github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #6355] LLama 3.1 Tools do not work properly #66025

New Issue
Closed
opened 2026-05-03 23:39:24 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-03 23:39:24 -05:00
Owner
Copy Link

Originally created by @nomisto on GitHub (Aug 14, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6355

What is the issue?

I can't get the tooling function of llama 3.1 to work properly.

messages = [
 {'role': 'system', 'content': 'You are a helpful assistant with tool calling capabilities. When you receive a tool call response, use the output to format an answer to the orginal use question.'},
 {'role': 'user', 'content': 'What is the weather in Toronto?'},
 {'role': 'assistant',
  'content': '',
  'tool_calls': [{'function': {'name': 'get_current_weather',
     'arguments': {'city': 'Toronto'}},
    'id': 'call_6duDxk',
    'type': 'function'}]},
 {'role': 'ipython',
  'tool_call_id': 'call_6duDxk',
  'name': 'get_current_weather',
  'content': '{"city": "Toronto", "weather": "sunny"}'}]

Using ollama, I get the following (hallucinated) output:

response = ollama.chat(
    model='llama3.1:8b-instruct-fp16',
    messages=messages,
    tools=tools
)
response

{'model': 'llama3.1:8b-instruct-fp16',
 'created_at': '2024-08-14T09:12:15.708397447Z',
 'message': {'role': 'assistant',
  'content': ' \n\nThe current weather in Toronto is mostly cloudy with a temperature of 22°C (72°F) and a gentle breeze of 15 km/h (9 mph).'},
 'done_reason': 'stop',
 'done': True,
 'total_duration': 1352319726,
 'load_duration': 70634597,
 'prompt_eval_count': 101,
 'prompt_eval_duration': 44061000,
 'eval_count': 34,
 'eval_duration': 1013216000}

Using plain transformers I get the correct answer

import transformers
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "meta-llama/Meta-Llama-3.1-8B-Instruct"
device = "cuda:5"
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16)
model.to(device)
tokenizer = AutoTokenizer.from_pretrained(model_id)

inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=True, return_tensors="pt").to(device)
outputs = model.generate(inputs, do_sample=False, max_new_tokens=256)
response = tokenizer.batch_decode(outputs[:, inputs.shape[1]:], skip_special_tokens=True)[0]
response

'The current weather in Toronto is sunny.'

Maybe this is related to #6129?

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.3.5

Originally created by @nomisto on GitHub (Aug 14, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6355 ### What is the issue? I can't get the tooling function of llama 3.1 to work properly. ```python messages = [ {'role': 'system', 'content': 'You are a helpful assistant with tool calling capabilities. When you receive a tool call response, use the output to format an answer to the orginal use question.'}, {'role': 'user', 'content': 'What is the weather in Toronto?'}, {'role': 'assistant', 'content': '', 'tool_calls': [{'function': {'name': 'get_current_weather', 'arguments': {'city': 'Toronto'}}, 'id': 'call_6duDxk', 'type': 'function'}]}, {'role': 'ipython', 'tool_call_id': 'call_6duDxk', 'name': 'get_current_weather', 'content': '{"city": "Toronto", "weather": "sunny"}'}] ``` Using ollama, I get the following (hallucinated) output: ```python response = ollama.chat( model='llama3.1:8b-instruct-fp16', messages=messages, tools=tools ) response ``` ```python {'model': 'llama3.1:8b-instruct-fp16', 'created_at': '2024-08-14T09:12:15.708397447Z', 'message': {'role': 'assistant', 'content': ' \n\nThe current weather in Toronto is mostly cloudy with a temperature of 22°C (72°F) and a gentle breeze of 15 km/h (9 mph).'}, 'done_reason': 'stop', 'done': True, 'total_duration': 1352319726, 'load_duration': 70634597, 'prompt_eval_count': 101, 'prompt_eval_duration': 44061000, 'eval_count': 34, 'eval_duration': 1013216000} ``` Using *plain* transformers I get the correct answer ```python import transformers import torch from transformers import AutoModelForCausalLM, AutoTokenizer model_id = "meta-llama/Meta-Llama-3.1-8B-Instruct" device = "cuda:5" model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16) model.to(device) tokenizer = AutoTokenizer.from_pretrained(model_id) inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=True, return_tensors="pt").to(device) outputs = model.generate(inputs, do_sample=False, max_new_tokens=256) response = tokenizer.batch_decode(outputs[:, inputs.shape[1]:], skip_special_tokens=True)[0] response ``` ```text 'The current weather in Toronto is sunny.' ``` Maybe this is related to #6129? ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.3.5
GiteaMirror added the bug label 2026-05-03 23:39:24 -05:00
GiteaMirror commented 2026-05-03 23:39:29 -05:00
Author
Owner
Copy Link

@nomisto commented on GitHub (Aug 14, 2024):

Using "role": "tool" instead of "role": "ipython" resolved the issue. I don't know if this is intended or not.

Closing this since template of llama 3.1 hints that this is intended:
{{- else if eq .Role "tool" }}<|start_header_id|>ipython<|end_header_id|>

<!-- gh-comment-id:2288281257 --> @nomisto commented on GitHub (Aug 14, 2024): Using "role": "tool" instead of "role": "ipython" resolved the issue. I don't know if this is intended or not. Closing this since template of llama 3.1 hints that this is intended: `{{- else if eq .Role "tool" }}<|start_header_id|>ipython<|end_header_id|>`
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#66025
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?