feat: replace cuda13 flavour with per-GPU-generation packages

Build separate packages for each GPU generation instead of a single
cuda13 package:
- mistralrs-ampere (sm_86, RTX 3060)
- mistralrs-ada (sm_89, RTX 4090)
- mistralrs-blackwell (sm_120, RTX 5090)

All use the same CUDA 13.0 toolkit and features (cuda, cudnn, flash-attn,
nccl), varying only the compute capability target.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.gitea/workflows/build-prerelease.yml

@@ -30,7 +30,23 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - name: cuda13
+          - name: ampere
+            fedora_version: "43"
+            runner: cuda-13.0
+            cuda_home: /usr/local/cuda-13.0
+            cargo_features: "cuda cudnn flash-attn nccl"
+            compute_caps: "86"
+            build_jobs: 12
+            nvcc_threads: 4
+          - name: ada
+            fedora_version: "43"
+            runner: cuda-13.0
+            cuda_home: /usr/local/cuda-13.0
+            cargo_features: "cuda cudnn flash-attn nccl"
+            compute_caps: "89"
+            build_jobs: 12
+            nvcc_threads: 4
+          - name: blackwell
             fedora_version: "43"
             runner: cuda-13.0
             cuda_home: /usr/local/cuda-13.0
@@ -89,7 +105,11 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - name: cuda13
+          - name: ampere
+            fedora_version: "43"
+          - name: ada
+            fedora_version: "43"
+          - name: blackwell
             fedora_version: "43"
     steps:
       - uses: actions/checkout@v4

.gitea/workflows/build-release.yml

@@ -18,7 +18,23 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - name: cuda13
+          - name: ampere
+            fedora_version: "43"
+            runner: cuda-13.0
+            cuda_home: /usr/local/cuda-13.0
+            cargo_features: "cuda cudnn flash-attn nccl"
+            compute_caps: "86"
+            build_jobs: 12
+            nvcc_threads: 4
+          - name: ada
+            fedora_version: "43"
+            runner: cuda-13.0
+            cuda_home: /usr/local/cuda-13.0
+            cargo_features: "cuda cudnn flash-attn nccl"
+            compute_caps: "89"
+            build_jobs: 12
+            nvcc_threads: 4
+          - name: blackwell
             fedora_version: "43"
             runner: cuda-13.0
             cuda_home: /usr/local/cuda-13.0
@@ -76,7 +92,11 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - name: cuda13
+          - name: ampere
+            fedora_version: "43"
+          - name: ada
+            fedora_version: "43"
+          - name: blackwell
             fedora_version: "43"
     steps:
       - uses: actions/checkout@v4

.gitea/workflows/poll-upstream.yml

@@ -28,7 +28,7 @@ jobs:
         run: |
           version="${UPSTREAM_TAG#v}"
           needs_build=false
-          for target in "43:cuda13"; do
+          for target in "43:ampere" "43:ada" "43:blackwell"; do
             fedora_version="${target%%:*}"
             flavour="${target##*:}"
             base_url="https://rpm.lair.cafe/fedora/${fedora_version}/x86_64"
@@ -116,7 +116,7 @@ jobs:
         run: |
           prerelease="0.1.${UPSTREAM_DATE}git${UPSTREAM_SHORT_SHA}"
           needs_build=false
-          for target in "43:cuda13"; do
+          for target in "43:ampere" "43:ada" "43:blackwell"; do
             fedora_version="${target%%:*}"
             flavour="${target##*:}"
             base_url="https://rpm.lair.cafe/fedora/${fedora_version}/x86_64/unstable"