README.md

metadata

license: llama2
datasets:
  - HiTZ/latxa-corpus-v1.1
language:
  - eu
  - en
metrics:
  - accuracy
  - f1
  - perplexity
pipeline_tag: text-generation
tags:
  - TensorBlock
  - GGUF
base_model: HiTZ/latxa-7b-v1.2
model-index:
  - name: Latxa-7b-v1.2
    results:
      - task:
          type: multiple-choice
        dataset:
          name: xstory_cloze
          type: XStory
        metrics:
          - type: Accuracy (0-shot)
            value: 65.45
            name: Accuracy (0-shot)
        source:
          url: https://arxiv.org/abs/2403.20266
          name: Paper
      - task:
          type: multiple-choice
        dataset:
          name: belebele
          type: Belebele
        metrics:
          - type: Accuracy (5-shot)
            value: 37.33
            name: Accuracy (5-shot)
        source:
          url: https://arxiv.org/abs/2403.20266
          name: Paper
      - task:
          type: mix
        dataset:
          name: basque_glue
          type: BasqueGLUE
        metrics:
          - type: Average scores (5-shot)
            value: 52.56
            name: Average scores (5-shot)
        source:
          url: https://arxiv.org/abs/2403.20266
          name: Paper
      - task:
          type: multiple_choice
        dataset:
          name: eus_proficiency
          type: EusProficiency
        metrics:
          - type: Accuracy (5-shot)
            value: 30.26
            name: Accuracy (5-shot)
        source:
          url: https://arxiv.org/abs/2403.20266
          name: Paper
      - task:
          type: multiple_choice
        dataset:
          name: eus_reading
          type: EusReading
        metrics:
          - type: Accuracy (5-shot)
            value: 25
            name: Accuracy (5-shot)
        source:
          url: https://arxiv.org/abs/2403.20266
          name: Paper
      - task:
          type: multiple_choice
        dataset:
          name: eus_trivia
          type: EusTrivia
        metrics:
          - type: Accuracy (5-shot)
            value: 42.16
            name: Accuracy (5-shot)
        source:
          url: https://arxiv.org/abs/2403.20266
          name: Paper
      - task:
          type: multiple_choice
        dataset:
          name: eus_exams
          type: EusExams
        metrics:
          - type: Accuracy (5-shot)
            value: 33.82
            name: Accuracy (5-shot)
        source:
          url: https://arxiv.org/abs/2403.20266
          name: Paper

HiTZ/latxa-7b-v1.2 - GGUF

This repo contains GGUF format model files for HiTZ/latxa-7b-v1.2.

The files were quantized using machines provided by TensorBlock, and they are compatible with llama.cpp as of commit b4011.

Our projects

Forge
An OpenAI-compatible multi-provider routing layer.
🚀 Try it now! 🚀
Awesome MCP Servers	TensorBlock Studio
A comprehensive collection of Model Context Protocol (MCP) servers.	A lightweight, open, and extensible multi-LLM interaction studio.
👀 See what we built 👀	👀 See what we built 👀

## Prompt template

Model file specification

Filename	Quant type	File Size	Description
latxa-7b-v1.2-Q2_K.gguf	Q2_K	2.533 GB	smallest, significant quality loss - not recommended for most purposes
latxa-7b-v1.2-Q3_K_S.gguf	Q3_K_S	2.948 GB	very small, high quality loss
latxa-7b-v1.2-Q3_K_M.gguf	Q3_K_M	3.298 GB	very small, high quality loss
latxa-7b-v1.2-Q3_K_L.gguf	Q3_K_L	3.597 GB	small, substantial quality loss
latxa-7b-v1.2-Q4_0.gguf	Q4_0	3.826 GB	legacy; small, very high quality loss - prefer using Q3_K_M
latxa-7b-v1.2-Q4_K_S.gguf	Q4_K_S	3.857 GB	small, greater quality loss
latxa-7b-v1.2-Q4_K_M.gguf	Q4_K_M	4.081 GB	medium, balanced quality - recommended
latxa-7b-v1.2-Q5_0.gguf	Q5_0	4.652 GB	legacy; medium, balanced quality - prefer using Q4_K_M
latxa-7b-v1.2-Q5_K_S.gguf	Q5_K_S	4.652 GB	large, low quality loss - recommended
latxa-7b-v1.2-Q5_K_M.gguf	Q5_K_M	4.783 GB	large, very low quality loss - recommended
latxa-7b-v1.2-Q6_K.gguf	Q6_K	5.529 GB	very large, extremely low quality loss
latxa-7b-v1.2-Q8_0.gguf	Q8_0	7.161 GB	very large, extremely low quality loss - not recommended

Downloading instruction

Command line

Firstly, install Huggingface Client

pip install -U "huggingface_hub[cli]"

Then, downoad the individual model file the a local directory

huggingface-cli download tensorblock/latxa-7b-v1.2-GGUF --include "latxa-7b-v1.2-Q2_K.gguf" --local-dir MY_LOCAL_DIR

If you wanna download multiple model files with a pattern (e.g., *Q4_K*gguf), you can try:

huggingface-cli download tensorblock/latxa-7b-v1.2-GGUF --local-dir MY_LOCAL_DIR --local-dir-use-symlinks False --include='*Q4_K*gguf'