Got Internal Server Error when CURL DeepSeek-R1 on Azure AI

Question

Got Internal Server Error when CURL DeepSeek-R1 on Azure AI

Qumeng Sun 5

$       curl "https://ai-<XXX>.openai.azure.com/openai/deployments/DeepSeek-R1/chat/completions?api-version=2024-02-15-preview" \
        -H "Content-Type: application/json" \
        -H "api-key: <XXX>" \
        -d '{
          "messages": [{"role":"system","content":"<PROMPT>"}],
          "max_tokens": 800,
          "temperature": 0.7,
          "frequency_penalty": 0,
          "presence_penalty": 0,
          "best_of": 1,
          "top_p": 0.95,
          "stop": null,
          "model": "DeepSeek-R1"
        }'

{"error":{"code":"InternalServerError","message":"Backend returned unexpected response. Please contact Microsoft for help."}}

Vikram Singh 2,585 Reputation points Microsoft Employee Moderator

2025-02-10T05:57:58.9566667+00:00
Hi Qumeng Sun,

Thank you for posting your question on Microsoft Q&A. I apologize for any inconvenience this issue may have caused.

The "InternalServerError: Backend returned unexpected response" error can be challenging to diagnose. Here are some steps and quick workarounds that might help resolve this issue

Verify API Endpoint and Deployment

Endpoint Accuracy: Ensure that the API endpoint URL is correctly formatted and matches your deployment details.

Deployment Status: Confirm that the DeepSeek-R1 deployment is active and correctly configured in your Azure OpenAI resource.

Check API Key and Permissions

API Key Validity: Ensure that the API key used is valid and has the necessary permissions to access the deployment.

Subscription Limits: Verify that your subscription has not exceeded its usage limits, which could cause such errors.

Review Request Payload

JSON Structure: Ensure that the JSON payload is correctly structured and all required fields are present.

Parameter Values: Double-check the values of parameters like max_tokens, temperature, and others to ensure they are within acceptable ranges.

Monitor Service Health

Azure Service Status: Check the Azure Status Page to see if there are any ongoing issues with the OpenAI service in your region.

Regional Availability: Ensure that the service is available in the region where your resource is deployed.

Implement Retry Logic: Transient Errors: Sometimes, 500 errors can be transient. Implementing a retry mechanism with exponential backoff can help mitigate temporary issues.

If the issue persists after performing the above steps, capture the full error response, including any error codes or messages, to assist in troubleshooting. If available, note the correlation ID from the error response to provide for further investigation.

Thanks
Michal Simonfy 80 Reputation points

2025-02-10T11:41:23.1533333+00:00
@Qumeng Sun we were able to get R1 running by changing the base URL to

https://<XXX>.openai.azure.com/models/chat/completions?api-version=2024-05-01-preview

Hope this helps.
Vikram Singh 2,585 Reputation points Microsoft Employee Moderator

2025-02-12T12:09:09.35+00:00

Hi Qumeng Sun,

Greetings.

Just following up to check if my suggestion helped or let me know if you are facing the issue. Please do not forget to "Accept the answer” and “up-vote” wherever the information provided helps you, this can be beneficial to other community members.

Thank you
Vikram Singh 2,585 Reputation points Microsoft Employee Moderator

2025-02-17T04:50:48.6133333+00:00

Hi Qumeng Sun,

Greetings.

If the response helped, please do click Accept Answer and Yes for was this answer helpful.

Doing so would help other community members with similar issue identify the solution. I highly appreciate your contribution to the community.

Thank You.
Vikram Singh 2,585 Reputation points Microsoft Employee Moderator

2025-02-20T04:29:18.1366667+00:00

Hi Qumeng Sun,

Greetings.

Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.

Doing so would help other community members with similar issue identify the solution. I highly appreciate your contribution to the community.

Thank You.

Your answer

Vikram Singh 2,585 Reputation points Microsoft Employee Moderator

2025-02-10T05:57:58.9566667+00:00

Hi Qumeng Sun,

Thank you for posting your question on Microsoft Q&A. I apologize for any inconvenience this issue may have caused.

The "InternalServerError: Backend returned unexpected response" error can be challenging to diagnose. Here are some steps and quick workarounds that might help resolve this issue

Verify API Endpoint and Deployment

Endpoint Accuracy: Ensure that the API endpoint URL is correctly formatted and matches your deployment details.

Deployment Status: Confirm that the DeepSeek-R1 deployment is active and correctly configured in your Azure OpenAI resource.

Check API Key and Permissions

API Key Validity: Ensure that the API key used is valid and has the necessary permissions to access the deployment.

Subscription Limits: Verify that your subscription has not exceeded its usage limits, which could cause such errors.

Review Request Payload

JSON Structure: Ensure that the JSON payload is correctly structured and all required fields are present.

Parameter Values: Double-check the values of parameters like max_tokens, temperature, and others to ensure they are within acceptable ranges.

Monitor Service Health

Azure Service Status: Check the Azure Status Page to see if there are any ongoing issues with the OpenAI service in your region.

Regional Availability: Ensure that the service is available in the region where your resource is deployed.

Implement Retry Logic: Transient Errors: Sometimes, 500 errors can be transient. Implementing a retry mechanism with exponential backoff can help mitigate temporary issues.

If the issue persists after performing the above steps, capture the full error response, including any error codes or messages, to assist in troubleshooting. If available, note the correlation ID from the error response to provide for further investigation.

Thanks
Michal Simonfy 80 Reputation points

2025-02-10T11:41:23.1533333+00:00

@Qumeng Sun we were able to get R1 running by changing the base URL to

https://<XXX>.openai.azure.com/models/chat/completions?api-version=2024-05-01-preview

Hope this helps.
Vikram Singh 2,585 Reputation points Microsoft Employee Moderator

2025-02-12T12:09:09.35+00:00

Hi Qumeng Sun,

Greetings.

Just following up to check if my suggestion helped or let me know if you are facing the issue. Please do not forget to "Accept the answer” and “up-vote” wherever the information provided helps you, this can be beneficial to other community members.

Thank you
Vikram Singh 2,585 Reputation points Microsoft Employee Moderator

2025-02-17T04:50:48.6133333+00:00

Hi Qumeng Sun,

Greetings.

If the response helped, please do click Accept Answer and Yes for was this answer helpful.

Doing so would help other community members with similar issue identify the solution. I highly appreciate your contribution to the community.

Thank You.
Vikram Singh 2,585 Reputation points Microsoft Employee Moderator

2025-02-20T04:29:18.1366667+00:00

Hi Qumeng Sun,

Greetings.

Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.

Doing so would help other community members with similar issue identify the solution. I highly appreciate your contribution to the community.

Thank You.

Share via

Got Internal Server Error when CURL DeepSeek-R1 on Azure AI

Your answer