use tinyllama

Signed-off-by: Sertac Ozercan <[email protected]>

.github/workflows/pre-release.yaml

@@ -51,3 +51,11 @@ jobs:
         env:
           DIGEST: ${{ steps.build-and-push.outputs.digest }}
         run: cosign sign --yes "ghcr.io/sozercan/aikit@${DIGEST}"
+
+      - name: Verify image signature
+        env:
+          DIGEST: ${{ steps.build-and-push.outputs.digest }}
+        run: |
+          cosign verify ghcr.io/sozercan/${{ env.MODEL_NAME }}:${{ env.MODEL_SIZE }}${{ env.MODEL_RUNTIME }}@${DIGEST} \
+            --certificate-identity https://github.com/sozercan/aikit/.github/workflows/pre-release.yaml@refs/heads/main \
+            --certificate-oidc-issuer https://token.actions.githubusercontent.com | jq

.github/workflows/release.yaml

@@ -58,3 +58,11 @@ jobs:
         env:
           DIGEST: ${{ steps.build-and-push.outputs.digest }}
         run: cosign sign --yes "ghcr.io/sozercan/aikit@${DIGEST}"
+
+      - name: Verify image signature
+        env:
+          DIGEST: ${{ steps.build-and-push.outputs.digest }}
+        run: |
+          cosign verify ghcr.io/sozercan/${{ env.MODEL_NAME }}:${{ env.MODEL_SIZE }}${{ env.MODEL_RUNTIME }}@${DIGEST} \
+            --certificate-identity https://github.com/sozercan/aikit/.github/workflows/release.yaml@refs/heads/main \
+            --certificate-oidc-issuer https://token.actions.githubusercontent.com | jq

.github/workflows/test-docker.yaml

@@ -42,10 +42,19 @@ jobs:
       - uses: crazy-max/ghaction-github-runtime@v3
 
       - name: build aikit
-        run: docker buildx build . -t aikit:test --load --cache-from=type=gha,scope=aikit --cache-to=type=gha,scope=aikit,mode=max
+        run: |
+          docker buildx build . -t aikit:test \
+            --load --progress plain \
+            --cache-from=type=gha,scope=aikit \
+            --cache-to=type=gha,scope=aikit,mode=max
 
       - name: build test model
-        run: docker buildx build . -t testmodel:test -f test/aikitfile.yaml --load --cache-from=type=gha,scope=testmodel --cache-to=type=gha,scope=testmodel,mode=max
+        run: |
+          docker buildx build . -t testmodel:test \
+            -f test/aikitfile.yaml \
+            --load --progress plain \
+            --cache-from=type=gha,scope=testmodel \
+            --cache-to=type=gha,scope=testmodel,mode=max
 
       - name: list images
         run: docker images

.github/workflows/update-models.yaml

@@ -47,6 +47,7 @@ jobs:
 
       - name: Build and push
         uses: docker/build-push-action@v5
+        id: build-and-push
         with:
           push: true
           tags: ghcr.io/sozercan/${{ env.MODEL_NAME }}:${{ env.MODEL_SIZE }}${{ env.MODEL_RUNTIME }}
@@ -59,7 +60,7 @@ jobs:
           DIGEST: ${{ steps.build-and-push.outputs.digest }}
         run: cosign sign --yes "ghcr.io/sozercan/${{ env.MODEL_NAME }}:${{ env.MODEL_SIZE }}${{ env.MODEL_RUNTIME }}@${DIGEST}"
 
-      - name: verify image signature
+      - name: Verify image signature
         env:
           DIGEST: ${{ steps.build-and-push.outputs.digest }}
         run: |

test/aikitfile.yaml

@@ -14,3 +14,4 @@ config: |
       top_p: 0.7
       model: tinyllama-1.1b-chat-v0.3.Q4_K_M.gguf
     context_size: 1000
+    mmap: false

test/bats/test.bats

@@ -5,9 +5,9 @@ load helpers
 WAIT_TIME=120
 SLEEP_TIME=1
 
-@test "send request to llama-2-7b-chat" {
+@test "send request to tinyllama-1.1b-chat" {
     run curl --retry 20 --retry-all-errors http://localhost:8080/v1/chat/completions -H "Content-Type: application/json" -d '{
-     "model": "llama-2-7b-chat",
+     "model": "tinyllama-1.1b-chat",
      "messages": [{"role": "user", "content": "explain kubernetes in a sentence"}],
     }'
     assert_success