fix: revert skips and prefer updated ci token for tests

.github/workflows/build.yaml

@@ -141,7 +141,7 @@ jobs:
         run: |
           export DOCKER_VOLUME=/mnt/cache
           export DOCKER_IMAGE=registry.internal.huggingface.tech/api-inference/community/text-generation-inference:sha-${{ env.GITHUB_SHA_SHORT }}
-          export HUGGING_FACE_HUB_TOKEN=${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+          export HUGGING_FACE_HUB_TOKEN=${{ secrets.HF_TOKEN }}
           pytest -s -vv integration-tests
       - name: Tailscale Wait
         if: ${{ failure() || runner.debug == '1' }}

.github/workflows/tests.yaml

@@ -72,7 +72,7 @@ jobs:
       - name: Run server tests
         run: |
           pip install pytest
-          export HUGGING_FACE_HUB_TOKEN=${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+          export HUGGING_FACE_HUB_TOKEN=${{ secrets.HF_TOKEN }}
           pytest -s -vv server/tests
       - name: Pre-commit checks
         run: |

integration-tests/models/test_flash_llama.py

@@ -1,8 +1,5 @@
 import pytest
 
-# TODO: avoid skipping module when CI permissions are fixed
-pytest.skip(allow_module_level=True)
-
 
 @pytest.fixture(scope="module")
 def flash_llama_handle(launcher):

integration-tests/models/test_flash_llama_gptq.py

@@ -1,8 +1,5 @@
 import pytest
 
-# TODO: avoid skipping module when CI permissions are fixed
-pytest.skip(allow_module_level=True)
-
 
 @pytest.fixture(scope="module")
 def flash_llama_gptq_handle(launcher):

server/tests/models/test_model.py

@@ -5,9 +5,6 @@ from transformers import AutoTokenizer
 
 from text_generation_server.models import Model
 
-# TODO: avoid skipping module when CI permissions are fixed
-pytest.skip(allow_module_level=True)
-
 
 def get_test_model():
     class TestModel(Model):