-
Hi,
Hope you are doing well!
Is there a way you can harness the `tesseract` (which is already installed on my system - it just needs to run using a shortcut or be available via the menubar like …
-
Hi,
It seems there is a problem with lm_eval when I am not setting 'max_length' with some tasks (at least GEM/wiki_lingua_en). When I am letting 'max_length' with default value, I get this error mess…
-
When the harness/fuzzers are running properly, we should report some sort of status to EC2Reporter.
Probably:
* Platform detection
* Harness is running
* Which fuzzer is running
* How much fuzz…
-
Hello,
I usually use additional parameters in my harness plans like the signal direction or the voltage levels as additional layers of security. One could add them to the pinlabels ofc but I would …
T0jan updated
1 month ago
-
The HumanEval and HumanEval+ benchmarks are stables for benchmarking code capabilities of base LLMs. It would be nice to include them in `lighteval` so one doesn't have to switch to another framework …
-
### Describe the bug
- Use a callable to initialise a component on load
- Setup a change event handler
Expect: Change event fires since callable has updated value on component.
Actual:
Markdown -…
-
Hi, I'm currently using kAFL 0.7
When the Sleep() function includes the harness, the ntoskrnl crashes.
The reason for this is that if I run a harness without Sleep() function, it will work normally,…
-
Hi, is there any plan for adding a markdown preview?
Thanks in advance.
-
### Your current environment
```text
The output of `python collect_env.py`
```
### 🐛 Describe the bug
vLLM has an issue where we can go OOM if too many `logprobs` are requested.
The reason t…
-
Harness new version so called NextGen doesn't use graphQL API anymore and hence the connection is broken.
The connection tests failed.
Please verify your API ID or key are correct. Error: GraphQL…