mradermacher/smol_llama-4x220m-moe-i1-gguf overview

{
  "metadata": {},
  "card_data": {
    "base_model": "Isotonic/smol_llama-4x220M-MoE",
    "datasets": [
      "JeanKaddour/minipile",
      "pszemraj/simple_wikipedia_LM",
      "mattymchen/refinedweb-3m",
      "HuggingFaceH4/ultrachat_200k",
      "teknium/openhermes",
      "HuggingFaceH4/ultrafeedback_binarized",
      "EleutherAI/proof-pile-2",
      "bigcode/the-stack-smol-xl"
    ],
    "language": [
      "en"
    ],
    "library_name": "transformers",
    "license": "apache-2.0",
    "quantized_by": "mradermacher",
    "tags": [
      "moe",
      "merge",
      "mergekit",
      "lazymergekit",
      "BEE-spoke-data/smol_llama-220M-openhermes",
      "BEE-spoke-data/beecoder-220M-python",
      "BEE-spoke-data/zephyr-220m-sft-full",
      "BEE-spoke-data/zephyr-220m-dpo-full",
      "text-generation"
    ],
    "frontmatter": {
      "base_model": "Isotonic/smol_llama-4x220M-MoE",
      "datasets": [
        "JeanKaddour/minipile",
        "pszemraj/simple_wikipedia_LM",
        "mattymchen/refinedweb-3m",
        "HuggingFaceH4/ultrachat_200k",
        "teknium/openhermes",
        "HuggingFaceH4/ultrafeedback_binarized",
        "EleutherAI/proof-pile-2",
        "bigcode/the-stack-smol-xl"
      ],
      "language": [
        "en"
      ],
      "library_name": "transformers",
      "license": "apache-2.0",
      "quantized_by": "mradermacher",
      "tags": [
        "moe",
        "merge",
        "mergekit",
        "lazymergekit",
        "BEE-spoke-data/smol_llama-220M-openhermes",
        "BEE-spoke-data/beecoder-220M-python",
        "BEE-spoke-data/zephyr-220m-sft-full",
        "BEE-spoke-data/zephyr-220m-dpo-full",
        "text-generation"
      ]
    },
    "hero_image_url": "https://www.nethype.de/huggingface_embed/quantpplgraph.png",
    "summary": "## About      weighted/imatrix quants of https://huggingface.co/Isotonic/smol_llama-4x220M-MoE  static quants are available at https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-GGUF",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nbase_model: Isotonic/smol_llama-4x220M-MoE\ndatasets:\n- JeanKaddour/minipile\n- pszemraj/simple_wikipedia_LM\n- mattymchen/refinedweb-3m\n- HuggingFaceH4/ultrachat_200k\n- teknium/openhermes\n- HuggingFaceH4/ultrafeedback_binarized\n- EleutherAI/proof-pile-2\n- bigcode/the-stack-smol-xl\nlanguage:\n- en\nlibrary_name: transformers\nlicense: apache-2.0\nquantized_by: mradermacher\ntags:\n- moe\n- merge\n- mergekit\n- lazymergekit\n- BEE-spoke-data/smol_llama-220M-openhermes\n- BEE-spoke-data/beecoder-220M-python\n- BEE-spoke-data/zephyr-220m-sft-full\n- BEE-spoke-data/zephyr-220m-dpo-full\n- text-generation\n---\n## About\n\n<!-- ### quantize_version: 2 -->\n<!-- ### output_tensor_quantised: 1 -->\n<!-- ### convert_type: hf -->\n<!-- ### vocab_type:  -->\n<!-- ### tags: nicoboss -->\nweighted/imatrix quants of https://huggingface.co/Isotonic/smol_llama-4x220M-MoE\n\n<!-- provided-files -->\nstatic quants are available at https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-GGUF\n## Usage\n\nIf you are unsure how to use GGUF files, refer to one of [TheBloke's\nREADMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for\nmore details, including on how to concatenate multi-part files.\n\n## Provided Quants\n\n(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)\n\n| Link | Type | Size/GB | Notes |\n|:-----|:-----|--------:|:------|\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ1_S.gguf) | i1-IQ1_S | 0.2 | for the desperate |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ1_M.gguf) | i1-IQ1_M | 0.3 | mostly desperate |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 0.3 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ2_XS.gguf) | i1-IQ2_XS | 0.3 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ2_S.gguf) | i1-IQ2_S | 0.3 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ2_M.gguf) | i1-IQ2_M | 0.3 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q2_K_S.gguf) | i1-Q2_K_S | 0.3 | very low quality |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q2_K.gguf) | i1-Q2_K | 0.3 | IQ3_XXS probably better |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 0.3 | lower quality |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ3_XS.gguf) | i1-IQ3_XS | 0.4 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q3_K_S.gguf) | i1-Q3_K_S | 0.4 | IQ3_XS probably better |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ3_S.gguf) | i1-IQ3_S | 0.4 | beats Q3_K* |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ3_M.gguf) | i1-IQ3_M | 0.4 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q3_K_M.gguf) | i1-Q3_K_M | 0.4 | IQ3_S probably better |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q3_K_L.gguf) | i1-Q3_K_L | 0.4 | IQ3_M probably better |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ4_XS.gguf) | i1-IQ4_XS | 0.4 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-IQ4_NL.gguf) | i1-IQ4_NL | 0.4 | prefer IQ4_XS |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q4_0.gguf) | i1-Q4_0 | 0.4 | fast, low quality |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q4_K_S.gguf) | i1-Q4_K_S | 0.4 | optimal size/speed/quality |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q4_K_M.gguf) | i1-Q4_K_M | 0.5 | fast, recommended |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q4_1.gguf) | i1-Q4_1 | 0.5 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q5_K_S.gguf) | i1-Q5_K_S | 0.5 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q5_K_M.gguf) | i1-Q5_K_M | 0.5 |  |\n| [GGUF](https://huggingface.co/mradermacher/smol_llama-4x220M-MoE-i1-GGUF/resolve/main/smol_llama-4x220M-MoE.i1-Q6_K.gguf) | i1-Q6_K | 0.6 | practically like static Q6_K |\n\nHere is a handy graph by ikawrakow comparing some lower-quality quant\ntypes (lower is better):\n\n![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)\n\nAnd here are Artefact2's thoughts on the matter:\nhttps://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9\n\n## FAQ / Model Request\n\nSee https://huggingface.co/mradermacher/model_requests for some answers to\nquestions you might have and/or if you want some other model quantized.\n\n## Thanks\n\nI thank my company, [nethype GmbH](https://www.nethype.de/), for letting\nme use its servers and providing upgrades to my workstation to enable\nthis work in my free time. Additional thanks to [@nicoboss](https://huggingface.co/nicoboss) for giving me access to his private supercomputer, enabling me to provide many more imatrix quants, at much higher quality, than I would otherwise be able to.\n\n<!-- end -->\n",
    "related_quantizations": []
  },
  "tags": [
    "transformers",
    "gguf",
    "moe",
    "merge",
    "mergekit",
    "lazymergekit",
    "BEE-spoke-data/smol_llama-220M-openhermes",
    "BEE-spoke-data/beecoder-220M-python",
    "BEE-spoke-data/zephyr-220m-sft-full",
    "BEE-spoke-data/zephyr-220m-dpo-full",
    "text-generation",
    "en",
    "dataset:JeanKaddour/minipile",
    "dataset:pszemraj/simple_wikipedia_LM",
    "dataset:mattymchen/refinedweb-3m",
    "dataset:HuggingFaceH4/ultrachat_200k",
    "dataset:teknium/openhermes",
    "dataset:HuggingFaceH4/ultrafeedback_binarized",
    "dataset:EleutherAI/proof-pile-2",
    "dataset:bigcode/the-stack-smol-xl",
    "base_model:Isotonic/smol_llama-4x220M-MoE",
    "base_model:quantized:Isotonic/smol_llama-4x220M-MoE",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "imatrix"
  ],
  "likes": 0,
  "downloads": 298,
  "gated": false,
  "private": false,
  "last_modified": "2025-01-24T03:58:06.000Z",
  "created_at": "2025-01-24T03:48:19.000Z",
  "pipeline_tag": "text-generation",
  "library_name": "transformers"
}

{
  "_id": "67930d8357b2fe2b1eed858b",
  "id": "mradermacher/smol_llama-4x220M-MoE-i1-GGUF",
  "modelId": "mradermacher/smol_llama-4x220M-MoE-i1-GGUF",
  "sha": "26e7afbccdfe265352e0bd364853ee9c68e6f365",
  "createdAt": "2025-01-24T03:48:19.000Z",
  "lastModified": "2025-01-24T03:58:06.000Z",
  "author": "mradermacher",
  "downloads": 298,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "text-generation",
  "library_name": "transformers",
  "siblings_count": 27
}