github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #6561] Inconsistent API Behavior #50640

New Issue
Open
opened 2026-04-28 16:42:07 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-28 16:42:07 -05:00
Owner
Copy Link

Originally created by @negaralizadeh on GitHub (Aug 29, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6561

What is the issue?

I'm calling the generate API as follows:

url = 'http://localhost:11434/api/generate'
data = {
  "model": model_name,
  "stream": False,
  "options": {
      "temperature": 0.2,
      "top_p": 0.8,
      "seed": 42,
      "num_predict": 300,
  },
  "system": set_role()
}

response = requests.post(url, json=data).json()

Although I set the stream flag to false, sometimes I don't get the whole response in one package (done = False in the first received package).
The other problem is that sometimes, even when the package is final (done=true), I don't see all the expected additional information there, for example, prompt_eval_count is missing.

This problem also persists with the Python library. I’ve carefully checked the documentation, and I believe it might be some sort of bug.

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.3.5

Originally created by @negaralizadeh on GitHub (Aug 29, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6561 ### What is the issue? I'm calling the generate API as follows: ``` url = 'http://localhost:11434/api/generate' data = { "model": model_name, "stream": False, "options": { "temperature": 0.2, "top_p": 0.8, "seed": 42, "num_predict": 300, }, "system": set_role() } response = requests.post(url, json=data).json() ``` Although I set the stream flag to false, sometimes I don't get the whole response in one package (done = False in the first received package). The other problem is that sometimes, even when the package is final (done=true), I don't see all the expected additional information there, for example, prompt_eval_count is missing. This problem also persists with the Python library. I’ve carefully checked the documentation, and I believe it might be some sort of bug. ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.3.5
GiteaMirror added the bug label 2026-04-28 16:42:07 -05:00
GiteaMirror commented 2026-04-28 16:42:09 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 29, 2024):

What is the system field supposed to be? What model are you seeing issues with?

<!-- gh-comment-id:2319440746 --> @rick-github commented on GitHub (Aug 29, 2024): What is the `system` field supposed to be? What model are you seeing issues with?
GiteaMirror commented 2026-04-28 16:42:10 -05:00
Author
Owner
Copy Link

@negaralizadeh commented on GitHub (Aug 30, 2024):

What is the system field supposed to be? What model are you seeing issues with?

According to the documentation, the system field overwrites the system message in the Modelfile.

I mostly encounter this issue with "codellama:7b-instruct-q5_K_M" but it has also been observed with "granite-code:8b-instruct-q4_K_M" and "llama3:8b-instruct-q5_K_M"

To be more specific, I'm doing experiments with the HumanEval (Python) dataset.

data = {
"model": codellama:7b-instruct-q5_K_M
"prompt": "Generate only the Check Function in Python with at most 10 assertions according to the Instruction: " + prompt
"stream": False,
    "options": {
        "temperature": 0.2,
        "top_p": 0.8,
        "seed": 42,
        "num_predict": 300,
    },
"system": 'You are an exceptionally intelligent AI assistant that consistently delivers accurate and reliable Python test cases for a function. Without any explanations in natural language.'
}
response = requests.post(url, json=data).json()

And this is what I got as a response for task Python/84:
{'model': 'codellama:7b-instruct-q5_K_M', 'created_at': '2024-08-30T08:24:41.524783507Z', 'response': '```\ndef check(solve):\n assert solve(1000) == "1"\n assert solve(150) == "110"\n assert solve(147) == "1100"\n assert solve(123456789) == "1001001001001001001001001001001"\n assert solve(1000000000) == "1111111111111111111111111111111', 'done': False}

However it's flaky, I cannot always reproduce this bug.

<!-- gh-comment-id:2320505742 --> @negaralizadeh commented on GitHub (Aug 30, 2024): > What is the `system` field supposed to be? What model are you seeing issues with? According to the [documentation](https://github.com/ollama/ollama/blob/56346ccfa3e51eec51fc26ae8e91fc88cb74a9b8/docs/api.md?plain=1#L50), the system field overwrites the system message in the Modelfile. I mostly encounter this issue with "codellama:7b-instruct-q5_K_M" but it has also been observed with "granite-code:8b-instruct-q4_K_M" and "llama3:8b-instruct-q5_K_M" To be more specific, I'm doing experiments with the HumanEval (Python) dataset. ``` data = { "model": codellama:7b-instruct-q5_K_M "prompt": "Generate only the Check Function in Python with at most 10 assertions according to the Instruction: " + prompt "stream": False, "options": { "temperature": 0.2, "top_p": 0.8, "seed": 42, "num_predict": 300, }, "system": 'You are an exceptionally intelligent AI assistant that consistently delivers accurate and reliable Python test cases for a function. Without any explanations in natural language.' } response = requests.post(url, json=data).json() ``` And this is what I got as a response for task Python/84: {'model': 'codellama:7b-instruct-q5_K_M', 'created_at': '2024-08-30T08:24:41.524783507Z', 'response': '```\ndef check(solve):\n assert solve(1000) == "1"\n assert solve(150) == "110"\n assert solve(147) == "1100"\n assert solve(123456789) == "1001001001001001001001001001001"\n assert solve(1000000000) == "1111111111111111111111111111111', 'done': False} However it's flaky, I cannot always reproduce this bug.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#50640
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?