            _flag_for_generation = self._flag_for_generation,
        )
-        hidden_states = residual + hidden_states
+        hidden_states += hidden_states


Preserve residual connection in Qwen3Moe attention path

When use_cache is enabled, Qwen3MoeDecoderLayer_fast_forward now does hidden_states += hidden_states instead of adding the saved residual. This doubles the self-attention output and removes the skip connection, so cached generation will produce incorrect activations compared to training/eager mode. It should add the stored residual back in this spot.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2025-12-01T15:15:32Z

unsloth/models/qwen3_moe.py

            self.mlp, hidden_states
        )
-        hidden_states = residual + hidden_states
+        hidden_states += hidden_states


Restore MoE residual addition in generation branch

In the same use_cache branch, the MoE block now applies hidden_states += hidden_states instead of combining with the pre-MLP residual. This removes the skip connection and doubles the MoE activations during cached generation, yielding different outputs from the non-cached path. It should add the stored residual instead of self-adding.

Useful? React with 👍 / 👎.

gemini-code-assist

Code Review

This pull request includes several updates, likely part of a nightly build. It bumps the unsloth_zoo dependency version, fixes some typos in comments, and adds a patch for vLLM compatibility. However, I've found two critical bugs in unsloth/models/qwen3_moe.py where residual connections are incorrectly implemented, causing the hidden states to be doubled instead of having the residual added. These need to be fixed to ensure correct model behavior.

gemini-code-assist · 2025-12-01T15:16:25Z

unsloth/models/qwen3_moe.py

            _flag_for_generation = self._flag_for_generation,
        )
-        hidden_states = residual + hidden_states
+        hidden_states += hidden_states


This line seems to have a typo. It should be hidden_states += residual to add the residual connection back. Currently, it's hidden_states += hidden_states, which doubles the hidden_states tensor. This is likely not the intended behavior for a residual connection.

Suggested change

hidden_states += hidden_states

hidden_states += residual

gemini-code-assist · 2025-12-01T15:16:25Z

unsloth/models/qwen3_moe.py

            self.mlp, hidden_states
        )
-        hidden_states = residual + hidden_states
+        hidden_states += hidden_states


Similar to the previous comment, this line appears to be incorrect for a residual connection. It should probably be hidden_states += residual instead of hidden_states += hidden_states to correctly add the residual.

Suggested change

hidden_states += hidden_states

hidden_states += residual

gemini-code-assist · 2025-12-01T15:16:27Z

Summary of Changes

Hello @danielhanchen, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request delivers a set of maintenance and compatibility updates across the Unsloth library. It ensures that dependencies are current, addresses a specific compatibility challenge with vLLM and trl by introducing a targeted patch, and includes minor code refinements such as typo corrections and improved error handling in the RL training module. The overall version of the library is also updated.

Highlights

Dependency Update: The unsloth_zoo dependency has been updated to >=2025.11.6 in the huggingface and colab-new extra requirements within pyproject.toml.
vLLM Compatibility Patch: A new fix, fix_vllm_guided_decoding_params, was introduced to address a renaming issue in vLLM where GuidedDecodingParams became StructuredOutputsParams, ensuring compatibility with trl.
Version Bump: The internal version of Unsloth has been incremented from 2025.11.4 to 2025.11.5.
Typo Corrections: A consistent typo 'Inferene' was corrected to 'Inference' in comments across multiple model files (falcon_h1, gemma, gemma2, llama, mistral, qwen2, qwen3) related to KV Cache and CUDAGraphing.
Qwen3 MoE Model Enhancements: Changes were made to Qwen3MoeSparseMoeBlock_fast_forward to explicitly use torch.float32 for softmax and tensor initialization, and hidden_states = residual + hidden_states was optimized to hidden_states += hidden_states in Qwen3MoeDecoderLayer_fast_forward.
Improved RL Trainer Patching: Error logging was enhanced for trl.trainer patching, dtype retrieval was refined to use model.get_input_embeddings().weight.dtype, the overwrite parameter was set to False, and base_trainer is now explicitly excluded from patching.
Statistics Output Formatting: The display of arch_name in the statistics output for vision models was improved by standardizing casing for _Vl_ and _Moe.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

gemini-code-assist

Code Review

The pull request primarily focuses on version updates, minor bug fixes, and logging improvements across various model implementations. A new vLLM compatibility patch has been added to import_fixes.py to handle GuidedDecodingParams renaming. Several model files (falcon_h1.py, gemma.py, gemma2.py, llama.py, mistral.py, qwen2.py, qwen3.py, qwen3_moe.py) include fixes for a typo in comments. The pyproject.toml file has been updated to reflect a new unsloth_zoo version. Logging in unsloth/models/rl.py has been enhanced to provide more detailed messages during trainer patching. A critical bug was identified in unsloth/models/qwen3_moe.py where residual connections are incorrectly applied.

gemini-code-assist · 2025-12-01T15:16:53Z

unsloth/models/qwen3_moe.py

            _flag_for_generation = self._flag_for_generation,
        )
-        hidden_states = residual + hidden_states
+        hidden_states += hidden_states


This line appears to be a logical error. hidden_states += hidden_states will double the hidden_states value, effectively computing 2 * hidden_states. Based on the previous line residual = hidden_states, it seems the intention was to add the residual connection, i.e., hidden_states = residual + hidden_states or hidden_states += residual.

Suggested change

hidden_states += hidden_states

hidden_states += residual

gemini-code-assist · 2025-12-01T15:16:53Z

unsloth/models/qwen3_moe.py

            self.mlp, hidden_states
        )
-        hidden_states = residual + hidden_states
+        hidden_states += hidden_states


This line appears to be a logical error. hidden_states += hidden_states will double the hidden_states value, effectively computing 2 * hidden_states. Based on the previous line residual = hidden_states, it seems the intention was to add the residual connection, i.e., hidden_states = residual + hidden_states or hidden_states += residual.

Suggested change

hidden_states += hidden_states

hidden_states += residual

gemini-code-assist

Code Review

This pull request includes nightly updates, primarily version bumps and a new fix for vLLM guided decoding. While most changes are improvements, I've identified two critical bugs in unsloth/models/qwen3_moe.py related to incorrect residual connections. These have been commented on with suggested fixes. Other changes include better debugging logs and minor code refinements.

gemini-code-assist · 2025-12-01T15:17:51Z

unsloth/models/qwen3_moe.py

            _flag_for_generation = self._flag_for_generation,
        )
-        hidden_states = residual + hidden_states
+        hidden_states += hidden_states


This appears to be a bug in the residual connection. The code was changed from hidden_states = residual + hidden_states to hidden_states += hidden_states, which is equivalent to hidden_states = 2 * hidden_states. This doubles the attention output instead of adding the residual. It should be hidden_states += residual.

Suggested change

hidden_states += hidden_states

hidden_states += residual

unsloth/models/rl.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

for more information, see https://pre-commit.ci

This reverts commit 9bf82bb.

for more information, see https://pre-commit.ci

This reverts commit 1146769.

for more information, see https://pre-commit.ci

This reverts commit c427be9.

andrewor14 and others added 30 commits November 19, 2025 23:51

Update __init__.py

919fb1e

Update loader_utils.py

c142283

Update loader_utils.py

e28b7c2

Version

c11af7a

Merge branch 'main' into nightly

4fb4ec0

Update vision.py

2bd26ee

Update rl.py

1f6c0a9

Add torch 2.9.1

874e6cb

Fix auto installer

822eb94

Update fp8.py

95b0dbc

Float8

4246d3e

Update fp8.py

a3e6ef0

Update mapper.py

c4d5e9a

Update mapper.py

acf9d53

Update loader_utils.py

05d80fb

Update loader.py

5e419d5

Update fp8.py

ed43bcb

Versioning

51247e9

[pre-commit.ci] auto fixes from pre-commit.com hooks

c1a0208

for more information, see https://pre-commit.ci

Merge branch 'main' into nightly

7f2eb33

Merge branch 'main' into nightly

b9c963c

Merge branch 'main' into nightly

0581621

Merge branch 'main' into nightly

43d931c

Merge branch 'main' into nightly

c04d140

Merge branch 'main' into nightly

6234e90

Update rl.py

5aa7c17

Update vision.py

f7790f0

danielhanchen and others added 9 commits November 30, 2025 23:37

Update qwen3_moe.py

2b37b9e

Typos

029f33d

Update vision.py

ee4980f

Merge branch 'main' into nightly

afc6ac8

Revert "[FIX] Vllm guided decoding params (#3662)"

3d0be20

This reverts commit 6c47dc5.

Update rl.py

d7c6fc1

Verisoning

6e8de57

chatgpt-codex-connector bot reviewed Dec 1, 2025

View reviewed changes

gemini-code-assist bot reviewed Dec 1, 2025

View reviewed changes

danielhanchen and others added 9 commits December 1, 2025 07:19

Update qwen3_moe.py

d71186a

Update unsloth/models/rl.py

a9a3db3

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Update unsloth/models/rl.py

68c70d1

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

[pre-commit.ci] auto fixes from pre-commit.com hooks

9bf82bb

for more information, see https://pre-commit.ci

Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks"

d5ab983

This reverts commit 9bf82bb.

[pre-commit.ci] auto fixes from pre-commit.com hooks

1146769

for more information, see https://pre-commit.ci

Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks"

77bb017

This reverts commit 1146769.

[pre-commit.ci] auto fixes from pre-commit.com hooks

c427be9

for more information, see https://pre-commit.ci

Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks"

1679dde

This reverts commit c427be9.

danielhanchen merged commit 1679dde into main Dec 1, 2025
2 of 3 checks passed

Uh oh!

Nightly #3664

Nightly #3664

Uh oh!

Conversation

danielhanchen commented Dec 1, 2025

Uh oh!

gemini-code-assist bot commented Dec 1, 2025

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot commented Dec 1, 2025

Summary of Changes

Highlights

Footnotes

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Dec 1, 2025

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot Dec 1, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Dec 1, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Dec 1, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot commented Dec 1, 2025

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Dec 1, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Dec 1, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Dec 1, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants