feat(vlm): add num_tokens as attribtue for VlmPrediction (#2489)

* feat: add num_tokens as attribtue for VlmPrediction

* feat: implement tokens tracking for api_vlm

Signed-off-by: Peter El Hachem <peter.el.hachem@ibm.com>

* DCO Remediation Commit for ElHachem02 <peterelhachem02@gmail.com>

I, ElHachem02 <peterelhachem02@gmail.com>, hereby add my Signed-off-by to this commit: 311287f562

Signed-off-by: Peter El Hachem <peter.el.hachem@ibm.com>

* DCO Remediation Commit for ElHachem02 <peterelhachem02@gmail.com>

I, ElHachem02 <peterelhachem02@gmail.com>, hereby add my Signed-off-by to this commit: 311287f562

Signed-off-by: ElHachem02 <peterelhachem02@gmail.com>

* update return type

Signed-off-by: ElHachem02 <peterelhachem02@gmail.com>

* add time recorder for vlm inference and track generated token ids depending on config

Signed-off-by: ElHachem02 <peterelhachem02@gmail.com>

* update num_tokens to have None as value on exception

Signed-off-by: ElHachem02 <peterelhachem02@gmail.com>

* set default value of num_tokens to None

Signed-off-by: ElHachem02 <peterelhachem02@gmail.com>

---------

Signed-off-by: Peter El Hachem <peter.el.hachem@ibm.com>
Signed-off-by: ElHachem02 <peterelhachem02@gmail.com>
Signed-off-by: peets <100425207+ElHachem02@users.noreply.github.com>
Co-authored-by: Peter El Hachem <peter.el.hachem@ibm.com>
This commit is contained in:
peets
2025-10-28 17:18:44 +01:00
committed by GitHub
parent cdffb47b9a
commit b6c892b505
8 changed files with 71 additions and 20 deletions

View File

@@ -207,6 +207,8 @@ class VlmPrediction(BaseModel):
text: str = ""
generated_tokens: list[VlmPredictionToken] = []
generation_time: float = -1
num_tokens: Optional[int] = None
stop_reason: Optional[str] = None # todo define an enum for possible stop reasons
class ContainerElement(