github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-17 05:36:16 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #4724] empty response #80659

New Issue
Closed
opened 2026-05-09 09:16:37 -05:00 by GiteaMirror · 7 comments
GiteaMirror commented 2026-05-09 09:16:37 -05:00
Owner
Copy Link

Originally created by @themw123 on GitHub (May 30, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/4724

What is the issue?

I am getting an empty response string with llama3:8b.
Different models like mistral-instruct are working fine.

Setup:

  • windows 11
  • newest ollama version
  • llama3:8b(latest)

When context gets to high (approximately after exchanging 20 question/answer pairs) by appending the history with every request, i get an empty response from time to time like in 30% of responses.

Here is my code:

self.messages = [{"role": "system", "content": system_role}]
self.messages.append({"role": "user", "content": my_input_text})

url = "http://localhost:11434/api/chat"
data = {
    "model": "llama3:8b",
    "messages": self.messages,
    "stream": True
}
response = requests.post(url, json=data)   

for chunk in response.iter_lines():
    if chunk:
        chunk = chunk.decode('utf-8')
        chunk = json.loads(chunk)
        content = chunk['message']['content']
        print(content, end="", flush=True)
        yield content

OS

Windows

GPU

Nvidia

CPU

AMD

Ollama version

0.1.39

Originally created by @themw123 on GitHub (May 30, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/4724 ### What is the issue? I am getting an empty response string with llama3:8b. Different models like mistral-instruct are working fine. Setup: - windows 11 - newest ollama version - llama3:8b(latest) When context gets to high (approximately after exchanging 20 question/answer pairs) by appending the history with every request, i get an empty response from time to time like in 30% of responses. Here is my code: ``` self.messages = [{"role": "system", "content": system_role}] self.messages.append({"role": "user", "content": my_input_text}) url = "http://localhost:11434/api/chat" data = { "model": "llama3:8b", "messages": self.messages, "stream": True } response = requests.post(url, json=data) for chunk in response.iter_lines(): if chunk: chunk = chunk.decode('utf-8') chunk = json.loads(chunk) content = chunk['message']['content'] print(content, end="", flush=True) yield content ``` ### OS Windows ### GPU Nvidia ### CPU AMD ### Ollama version 0.1.39
GiteaMirror added the bug label 2026-05-09 09:16:37 -05:00
GiteaMirror commented 2026-05-09 09:16:40 -05:00
Author
Owner
Copy Link

@dominikvalk commented on GitHub (Jun 5, 2024):

same for me.

<!-- gh-comment-id:2150666245 --> @dominikvalk commented on GitHub (Jun 5, 2024): same for me.
GiteaMirror commented 2026-05-09 09:16:40 -05:00
Author
Owner
Copy Link

@MoritzHinterleitner commented on GitHub (Jun 5, 2024):

same for me. llama 3 and 2 with several messages and instant empty response
Edit: it seems to only happen for me if I give the array of messages and a promt at the same time. Might not be the same issue as it works for me now

<!-- gh-comment-id:2150790974 --> @MoritzHinterleitner commented on GitHub (Jun 5, 2024): same for me. llama 3 and 2 with several messages and instant empty response Edit: it seems to only happen for me if I give the array of messages and a promt at the same time. Might not be the same issue as it works for me now
GiteaMirror commented 2026-05-09 09:16:40 -05:00
Author
Owner
Copy Link

@Qualzz commented on GitHub (Jun 30, 2024):

Same for me with gemma2, llama3.

What is weird is that using open web ui, long contexts works fine, but using the ollama.js chat function, > 512 token and the answer are empty with "model does not support sliding window, limiting generation" visible in the ollama logs

<!-- gh-comment-id:2198797693 --> @Qualzz commented on GitHub (Jun 30, 2024): Same for me with gemma2, llama3. What is weird is that using open web ui, long contexts works fine, but using the ollama.js chat function, > 512 token and the answer are empty with "`model does not support sliding window, limiting generation`" visible in the ollama logs
GiteaMirror commented 2026-05-09 09:16:40 -05:00
Author
Owner
Copy Link

@SheMelody commented on GitHub (Jul 3, 2024):

This seems to be an issue specific to Windows. I'm having the same exact problem. When using ollama on Linux it's perfectly fine, I literally had zero issues in there, same versions.

EDIT: I have instantly found a solution. Before running your scripts and program, make sure you do it from powershell and use the following commands before running them:

$OutputEncoding=[Text.Encoding]::UTF8
[Console]::OutputEncoding=[Text.Encoding]::UTF8

As I imagined, this was simply a problem with text encoding and was causing wrong interpretation through client and server. It's all fine now for me.

Edit 2: The issue at hand says "getting empty responses 30% of the times", which was exactly what was happening to me especially when it was generating fast. I don't know if it's the same exact specific issue, but this seems to have solved it for me.

<!-- gh-comment-id:2206326895 --> @SheMelody commented on GitHub (Jul 3, 2024): This seems to be an issue specific to Windows. I'm having the same exact problem. When using ollama on Linux it's perfectly fine, I literally had zero issues in there, same versions. EDIT: I have instantly found a solution. Before running your scripts and program, make sure you do it from powershell and use the following commands before running them: ``` $OutputEncoding=[Text.Encoding]::UTF8 [Console]::OutputEncoding=[Text.Encoding]::UTF8 ``` As I imagined, this was simply a problem with text encoding and was causing wrong interpretation through client and server. It's all fine now for me. Edit 2: The issue at hand says "getting empty responses 30% of the times", which was exactly what was happening to me especially when it was generating fast. I don't know if it's the same exact specific issue, but this seems to have solved it for me.
GiteaMirror commented 2026-05-09 09:16:41 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Sep 12, 2024):

I'm having a hard time reproducing this in Windows (and haven't seen it in either Linux or MacOS). I've been trying with Windows 11 and a 7900 XTX, but no luck.

Given the issue was opened in May, let's go ahead and close it, but if someone can repro I can reopen.

<!-- gh-comment-id:2347398366 --> @pdevine commented on GitHub (Sep 12, 2024): I'm having a hard time reproducing this in Windows (and haven't seen it in either Linux or MacOS). I've been trying with Windows 11 and a 7900 XTX, but no luck. Given the issue was opened in May, let's go ahead and close it, but if someone can repro I can reopen.
GiteaMirror commented 2026-05-09 09:16:41 -05:00
Author
Owner
Copy Link

@v3ss0n commented on GitHub (Sep 25, 2024):

Is this bug solved? i am still encountering. Why this is closed as completed? which PR fixes it?

<!-- gh-comment-id:2374643944 --> @v3ss0n commented on GitHub (Sep 25, 2024): Is this bug solved? i am still encountering. Why this is closed as completed? which PR fixes it?
GiteaMirror commented 2026-05-09 09:16:41 -05:00
Author
Owner
Copy Link

@codejq commented on GitHub (Jan 25, 2025):

For me, the issue was, that I set the temperature to 0.5 and I had to change it ti to 0.1 and I had to omit the format: JSON

<!-- gh-comment-id:2614038251 --> @codejq commented on GitHub (Jan 25, 2025): For me, the issue was, that I set the temperature to 0.5 and I had to change it ti to 0.1 and I had to omit the format: JSON
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#80659
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?