Fabian Joswig
Fabian Joswig
This patch speeds up the construction of the exponential Clover term in the compact layout on GPU architectures. The exponentiation is now performed on the accelerator and the inverse is...
When running a `GenerationSession` with `gather_generation_logits` and the first token that is generated has `end_id` then `outputs['generation_logits']` is an empty list. This crashes the session in `_prepare_outputs` https://github.com/NVIDIA/TensorRT-LLM/blob/3d56a445e8ebf888e78be638faf6beec0a78f3c2/tensorrt_llm/runtime/model_runner.py#L253 with ```...
pyerrors raises multiple errors in combination with the numpy 2 release candidate - I fixed the relevant breaking changes in autograd in the following PR https://github.com/HIPS/autograd/pull/618 - matplotlib raises `ImportError:...
In regression tests of the `ModelRunnerCpp` vs `ModelRunner` we noticed that the `stop_words_list` feature does not work properly for the `ModelRunner` and `batch_size` > 1. Depending on the input we...
This PR includes a new module `ConservedBilinear` which can compute an NPR vertex with a conserved current insertion (so far only for 4d actions). I also started writing a test...
Running a module with an output used by no other module causes the following db error `database error: error executing query 'INSERT INTO "objects" VALUES(16,'module_name',4,0,'undef',4294967295);'` which does not directly hint...
Follow-up to https://github.com/NVIDIA/TensorRT-LLM/pull/2056, which I developed against v0.10. After upgrading to last week's dev release, I noticed that the `exclude_modules` matching pattern in quantize.py had changed to `fnmatchcase` in https://github.com/NVIDIA/TensorRT-LLM/pull/1793...
As a first step towards adding type hints, I added automatic type hints with monkeytype. Hints for the more complicated methods will need additional manual work but for the simpler...
In this PR I remove all mutables from default arguments (see for example [this article](https://florimond.dev/en/posts/2018/08/python-mutable-defaults-are-the-source-of-all-evil)). I also corrected smaller stilistic issues and added missing docstrings.
I sometimes want to do Gaussian error propagation and find the syntax for creating a scalar cov_Obs a bit inconvenient. For this reason I propose a `Meas` class that simplifies...