`ibm/granite-instruct` preset for the IBMvLLM adapter is vague and resolves incorrect max_sequence_length

[michael-desmond]

Describe the bug The ibm/granite-instruct preset for the IBMvLLM adapter is not specific to the model being used. I assume its ibm-granite/granite-3.0-8b-instruct and should be named as such to avoid confusion.

Assuming ibm-granite/granite-3.0-8b-instruct then the backend returns the incorrect max_sequence_length of 8192. The correct max_sequence_length is 4096

To Reproduce Steps to reproduce the behavior:

1. Create an GraniteBeeAgent with IBMvLLM backend using preset ibm/granite-instruct
2. Breakpoint the meta method in the corresponding llm.ts
3. Result will indicate a max_sequence_length of 8192

Expected behavior max_sequence_length of 4096

Screenshots / Code snippets

Set-up:

• Bee version: 0.0.41
• Model provider: IBMvLLM