[Bugfix][Habana_main] fix guided_decode HPU failing issue

[xuechendi]

FILL IN THE PR DESCRIPTION HERE

FIX ##198

After this change, we can see tool_calls can be returned successfully

Compiling FSM index for all state transitions: 100%|████████████████████████████████████████████████████████████████████████| 55/55 [00:01<00:00, 32.86it/s]INFO 09-04 02:15:34 metrics.py:406] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.
INFO 09-04 02:15:34 logger.py:36] Received request chat-0fd03b03ae05473488d9bce566401d91: prompt: "<|im_start|>user\nWhat's the weather like in Boston today?<|im_end|>\n<|im_start|>assistant\n", params: SamplingParams(n=1, best_of=1, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalty=1.0, temperature=0.7, top_p=1.0, top_k=-1, min_p=0.0, seed=None, use_beam_search=False, length_penalty=1.0, early_stopping=False, stop=[], stop_token_ids=[], include_stop_str_in_output=False, ignore_eos=False, max_tokens=1000, min_tokens=0, logprobs=None, prompt_logprobs=None, skip_special_tokens=True, spaces_between_special_tokens=True, truncate_prompt_tokens=None), prompt_token_ids: [27, 91, 318, 5011, 91, 29, 882, 198, 3923, 596, 279, 9282, 1093, 304, 10406, 3432, 76514, 91, 318, 6345, 91, 397, 27, 91, 318, 5011, 91, 29, 78191, 198], lora_request: None, prompt_adapter_request: None.
INFO 09-04 02:15:34 async_llm_engine.py:173] Added request chat-0fd03b03ae05473488d9bce566401d91.
INFO 09-04 02:15:36 async_llm_engine.py:140] Finished request chat-0fd03b03ae05473488d9bce566401d91.
INFO:     127.0.0.1:40452 - "POST /v1/chat/completions HTTP/1.1" 200 OK
Message: ChatCompletionMessage(content='', refusal=None, role='assistant', function_call=None, tool_calls=[ChatCompletionMessageToolCall(id='chatcmpl-tool-af3eac9372144f959ed0df7e16cf5da4', function=Function(arguments='{ "location": "Boston, MA", "unit": "fahrenheit" }', name='get_current_weather'), type='function')])

BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE

---

PR Checklist (Click to Expand)

Thank you for your contribution to vLLM! Before submitting the pull request, please ensure the PR meets the following criteria. This helps vLLM maintain the code quality and improve the efficiency of the review process.

PR Title and Classification

Only specific types of PRs will be reviewed. The PR title is prefixed appropriately to indicate the type of change. Please use one of the following:

• [Bugfix] for bug fixes.
• [CI/Build] for build or continuous integration improvements.
• [Doc] for documentation fixes and improvements.
• [Model] for adding a new model or improving an existing model. Model name should appear in the title.
• [Frontend] For changes on the vLLM frontend (e.g., OpenAI API server, LLM class, etc.)
• [Kernel] for changes affecting CUDA kernels or other compute kernels.
• [Core] for changes in the core vLLM logic (e.g., LLMEngine, AsyncLLMEngine, Scheduler, etc.)
• [Hardware][Vendor] for hardware-specific changes. Vendor name should appear in the prefix (e.g., [Hardware][AMD]).
• [Misc] for PRs that do not fit the above categories. Please use this sparingly.

Note: If the PR spans more than one category, please include all relevant prefixes.

Code Quality

The PR need to meet the following code quality standards:

• We adhere to Google Python style guide and Google C++ style guide.
• Pass all linter checks. Please use format.sh to format your code.
• The code need to be well-documented to ensure future contributors can easily understand the code.
• Include sufficient tests to ensure the project to stay correct and robust. This includes both unit tests and integration tests.
• Please add documentation to docs/source/ if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.

Notes for Large Changes

Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with rfc-required and might not go through the PR.

What to Expect for the Reviews

The goal of the vLLM team is to be a transparent reviewing machine. We would like to make the review process transparent and efficient and make sure no contributor feel confused or frustrated. However, the vLLM team is small, so we need to prioritize some PRs over others. Here is what you can expect from the review process:

• After the PR is submitted, the PR will be assigned to a reviewer. Every reviewer will pick up the PRs based on their expertise and availability.
• After the PR is assigned, the reviewer will provide status update every 2-3 days. If the PR is not reviewed within 7 days, please feel free to ping the reviewer or the vLLM team.
• After the review, the reviewer will put an action-required label on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR.
• Please respond to all comments within a reasonable time frame. If a comment isn't clear or you disagree with a suggestion, feel free to ask for clarification or discuss the suggestion.

Thank You

Finally, thank you for taking the time to read these guidelines and for your interest in contributing to vLLM. Your contributions make vLLM a great tool for everyone!

[xuechendi]

@kzawora-intel , please help to review this PR. This is a bug fix for issue #198

• With this PR, we will be able to support Agent Tool_Call for llama3.1 Instruct models.
• The original in-place tensor.add_ is not well supported in HPU. One easy fix is to change it to 'add'.

[xuechendi]

returned tool_calls:

tool_calls=[ChatCompletionMessageToolCall(id='chatcmpl-tool-af3eac9372144f959ed0df7e16cf5da4', function=Function(arguments='{ "location": "Boston, MA", "unit": "fahrenheit" }', name='get_current_weather'), type='function')])

test script as below

import asyncio, os, sys
import openai
from pathlib import Path


VLLM_PATH = os.path.join(Path(__file__).parent.parent, "vllm")
VLLM_PATH = os.path.join(Path(__file__).parent.parent, "vllm", "tests")
sys.path.append(VLLM_PATH)

from utils import RemoteOpenAIServer

"""Path to root of the vLLM repository."""

async def test_named_tool_use(client: openai.AsyncOpenAI, model_name, input, tools, tool_choice):
    # non-streaming
    chat_completion = await client.chat.completions.create(
        model=model_name,
        messages=input,
        max_tokens=1000,
        tools=tools,
        tool_choice=tool_choice)
    message = chat_completion.choices[0].message
    print(f"Message: {message}")

def test_multiple_sampling_params():
    
    model_name = "meta-llama/Meta-Llama-3.1-8B"
    args = [
        "--max-model-len", "8192",
        #"--enforce-eager",
    ]
    with RemoteOpenAIServer(model_name, args) as remote_server:
        
        client_inst = remote_server.get_async_client()
    
        input = [{"role": "user", "content": "What's the weather like in Boston today?"}]
        tools = [{
            "type": "function",
            "function": {
                "name": "get_current_weather",
                "description": "Get the current weather in a given location",
                "parameters": {
                    "type": "object",
                    "properties": {
                    "location": {
                        "type": "string",
                        "description": "The city and state, e.g. San Francisco, CA",
                    },
                    "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
                    },
                    "required": ["location"],
                },
            }
        }]
        tool_choice = {'function': {'name': 'get_current_weather'}}
        
        asyncio.run(test_named_tool_use(client_inst, model_name, input, tools, tool_choice=tool_choice))
        
if __name__ == "__main__":
    print("VLLM_PATH is ", VLLM_PATH)
    test_multiple_sampling_params()

[tae-su-kim]

Duplicate of PR #153 and PR #226 - please review those PRs as well.

[xuechendi]

Duplicate of PR #153 and PR #226 - please review those PRs as well.

Thanks, @tae-su-kim , I didn't notice your PR, it looks great. Either 226 or this PR works for me, I hope to make vllm-fork to support Agent tool_calls ASAP so it can be utilized in OPEA.

[tae-su-kim]

@xuechendi Great! We recently observed unexpected throughput degradation with guided_decode and submitted fix for it (commit 6d57c18 and #226 (comment)). If you are interested, please check it out. It would be really helpful if you could cross-check latency improvement with the test cases for tool_call and llama-3.1-8B.

[michalkuligowski]

LGTM

[tae-su-kim]

@michalkuligowski This feature seriously needs commit 6d57c18. Without it, e2e throughput degrades a lot. Please review #226 also.

[michalkuligowski]

@tae-su-kim hi, I reviewed the #226, it has unused imported function and ruff code analysis fails on that

[jikunshang]

just curious, what's the difference here?

[xuechendi]

should be no difference in result, however, 'add_' is executing in_place, while 'add' will return a tensor.
From my test, using 'add_' leads to "RuntimeError: synNodeCreateWithId failed for node: strided_insert with synStatus 1 [Invalid argument]. ", replacing with 'add' fixed above issue