richarderkhov/futurice_-_gpt3-finnish-xl-instruct-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "",
    "summary": "The model gpt3-finnish-xl-instruct is an instruction fine-tuned model intended for RAG type Q&A in Finnish.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "Quantization made by Richard Erkhov.\n\n[Github](https://github.com/RichardErkhov)\n\n[Discord](https://discord.gg/pvy7H8DZMG)\n\n[Request more models](https://github.com/RichardErkhov/quant_request)\n\n\ngpt3-finnish-xl-instruct - GGUF\n- Model creator: https://huggingface.co/Futurice/\n- Original model: https://huggingface.co/Futurice/gpt3-finnish-xl-instruct/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [gpt3-finnish-xl-instruct.Q2_K.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q2_K.gguf) | Q2_K | 2.52GB |\n| [gpt3-finnish-xl-instruct.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.IQ3_XS.gguf) | IQ3_XS | 2.52GB |\n| [gpt3-finnish-xl-instruct.IQ3_S.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.IQ3_S.gguf) | IQ3_S | 2.52GB |\n| [gpt3-finnish-xl-instruct.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q3_K_S.gguf) | Q3_K_S | 2.52GB |\n| [gpt3-finnish-xl-instruct.IQ3_M.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.IQ3_M.gguf) | IQ3_M | 2.52GB |\n| [gpt3-finnish-xl-instruct.Q3_K.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q3_K.gguf) | Q3_K | 2.57GB |\n| [gpt3-finnish-xl-instruct.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q3_K_M.gguf) | Q3_K_M | 2.57GB |\n| [gpt3-finnish-xl-instruct.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q3_K_L.gguf) | Q3_K_L | 2.59GB |\n| [gpt3-finnish-xl-instruct.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.IQ4_XS.gguf) | IQ4_XS | 2.53GB |\n| [gpt3-finnish-xl-instruct.Q4_0.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q4_0.gguf) | Q4_0 | 1.06GB |\n| [gpt3-finnish-xl-instruct.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.IQ4_NL.gguf) | IQ4_NL | 0.02GB |\n| [gpt3-finnish-xl-instruct.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q4_K_S.gguf) | Q4_K_S | 2.57GB |\n| [gpt3-finnish-xl-instruct.Q4_K.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q4_K.gguf) | Q4_K | 2.64GB |\n| [gpt3-finnish-xl-instruct.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q4_K_M.gguf) | Q4_K_M | 2.64GB |\n| [gpt3-finnish-xl-instruct.Q4_1.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q4_1.gguf) | Q4_1 | 1.15GB |\n| [gpt3-finnish-xl-instruct.Q5_0.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q5_0.gguf) | Q5_0 | 1.23GB |\n| [gpt3-finnish-xl-instruct.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q5_K_S.gguf) | Q5_K_S | 2.59GB |\n| [gpt3-finnish-xl-instruct.Q5_K.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q5_K.gguf) | Q5_K | 2.65GB |\n| [gpt3-finnish-xl-instruct.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q5_K_M.gguf) | Q5_K_M | 2.65GB |\n| [gpt3-finnish-xl-instruct.Q5_1.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q5_1.gguf) | Q5_1 | 1.32GB |\n| [gpt3-finnish-xl-instruct.Q6_K.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q6_K.gguf) | Q6_K | 2.71GB |\n| [gpt3-finnish-xl-instruct.Q8_0.gguf](https://huggingface.co/RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf/blob/main/gpt3-finnish-xl-instruct.Q8_0.gguf) | Q8_0 | 1.76GB |\n\n\n\n\nOriginal model description:\n---\nbase_model: TurkuNLP/gpt3-finnish-xl\nlicense: apache-2.0\ndatasets:\n- TurkuNLP/squad_v2_fi\nlanguage:\n- fi\npipeline_tag: text-generation\n---\n\n# Model Card for Model Futurice/gpt3-finnish-xl-instruct\n\nThe model gpt3-finnish-xl-instruct is an instruction fine-tuned model intended for RAG type Q&A in Finnish.\n\n## Model Details\n\n### Model Description\n\nThe gpt3-finnish-xl-instruct model is based on TurkuNLP Finnish GPT-3-models. They are a model family of pretrained monolingual GPT-style language models, based on BLOOM-architecture.\n\nThe model was fine-tuned using a sample of dataset TurkuNLP/squad_v2_fi, that was DeepL translated from SQuAD2.0.\n\n- **Developed by:** Martti Sutinen\n- **Model type:** Bloom\n- **Language(s) (NLP):** Finnish\n- **License:** Apache-2.0\n- **Finetuned from model:** TurkuNLP/gpt3-finnish-large\n\n## Uses\n\nIntended for RAG type Q&A in Finnish.\n\n### Direct Use\n\nIntended for text generation and RAG type Q&A in Finnish. Supply a context and ask a question about it.\n\n### Out-of-Scope Use\n\nPlease do not misuse the model. Not recommended for other use cases.\n\n## Bias, Risks, and Limitations\n\nA key limitation is simple and limited selection of fine-tuning data. Please do not expect high quality answers.\n\n### Recommendations\n\nRecommeded to continue fine-tuning with more data or newer architecture.\n\n## How to Get Started with the Model\n\n- Recommended system message: \"Olet avustaja. Seuraavaksi saat kysymyksen tai tehtävän. Kirjoita vastaus parhaasi mukaan siten että se täyttää kysymyksen tai tehtävän vaatimukset.\"\n- Recommended format for question about context: Tausta: \"{context} \\n\\nKäytä vain taustaa ja vastaa kysymykseen tai tehtävään: {question}\"\n- Prompt format: tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)\n\nWhere messages with typical format:\nmessages = [\n    {\"role\": \"system\", \"content\": system_message},\n    {\"role\": \"user\", \"content\": prompt_with_context}\n].\n\nHere is what the input could look like: \n\n\\<s><|im_start|>system  \nOlet avustaja. Seuraavaksi saat kysymyksen tai tehtävän. Kirjoita vastaus parhaasi mukaan siten että se täyttää kysymyksen tai tehtävän vaatimukset.<|im_end|>  \n<|im_start|>user  \nTausta:  \nDokumentti luotiin tammikuussa. Sen kirjoittajaa ei tunneta.  \n  \nKäytä vain taustaa ja vastaa kysymykseen tai tehtävään: Milloin dokumentti kirjoitettiin?<|im_end|>  \n<|im_start|>assistant  \n\n\nUse pipeline with task text-generation and the recommended format.\n\n## Training Details\n\n### Training Data\n\nTrained with 40000 random samples from test data in: [TurkuNLP/squad_v2_fi](https://huggingface.co/datasets/TurkuNLP/squad_v2_fi).\n\n### Training Procedure\n\nTraining was done for 4-bit base model with supervised fine-tuning and Lora.\n\n#### Training Hyperparameters\n\n- **Training regime:** 4-bit, batch size 4, max steps 20000, data collator for completion only\n\n## Evaluation\n\nEvaluation has not been done properly yet.\n\n### Testing Data, Factors & Metrics\n\n#### Testing Data\n\nEvaluated with 1000 random samples from test data in: [TurkuNLP/squad_v2_fi](https://huggingface.co/datasets/TurkuNLP/squad_v2_fi).\n\n#### Factors\n\nSame factors as in SQuAD2.0.\n\n#### Metrics\n\nLoss.\n\n### Results\n\nNo results to be shared yet.\n\n#### Summary\n\n## Environmental Impact\n\nEnvironmental impact not yet evaluated.\n\nCarbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).\n\n- **Hardware Type:** Mostly trained on A100\n- **Hours used:** 5-10 hours\n- **Cloud Provider:** GCP\n- **Compute Region:** Unknown\n- **Carbon Emitted:** Not evaluated\n\n### Model Architecture and Objective\n\nBloom.\n\n### Compute Infrastructure\n\nColab.\n\n#### Hardware\n\n1 x A100.\n\n#### Software\n\nTypical software used.\n\n## Model Card Contact\n\nMartti Sutinen\n\n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "arxiv:1910.09700",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 1,
  "downloads": 90,
  "gated": false,
  "private": false,
  "last_modified": "2024-10-12T08:04:11.000Z",
  "created_at": "2024-10-12T07:17:57.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "670a22a5b8b5fde64c41a13f",
  "id": "RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf",
  "modelId": "RichardErkhov/Futurice_-_gpt3-finnish-xl-instruct-gguf",
  "sha": "e68b2d45fcd4cbafd0b3428dd4710b17a33b93d9",
  "createdAt": "2024-10-12T07:17:57.000Z",
  "lastModified": "2024-10-12T08:04:11.000Z",
  "author": "RichardErkhov",
  "downloads": 90,
  "likes": 1,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 24
}