gorilla-openfunctions-v2-gguf

Gorilla-OpenFunctions-v2 GGUF Quantized Models

Gorilla-OpenFunctions-v2

💡 SoTA for open-source models. On-par with GPT-4.

🚀 Check out the Berkeley Function Calling Leaderboard
📣 Read more in our OpenFunctions v2 release blog and Berkeley Function Calling Leaderboard blog 🟢 Check out the original Gorilla OpenFunctions-v2 in gorilla-llm/gorilla-openfunctions-v2

Introduction

Gorilla OpenFunctions extends Large Language Model(LLM) Chat Completion feature to formulate executable APIs call given natural language instructions and API context. With OpenFunctions v2, we now support:

1. Multiple functions - choose betwen functions
2. Parallel functions - call the same function N time with different parameter values
3. Multiple & parallel - both of the above in a single chatcompletion call (one generation)
4. Relevance detection - when chatting, chat. When asked for function, returns a function
5. Python - supports string, number, boolean, list, tuple, dict parameter datatypes and Any for those not natively supported.
6. JAVA - support for byte, short, int, float, double, long, boolean, char, Array, ArrayList, Set, HashMap, Hashtable, Queue, Stack, and Any datatypes.
7. JavaScript - support for String, Number, Bigint, Boolean, dict (object), Array, Date, and Any datatypes.
8. REST - native REST support

We've quantized Gorilla-OpenFunctions-v2 based on llama.cpp as well as evaluated the quantized models on the Berkeley Function Call Leaderboard to benchmark their performance with the original model as well as other models.

Gorilla-OpenFunctions-v2 Quantized GGUF Models Evaluation

Here, we show some of the evaluation result summaries we have obtained from the evaluation.

We observe that the quantized models have a lower overall accuracy compared to the original model. Evaluation results for q4 or higher quantization methods are comparable, but q3 and q2 quantization methods have larger drop in overall accuracy.

---

How to use GGUF locally

To use GGUF locally, first download GGUF models locally.

One option you can use is to use huggingface-cli. To download huggingface-cli please follow tutorials in https://huggingface.co/docs/huggingface_hub/main/en/guides/cli.

Then, do command (also replace {QUANTIZATION_METHOD} with one of your chosen quantization method)

huggingface-cli download gorilla-llm/gorilla-openfunctions-v2-gguf gorilla-openfunctions-v2-{QUANTIZATION_METHOD}.gguf --local-dir gorilla-openfunctions-v2-GGUF

It will store the QUANTIZATION_METHOD GGUF file to your local directory, gorilla-openfunctions-v2-GGUF.

We support QUANTIZATION_METHOD = {q2_K, q3_K_S, q3_K_M, q3_K_L, q4_K_S, q4_K_M, q5_K_S, q5_K_M, q6_K}. Please let us know what other quantization methods you would like us to include!

Please follow the llama-cpp-python for llama-cpp-python package installation on your machine.

Then, you can run the following example script to see an example of local inference. Fill in YOUR_DIRECTORY in this code snippet. This script is adapted from https://github.com/abetlen/llama-cpp-python and https://github.com/ShishirPatil/gorilla/tree/main/openfunctions

from llama_cpp import Llama
import json

llm = Llama(model_path="YOUR_DIRECTORY/gorilla-openfunctions-v2-GGUF/gorilla-openfunctions-v2-q2_K.gguf", n_threads=8, n_gpu_layers=35)

def get_prompt(user_query: str, functions: list = []) -> str:
    """
    Generates a conversation prompt based on the user's query and a list of functions.

    Parameters:
    - user_query (str): The user's query.
    - functions (list): A list of functions to include in the prompt.

    Returns:
    - str: The formatted conversation prompt.
    """
    system = "You are an AI programming assistant, utilizing the Gorilla LLM model, developed by Gorilla LLM, and you only answer questions related to computer science. For politically sensitive questions, security and privacy issues, and other non-computer science questions, you will refuse to answer."
    if len(functions) == 0:
        return f"{system}\n### Instruction: <<question>> {user_query}\n### Response: "
    functions_string = json.dumps(functions)
    return f"{system}\n### Instruction: <<function>>{functions_string}\n<<question>>{user_query}\n### Response: "

query = "What's the weather like in the two cities of Boston and San Francisco?"
functions = [
    {
        "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"],
        },
    }
]

user_prompt = get_prompt(query, functions)

output = llm(user_prompt,
    max_tokens=512,  # Generate up to 512 tokens
    stop=["<|EOT|>"], 
    echo=True        # Whether to echo the prompt
)

print("Output: ", output)

The expected output of successfully running this script is the following (tested on March 3, 2024)

❯ python quantized_inference.py
llama_model_loader: loaded meta data with 22 key-value pairs and 273 tensors from /Users/charliecheng-jieji/Downloads/codebase/quantized_eval/gorilla-openfunctions-v2-GGUF/gorilla-openfunctions-v2-q2_K.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = llama
llama_model_loader: - kv   1:                               general.name str              = LLaMA v2
llama_model_loader: - kv   2:                       llama.context_length u32              = 4096
llama_model_loader: - kv   3:                     llama.embedding_length u32              = 4096
llama_model_loader: - kv   4:                          llama.block_count u32              = 30
llama_model_loader: - kv   5:                  llama.feed_forward_length u32              = 11008
llama_model_loader: - kv   6:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv   7:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv   8:              llama.attention.head_count_kv u32              = 32
llama_model_loader: - kv   9:     llama.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  10:                       llama.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv  11:                          general.file_type u32              = 10
llama_model_loader: - kv  12:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  13:                      tokenizer.ggml.tokens arr[str,102400]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  14:                      tokenizer.ggml.scores arr[f32,102400]  = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  15:                  tokenizer.ggml.token_type arr[i32,102400]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  16:                      tokenizer.ggml.merges arr[str,99757]   = ["Ġ Ġ", "Ġ t", "Ġ a", "i n", "h e"...
llama_model_loader: - kv  17:                tokenizer.ggml.bos_token_id u32              = 100000
llama_model_loader: - kv  18:                tokenizer.ggml.eos_token_id u32              = 100015
llama_model_loader: - kv  19:            tokenizer.ggml.padding_token_id u32              = 100001
llama_model_loader: - kv  20:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
llama_model_loader: - kv  21:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   61 tensors
llama_model_loader: - type q2_K:  121 tensors
llama_model_loader: - type q3_K:   90 tensors
llama_model_loader: - type q6_K:    1 tensors
llm_load_vocab: mismatch in special tokens definition ( 2387/102400 vs 2400/102400 ).
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = llama
llm_load_print_meta: vocab type       = BPE
llm_load_print_meta: n_vocab          = 102400
llm_load_print_meta: n_merges         = 99757
llm_load_print_meta: n_ctx_train      = 4096
llm_load_print_meta: n_embd           = 4096
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 32
llm_load_print_meta: n_layer          = 30
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_embd_head_k    = 128
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 4096
llm_load_print_meta: n_embd_v_gqa     = 4096
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-06
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: n_ff             = 11008
llm_load_print_meta: n_expert         = 0
llm_load_print_meta: n_expert_used    = 0
llm_load_print_meta: pooling type     = 0
llm_load_print_meta: rope type        = 0
llm_load_print_meta: rope scaling     = linear
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx  = 4096
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: model type       = ?B
llm_load_print_meta: model ftype      = Q2_K - Medium
llm_load_print_meta: model params     = 6.91 B
llm_load_print_meta: model size       = 2.53 GiB (3.14 BPW) 
llm_load_print_meta: general.name     = LLaMA v2
llm_load_print_meta: BOS token        = 100000 '<｜begin▁of▁sentence｜>'
llm_load_print_meta: EOS token        = 100015 '<|EOT|>'
llm_load_print_meta: PAD token        = 100001 '<｜end▁of▁sentence｜>'
llm_load_print_meta: LF token         = 126 'Ä'
llm_load_tensors: ggml ctx size =    0.21 MiB
ggml_backend_metal_buffer_from_ptr: allocated buffer, size =  2457.45 MiB, ( 2457.52 / 10922.67)
llm_load_tensors: offloading 30 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 31/31 layers to GPU
llm_load_tensors:        CPU buffer size =   131.25 MiB
llm_load_tensors:      Metal buffer size =  2457.45 MiB
.....................................................................................
llama_new_context_with_model: n_ctx      = 512
llama_new_context_with_model: freq_base  = 10000.0
llama_new_context_with_model: freq_scale = 1
ggml_metal_init: allocating
ggml_metal_init: found device: Apple M1
ggml_metal_init: picking default device: Apple M1
ggml_metal_init: default.metallib not found, loading from source
ggml_metal_init: GGML_METAL_PATH_RESOURCES = nil
ggml_metal_init: loading '/Users/charliecheng-jieji/miniconda3/envs/public-api/lib/python3.12/site-packages/llama_cpp/ggml-metal.metal'
ggml_metal_init: GPU name:   Apple M1
ggml_metal_init: GPU family: MTLGPUFamilyApple7  (1007)
ggml_metal_init: GPU family: MTLGPUFamilyCommon3 (3003)
ggml_metal_init: GPU family: MTLGPUFamilyMetal3  (5001)
ggml_metal_init: simdgroup reduction support   = true
ggml_metal_init: simdgroup matrix mul. support = true
ggml_metal_init: hasUnifiedMemory              = true
ggml_metal_init: recommendedMaxWorkingSetSize  = 11453.25 MB
ggml_backend_metal_buffer_type_alloc_buffer: allocated buffer, size =   240.00 MiB, ( 2699.33 / 10922.67)
llama_kv_cache_init:      Metal KV buffer size =   240.00 MiB
llama_new_context_with_model: KV self size  =  240.00 MiB, K (f16):  120.00 MiB, V (f16):  120.00 MiB
llama_new_context_with_model:        CPU input buffer size   =    10.01 MiB
ggml_backend_metal_buffer_type_alloc_buffer: allocated buffer, size =   208.00 MiB, ( 2907.33 / 10922.67)
llama_new_context_with_model:      Metal compute buffer size =   208.00 MiB
llama_new_context_with_model:        CPU compute buffer size =     8.00 MiB
llama_new_context_with_model: graph splits (measure): 2
AVX = 0 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 0 | NEON = 1 | ARM_FMA = 1 | F16C = 0 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | SSSE3 = 0 | VSX = 0 | MATMUL_INT8 = 0 | 
Model metadata: {'general.quantization_version': '2', 'tokenizer.chat_template': "{% if not add_generation_prompt is defined %}\n{% set add_generation_prompt = false %}\n{% endif %}\n{%- set ns = namespace(found=false) -%}\n{%- for message in messages -%}\n    {%- if message['role'] == 'system' -%}\n        {%- set ns.found = true -%}\n    {%- endif -%}\n{%- endfor -%}\n{{bos_token}}{%- if not ns.found -%}\n{{'You are an AI programming assistant, utilizing the Gorilla LLM model, developed by Gorilla LLM, and you only answer questions related to computer science. For politically sensitive questions, security and privacy issues, and other non-computer science questions, you will refuse to answer\\n'}}\n{%- endif %}\n{%- for message in messages %}\n    {%- if message['role'] == 'system' %}\n{{ message['content'] }}\n    {%- else %}\n        {%- if message['role'] == 'user' %}\n{{'### Instruction:\\n' + message['content'] + '\\n'}}\n        {%- else %}\n{{'### Response:\\n' + message['content'] + '\\n<|EOT|>\\n'}}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{% if add_generation_prompt %}\n{{'### Response:'}}\n{% endif %}", 'tokenizer.ggml.padding_token_id': '100001', 'tokenizer.ggml.eos_token_id': '100015', 'tokenizer.ggml.bos_token_id': '100000', 'tokenizer.ggml.model': 'gpt2', 'llama.attention.head_count_kv': '32', 'llama.context_length': '4096', 'llama.attention.head_count': '32', 'llama.rope.freq_base': '10000.000000', 'llama.rope.dimension_count': '128', 'general.file_type': '10', 'llama.feed_forward_length': '11008', 'llama.embedding_length': '4096', 'llama.block_count': '30', 'general.architecture': 'llama', 'llama.attention.layer_norm_rms_epsilon': '0.000001', 'general.name': 'LLaMA v2'}
Using gguf chat template: {% if not add_generation_prompt is defined %}
{% set add_generation_prompt = false %}
{% endif %}
{%- set ns = namespace(found=false) -%}
{%- for message in messages -%}
    {%- if message['role'] == 'system' -%}
        {%- set ns.found = true -%}
    {%- endif -%}
{%- endfor -%}
{{bos_token}}{%- if not ns.found -%}
{{'You are an AI programming assistant, utilizing the Gorilla LLM model, developed by Gorilla LLM, and you only answer questions related to computer science. For politically sensitive questions, security and privacy issues, and other non-computer science questions, you will refuse to answer\n'}}
{%- endif %}
{%- for message in messages %}
    {%- if message['role'] == 'system' %}
{{ message['content'] }}
    {%- else %}
        {%- if message['role'] == 'user' %}
{{'### Instruction:\n' + message['content'] + '\n'}}
        {%- else %}
{{'### Response:\n' + message['content'] + '\n<|EOT|>\n'}}
        {%- endif %}
    {%- endif %}
{%- endfor %}
{% if add_generation_prompt %}
{{'### Response:'}}
{% endif %}
Using chat eos_token: <|EOT|>
Using chat bos_token: <｜begin▁of▁sentence｜>

llama_print_timings:        load time =    1890.11 ms
llama_print_timings:      sample time =      23.48 ms /    40 runs   (    0.59 ms per token,  1703.94 tokens per second)
llama_print_timings: prompt eval time =    1889.91 ms /   181 tokens (   10.44 ms per token,    95.77 tokens per second)
llama_print_timings:        eval time =    2728.54 ms /    39 runs   (   69.96 ms per token,    14.29 tokens per second)
llama_print_timings:       total time =    5162.12 ms /   220 tokens

Output:  {'id': 'cmpl-0679223d-578f-42be-bbce-0e307faddd28', 'object': 'text_completion', 'created': 1709525244, 'model': '/Users/charliecheng-jieji/Downloads/codebase/quantized_eval/gorilla-openfunctions-v2-GGUF/gorilla-openfunctions-v2-q2_K.gguf', 'choices': [{'text': 'You are an AI programming assistant, utilizing the Gorilla LLM model, developed by Gorilla LLM, and you only answer questions related to computer science. For politically sensitive questions, security and privacy issues, and other non-computer science questions, you will refuse to answer.\n### Instruction: <<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"]}}]\n<<question>>What\'s the weather like in the two cities of Boston and San Francisco?\n### Response:  <<function>>get_current_weather(location=\'Boston\', unit=\'fahrenheit\')<<function>>get_current_weather(location=\'San Francisco\', unit=\'fahrenheit\')', 'index': 0, 'logprobs': None, 'finish_reason': 'stop'}], 'usage': {'prompt_tokens': 181, 'completion_tokens': 39, 'total_tokens': 220}}