Rohit Singh Rathaur

[StatLog(German Credit Data)](https://archive.ics.uci.edu/ml/datasets/statlog+(german+credit+data))

Support for lecun normal weight initialization

1

feat: Add LeCun normal weight initialization - Implemented `lecun_normal` function for weight initialization based on fan-in. - Provided default initializer for flexibility. - Marked `lecun_normal` as non-differentiable using ChainRulesCore. It...

- integrated the property `bee_line_distance` to the `Channel` class

3

- integrated the property `latency` to the `OracleLightningNetwork` class

Update dependencies and outdated packages

1

update-modules because those were not maintained.

Made into a `pytest.fixture`

2

It resolves issue #295

https://github.com/RohitRathore1/elk/blob/84e99a36a5050881d85f1510a2486ce46ac1f942/tests/test_smoke_eval.py#L19C1-L20C35

Fix IndexError caused by invalid token IDs in CFGGuide

4

It fixes issue #1232 These changes fix the IndexError caused by invalid token IDs in allowed_tokens_concat by handling eos_token_id appropriately and adjusting token handling in CFGGuide. The updates maintain backward...

CUDA-based RMSNorm for Performance Optimization

2

In continuation of issue #194.. @erfanzar The current `RMSNorm` implementation in [rms_norm](https://github.com/erfanzar/EasyDeL/blob/main/easydel/kernels/rms_norm.py) uses `Pallas` for TPUs and falls back to basic JAX operations for other platforms. On NVIDIA GPUs, a...

[c10d] Fix _set_pg_timeout not working for Gloo backend

8

Fixes #165422 The issue was that `ProcessGroupGloo` operations were capturing the timeout value at `AsyncWork` construction time and storing it in a const member variable. When operations executed, they used...

[checkpoint] Fix get_optimizer_state_dict modifying optimizer state

10

Fixes #164929 cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @pragupta @msaroufim @dcci @LucasLLC @pradeepfn