Skip to content

Commit 7e99635

Browse files
committed
Debug
1 parent 1965a8e commit 7e99635

File tree

1 file changed

+15
-3
lines changed

1 file changed

+15
-3
lines changed

.github/workflows/additional_demo_notebook_tests.yaml

+15-3
Original file line numberDiff line numberDiff line change
@@ -85,18 +85,23 @@ jobs:
8585
# Remove login/logout cells, as KinD doesn't support authentication using token
8686
jq -r 'del(.cells[] | select(.source[] | contains("Create authentication object for user permissions")))' hf_interactive.ipynb > hf_interactive.ipynb.tmp && mv hf_interactive.ipynb.tmp hf_interactive.ipynb
8787
jq -r 'del(.cells[] | select(.source[] | contains("auth.logout()")))' hf_interactive.ipynb > hf_interactive.ipynb.tmp && mv hf_interactive.ipynb.tmp hf_interactive.ipynb
88+
# Rewrite cluster_uri() to local_client_url() to retrieve client URL available out of cluster, as the test is executed outside of cluster
89+
sed -i "s/cluster_uri()/local_client_url()/g" hf_interactive.ipynb
8890
# Replace async logs with waiting for job to finish, async logs don't work properly in papermill
8991
JOB_WAIT=$(jq -r '.' ${GITHUB_WORKSPACE}/.github/resources/wait_for_job_cell.json)
9092
jq --argjson job_wait "$JOB_WAIT" -r '(.cells[] | select(.source[] | contains("async for lines in client.tail_job_logs"))) |= $job_wait' hf_interactive.ipynb > hf_interactive.ipynb.tmp && mv hf_interactive.ipynb.tmp hf_interactive.ipynb
9193
# Set explicit namespace as SDK need it (currently) to resolve local queues
9294
sed -i "s/worker_cpu_requests=8,/worker_cpu_requests=1, namespace='default',/" hf_interactive.ipynb
9395
# Change cluster parameters (need to decrease)
9496
sed -i "s/{'nvidia.com\/gpu':1}/{'nvidia.com\/gpu':0}/g" hf_interactive.ipynb
97+
sed -i "s/worker_cpu_requests=8,/worker_cpu_requests=1,/" hf_interactive.ipynb
9598
sed -i "s/worker_cpu_limits=8,/worker_cpu_limits=1,/" hf_interactive.ipynb
9699
sed -i "s/worker_memory_requests=16,/worker_memory_requests=4,/" hf_interactive.ipynb
97100
sed -i "s/worker_memory_limits=8,/worker_memory_limits=4,/" hf_interactive.ipynb
98101
# Run notebook
99102
poetry run papermill hf_interactive.ipynb hf_interactive_out.ipynb --log-output --execution-timeout 1200
103+
env:
104+
GRPC_DNS_RESOLVER: "native"
100105
working-directory: demo-notebooks/additional-demos
101106

102107
- name: Print CodeFlare operator logs
@@ -135,7 +140,6 @@ jobs:
135140
136141
verify-local_interactive:
137142
# if: ${{ github.event.label.name == 'test-additional-notebooks' }}
138-
# runs-on: ubuntu-20.04-4core
139143
runs-on: ubuntu-20.04-4core
140144

141145
steps:
@@ -205,13 +209,17 @@ jobs:
205209
# Remove login/logout cells, as KinD doesn't support authentication using token
206210
jq -r 'del(.cells[] | select(.source[] | contains("Create authentication object and log in to desired user account")))' local_interactive.ipynb > local_interactive.ipynb.tmp && mv local_interactive.ipynb.tmp local_interactive.ipynb
207211
jq -r 'del(.cells[] | select(.source[] | contains("auth.logout()")))' local_interactive.ipynb > local_interactive.ipynb.tmp && mv local_interactive.ipynb.tmp local_interactive.ipynb
212+
# Rewrite cluster_uri() to local_client_url() to retrieve client URL available out of cluster, as the test is executed outside of cluster
213+
sed -i "s/cluster_uri()/local_client_url()/g" local_interactive.ipynb
208214
# Replace async logs with waiting for job to finish, async logs don't work properly in papermill
209215
JOB_WAIT=$(jq -r '.' ${GITHUB_WORKSPACE}/.github/resources/wait_for_job_cell.json)
210216
jq --argjson job_wait "$JOB_WAIT" -r '(.cells[] | select(.source[] | contains("async for lines in client.tail_job_logs"))) |= $job_wait' local_interactive.ipynb > local_interactive.ipynb.tmp && mv local_interactive.ipynb.tmp local_interactive.ipynb
211217
# Set explicit namespace as SDK need it (currently) to resolve local queues
212-
sed -i "s/worker_cpu_requests=1,/worker_cpu_requests=1, namespace='default',/" local_interactive.ipynb
218+
sed -i "s/worker_cpu_requests=1,/worker_cpu_requests='250m', namespace='default',/" local_interactive.ipynb
213219
# Run notebook
214-
poetry run papermill local_interactive.ipynb hf_interactive_out.ipynb --log-output --execution-timeout 1200
220+
poetry run papermill local_interactive.ipynb local_interactive_out.ipynb --log-output --execution-timeout 1200
221+
env:
222+
GRPC_DNS_RESOLVER: "native"
215223
working-directory: demo-notebooks/additional-demos
216224

217225
- name: Print CodeFlare operator logs
@@ -319,13 +327,17 @@ jobs:
319327
# Remove login/logout cells, as KinD doesn't support authentication using token
320328
jq -r 'del(.cells[] | select(.source[] | contains("Create authentication object for user permissions")))' ray_job_client.ipynb > ray_job_client.ipynb.tmp && mv ray_job_client.ipynb.tmp ray_job_client.ipynb
321329
jq -r 'del(.cells[] | select(.source[] | contains("auth.logout()")))' ray_job_client.ipynb > ray_job_client.ipynb.tmp && mv ray_job_client.ipynb.tmp ray_job_client.ipynb
330+
# Rewrite cluster_uri() to local_client_url() to retrieve client URL available out of cluster, as the test is executed outside of cluster
331+
sed -i "s/cluster_uri()/local_client_url()/g" ray_job_client.ipynb
322332
# Replace async logs with waiting for job to finish, async logs don't work properly in papermill
323333
JOB_WAIT=$(jq -r '.' ${GITHUB_WORKSPACE}/.github/resources/wait_for_job_cell.json)
324334
jq --argjson job_wait "$JOB_WAIT" -r '(.cells[] | select(.source[] | contains("async for lines in client.tail_job_logs"))) |= $job_wait' ray_job_client.ipynb > ray_job_client.ipynb.tmp && mv ray_job_client.ipynb.tmp ray_job_client.ipynb
325335
# Set explicit namespace as SDK need it (currently) to resolve local queues
326336
sed -i "s/worker_cpu_requests=1,/worker_cpu_requests=1, namespace='default',/" ray_job_client.ipynb
327337
# Run notebook
328338
poetry run papermill ray_job_client.ipynb hf_interactive_out.ipynb --log-output --execution-timeout 1200
339+
env:
340+
GRPC_DNS_RESOLVER: "native"
329341
working-directory: demo-notebooks/additional-demos
330342

331343
- name: Print CodeFlare operator logs

0 commit comments

Comments
 (0)