Anchen

fix(semantic_analyzers): fix the false error for noConstAssign

1

## Summary This PR is for fixing https://github.com/rome/tools/issues/3728. The current implementation of noConstAssign did not skip function assignment that caused a false error for the lint rule. ## Test Plan...

fix(semantic_analyzers): style/noShoutyConstants does not recognize multiple uses of a constant.

3

## Summary This is for fixing #3658. The current implementation fires false errors on exported shouty constants, and ignore the shouty constant checking when it's used more than once. ##...

feat(rome_js_analyze): `noRedundantUseStrict` rule

2

## Summary There used to be a parse check of the "use strict" directive and an error would be thrown when encountering redundant "use strict" directives; however, it should be...

feat(rome_js_analyze): add module and classes support for noRedundantUseStrict

1

## Summary Adding module and class support for noRedundantUseStrict rule to address https://github.com/rome/tools/issues/3901 ## Test Plan All tests added as part of PR. `cargo test` passed

cache costheta and sintheta in positional_encoding.py to reduce memory usage and computation

4

In the current implementation of Rope, we are not caching the costheta and sintheta. Instead, we compute the theta on-the-fly for every forward pass. It would be better if we...

Support quantization for non-multiples of 32.

5

Yayi 30b k,v layer has [input_dims=7168, out_dims=112], so it failed to quantize due to error `all dimensions should be divisible by 32 for now`. FYI, here is the implementation of...

Since we already support the hf format model, it is better to merge the deepseek example with the hf llm example to reduce maintenance overhead. so, linear scale rope support...

Can we make hf_llm as a Python package?

2

First of all, thank you for making it available to us. I would love to use it for my project, but currently I have to manually copy the source code...

How can I merge the Lora weight back to the original model weight?

6

Maybe I'm missing something, but I can't find any information on how to merge Lora weight back into the original model. Running the model with a Lora adapter will add...

feat: add yayi2-30b example

1

Yayi2-30b has a score of over 80 MMLU. I performed some fine-tuning on it using qlora, and from my quick test, it appears to be very promising. so, I created...