Skip to content

[dont merge]. Diff 0.2 -> 0.3 - #2

Open
rakkit wants to merge 407 commits into
TrustLLMeu:trustllm-0.2from
rakkit:trustllm-0.3
Open

[dont merge]. Diff 0.2 -> 0.3#2
rakkit wants to merge 407 commits into
TrustLLMeu:trustllm-0.2from
rakkit:trustllm-0.3

Conversation

@rakkit

@rakkit rakkit commented Nov 19, 2025

Copy link
Copy Markdown

No description provided.

idoh and others added 30 commits November 16, 2025 16:31
The current `convert_to_hf.py` does not support `export_dtype`, which
makes it `float32` by default. This PR adds support for export dtypes of
`["float16", "bfloat16", "float32"]`.
This PR integrates the changes in pytorch#1970 to compiler toolkit (applying
`joint_ac_pass` on the joint graph graph to tag nodes based on
`reshard_after_forward` flag)

Also did some refactor for applying graph passes in compiler toolkit
experiments. We will have two kinds of passes

1. joint_custom_passes: these are passes to be applied on the captured
joint graph before partitioner. By default we
`validate_flex_attn_annotation_pass` and `fsdp_reshard_after_fwd_pass`

2. compiler_passes: there are passes to be applied on partitioned fwd
and bwd graphs as backend optimizations. By default there is none. We
can indicate `autobucketing_reordering_pass` and
`regional_inductor_pass` using configs.
This adds JSON data support.
More reuse, inspired by BitBLAS modeling code.
Inspired by DeepSeekMoE.
- Close to original implementation
- Conflates light and constrained versions
- Uses (old) shared Muon functionality
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.

5 participants