richarderkhov/qwen_-_qwen1.5-moe-a2.7b-chat-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "",
    "summary": "",
    "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\nQwen1.5-MoE-A2.7B-Chat - GGUF\n- Model creator: https://huggingface.co/Qwen/\n- Original model: https://huggingface.co/Qwen/Qwen1.5-MoE-A2.7B-Chat/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [Qwen1.5-MoE-A2.7B-Chat.Q2_K.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q2_K.gguf) | Q2_K | 5.49GB |\n| [Qwen1.5-MoE-A2.7B-Chat.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.IQ3_XS.gguf) | IQ3_XS | 6.07GB |\n| [Qwen1.5-MoE-A2.7B-Chat.IQ3_S.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.IQ3_S.gguf) | IQ3_S | 6.37GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q3_K_S.gguf) | Q3_K_S | 6.37GB |\n| [Qwen1.5-MoE-A2.7B-Chat.IQ3_M.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.IQ3_M.gguf) | IQ3_M | 6.46GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q3_K.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q3_K.gguf) | Q3_K | 6.93GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q3_K_M.gguf) | Q3_K_M | 6.93GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q3_K_L.gguf) | Q3_K_L | 7.21GB |\n| [Qwen1.5-MoE-A2.7B-Chat.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.IQ4_XS.gguf) | IQ4_XS | 7.4GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q4_0.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q4_0.gguf) | Q4_0 | 7.59GB |\n| [Qwen1.5-MoE-A2.7B-Chat.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.IQ4_NL.gguf) | IQ4_NL | 7.68GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q4_K_S.gguf) | Q4_K_S | 8.11GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q4_K.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q4_K.gguf) | Q4_K | 8.84GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q4_K_M.gguf) | Q4_K_M | 8.84GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q4_1.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q4_1.gguf) | Q4_1 | 8.41GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q5_0.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q5_0.gguf) | Q5_0 | 9.22GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q5_K_S.gguf) | Q5_K_S | 9.46GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q5_K.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q5_K.gguf) | Q5_K | 10.09GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q5_K_M.gguf) | Q5_K_M | 10.09GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q5_1.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q5_1.gguf) | Q5_1 | 10.04GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q6_K.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q6_K.gguf) | Q6_K | 11.89GB |\n| [Qwen1.5-MoE-A2.7B-Chat.Q8_0.gguf](https://huggingface.co/RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf/blob/main/Qwen1.5-MoE-A2.7B-Chat.Q8_0.gguf) | Q8_0 | 14.18GB |\n\n\n\n\nOriginal model description:\n---\nlicense: other\nlicense_name: tongyi-qianwen\nlicense_link: >-\n  https://huggingface.co/Qwen/Qwen1.5-MoE-A2.7B-Chat/blob/main/LICENSE\nlanguage:\n- en\npipeline_tag: text-generation\ntags:\n- chat\n---\n\n# Qwen1.5-MoE-A2.7B-Chat\n\n\n## Introduction\n\nQwen1.5-MoE is a transformer-based MoE decoder-only language model pretrained on a large amount of data. \n\nFor more details, please refer to our [blog post](https://qwenlm.github.io/blog/qwen-moe/) and [GitHub repo](https://github.com/QwenLM/Qwen1.5).\n\n## Model Details\nQwen1.5-MoE employs Mixture of Experts (MoE) architecture, where the models are upcycled from dense language models. For instance, `Qwen1.5-MoE-A2.7B` is upcycled from `Qwen-1.8B`. It has 14.3B parameters in total and 2.7B activated parameters during runtime, while achieching comparable performance to `Qwen1.5-7B`, it only requires 25% of the training resources. We also observed that the inference speed is 1.74 times that of `Qwen1.5-7B`.\n\n## Training details\nWe pretrained the models with a large amount of data, and we post-trained the models with both supervised finetuning and direct preference optimization. \n\n## Requirements\nThe code of Qwen1.5-MoE has been in the latest Hugging face transformers and we advise you to build from source with command `pip install git+https://github.com/huggingface/transformers`, or you might encounter the following error:\n```\nKeyError: 'qwen2_moe'.\n```\n\n## Quickstart\n\nHere provides a code snippet with `apply_chat_template` to show you how to load the tokenizer and model and how to generate contents.\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\ndevice = \"cuda\" # the device to load the model onto\n\nmodel = AutoModelForCausalLM.from_pretrained(\n    \"Qwen/Qwen1.5-MoE-A2.7B-Chat\",\n    torch_dtype=\"auto\",\n    device_map=\"auto\"\n)\ntokenizer = AutoTokenizer.from_pretrained(\"Qwen/Qwen1.5-MoE-A2.7B-Chat\")\n\nprompt = \"Give me a short introduction to large language model.\"\nmessages = [\n    {\"role\": \"system\", \"content\": \"You are a helpful assistant.\"},\n    {\"role\": \"user\", \"content\": prompt}\n]\ntext = tokenizer.apply_chat_template(\n    messages,\n    tokenize=False,\n    add_generation_prompt=True\n)\nmodel_inputs = tokenizer([text], return_tensors=\"pt\").to(device)\n\ngenerated_ids = model.generate(\n    model_inputs.input_ids,\n    max_new_tokens=512\n)\ngenerated_ids = [\n    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)\n]\n\nresponse = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\n```\n\nFor quantized models, we advise you to use the GPTQ correspondents, namely `Qwen1.5-MoE-A2.7B-Chat-GPTQ-Int4`.\n\n\n## Tips\n\n* If you encounter code switching or other bad cases, we advise you to use our provided hyper-parameters in `generation_config.json`.\n* \n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 2,
  "downloads": 147,
  "gated": false,
  "private": false,
  "last_modified": "2024-05-13T19:05:37.000Z",
  "created_at": "2024-05-13T14:45:54.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "664227a2e9a084919edc5f67",
  "id": "RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf",
  "modelId": "RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-Chat-gguf",
  "sha": "947580e7b3904cd4081810d0efec4520a75609c7",
  "createdAt": "2024-05-13T14:45:54.000Z",
  "lastModified": "2024-05-13T19:05:37.000Z",
  "author": "RichardErkhov",
  "downloads": 147,
  "likes": 2,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 24
}