fix: pin transformers below 5.0.0 to fix generation_config warning

[Ravi-Poddar26]

Problem

Running any endpoint produces this warning on every request:

Both max_new_tokens (=1024) and max_length(=20) seem to
have been set. max_new_tokens will take precedence.

Root Cause

requirements.txt specifies transformers>=4.45.2 which
allows transformers 5.x to be installed. transformers 5.x
introduced stricter generation config handling — it raises
a warning when the model's generation_config.json and
pipeline parameters conflict.

transformers 4.x silently merged these configs.
transformers 5.x warns on every single request.

Fix

Pin transformers to <5.0.0 until the codebase is updated
for full transformers 5.x compatibility.

Testing

Tested on:

• Windows 11
• Python 3.13
• torch 2.10.0
• transformers 5.3.0 → warning appears
• transformers 4.x → warning gone

Fixes #63

---

## Before Clicking Submit — Check This

Make sure at the top of the PR page it shows:

base repository: sugarlabs/sugar-ai base: main
head repository: Ravi-Poddar26/sugar-ai compare: fix/pin-transformers-version