Skip to content

fix(model): replace INVALID_TOKEN sentinel with real first-token decode in prefill for whisper models #1028

fix(model): replace INVALID_TOKEN sentinel with real first-token decode in prefill for whisper models

fix(model): replace INVALID_TOKEN sentinel with real first-token decode in prefill for whisper models #1028

name: Trigger FSW PR CI for vLLM-RBLN
on:
pull_request:
types: [opened, synchronize, reopened]
workflow_dispatch:
inputs:
ref:
description: "ref to checkout"
required: false
type: string
python_version:
description: "Python version to use"
required: false
type: string
default: "3.12.9"
vllm_upstream_tag:
description: "vLLM upstream tag/version to use for tests (e.g., v0.9.1)"
required: false
type: string
jobs:
trigger-vllm-rbln-fsw-pr-ci:
if: github.repository_owner == 'RBLN-SW' && github.event_name == 'pull_request'
runs-on: fsw-r18s-atom-ext-rbln-sw-general-c1m12
strategy:
fail-fast: false
matrix:
vllm_version:
- "0.18.0+cpu"
steps:
- name: Dispatch to fsw-integration (vllm ${{ matrix.vllm_version }})
uses: Wandalen/wretry.action@v3
with:
action: peter-evans/repository-dispatch@v4
attempt_limit: 5
attempt_delay: 2000
with: |
token: ${{ secrets.GIT_PAT }}
repository: rebellions-sw/fsw-integration
event-type: vllm-rbln-pr-event
client-payload: |
{
"ref": "${{ github.event.pull_request.head.ref }}",
"sha": "${{ github.event.pull_request.head.sha }}",
"pr_number": "${{ github.event.number }}",
"actor": "${{ github.actor }}",
"vllm_version": "${{ matrix.vllm_version }}"
}