metadata

language:
  - en
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
model-index:
  - name: speechless-zephyr-code-functionary-7b
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: IFEval (0-Shot)
          type: HuggingFaceH4/ifeval
          args:
            num_few_shot: 0
        metrics:
          - type: inst_level_strict_acc and prompt_level_strict_acc
            value: 26.96
            name: strict accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=uukuguy/speechless-zephyr-code-functionary-7b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: BBH (3-Shot)
          type: BBH
          args:
            num_few_shot: 3
        metrics:
          - type: acc_norm
            value: 25.98
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=uukuguy/speechless-zephyr-code-functionary-7b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MATH Lvl 5 (4-Shot)
          type: hendrycks/competition_math
          args:
            num_few_shot: 4
        metrics:
          - type: exact_match
            value: 3.63
            name: exact match
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=uukuguy/speechless-zephyr-code-functionary-7b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GPQA (0-shot)
          type: Idavidrein/gpqa
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 6.71
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=uukuguy/speechless-zephyr-code-functionary-7b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MuSR (0-shot)
          type: TAUR-Lab/MuSR
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 11.61
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=uukuguy/speechless-zephyr-code-functionary-7b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU-PRO (5-shot)
          type: TIGER-Lab/MMLU-Pro
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 23.27
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=uukuguy/speechless-zephyr-code-functionary-7b
          name: Open LLM Leaderboard

speechless-zephyr-code-functionary-7b

4,5,8-bit GGUF models for CPU+GPU inference

This model is the one of the moloras (Mixture-of-Multi-LoRAs) experiments.

Extract LoRA modules from below models (all based Mistral-7B-v0.1), each LoRA module has its own unique skills. By using multi-loras, they can be combined together statically or dynamically to form a versatile new model.

HuggingFaceH4/zephyr-7b-beta (Uncensored Model)
meetkai/functionary-small-v2.2 (Execute functions/plugins)
uukuguy/speechless-code-mistral-7b-v1.0 (Enhance Coding)

The entire process is completed through the use of extract-lora, merge-lora, and lora-hub provided by multi-loras.

The router of mixture-of-multi-loras enables an automatic assembling of LoRA modules, using a gradientfree approach to obtain the coefficients of LoRA modules and requiring only a handful of inference steps for unseen tasks.

Code: https://github.com/uukuguy/multi_loras

LM-Evaluation-Harness

Open LLM Leaderboard

Metric	Value
ARC	61.52
HellaSwag	83.88
MMLU	64.71
TruthfulQA	44.99
Winogrande	78.69
GSM8K	43.82
Average	62.93

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	16.36
IFEval (0-Shot)	26.96
BBH (3-Shot)	25.98
MATH Lvl 5 (4-Shot)	3.63
GPQA (0-shot)	6.71
MuSR (0-shot)	11.61
MMLU-PRO (5-shot)	23.27