Mert Unsal
Mert Unsal
I realized that integration steps cannot be viewed for an integral such as exp(x)/(1+exp(2*x)) due to lack of support for printing of the PiecewiseRule. Added a very simple functionality to...
When I run ``` import sympy as sp from sympy.integrals.manualintegrate import integral_steps x = sp.Symbol('x') f = 1/sp.cos(x) integral_steps(f, x) ``` I get `DontKnowRule(integrand=1/cos(x), variable=x)` even though this is simple...
### Reminder - [x] I have read the above rules and searched the existing issues. ### System Info - `llamafactory` version: 0.9.2.dev0 - Platform: Linux-5.4.0-125-generic-x86_64-with-glibc2.31 - Python version: 3.12.7 -...
### Reminder - [x] I have read the above rules and searched the existing issues. ### System Info - `llamafactory` version: 0.9.2.dev0 - Platform: Linux-5.4.0-125-generic-x86_64-with-glibc2.31 - Python version: 3.12.7 -...
## Summary by cubic Updated langchain dependencies and added support for llama models using json_mode for tool calling. - **Dependencies** - Bumped langchain-core to 0.3.63 and langchain-openai to 0.3.18. -...
## Summary by cubic Added a file system for browser agents, allowing them to read, write, and append files during tasks. Agents now use a results.txt file to save findings...
1. Improved example agent messages to more complicated scenarios. 2. Fixed small syntax problems 3. Agent now needs to **first** reason and then conclude whether the previous step was successful...
## Summary by cubic Added a new reasoning_task_status field to track and explain progress toward the ultimate goal in agent responses. - **New Features** - Updated response format, logging, and...
During rollout, the train batch is passed as is and we generate n_samples for each prompt: However, during validation, the test batch is repeated as the number of validation samples...