diff --git a/catalog.json b/catalog.json index 134be18..af478fb 100644 --- a/catalog.json +++ b/catalog.json @@ -1,58 +1,58 @@ [ { "_descriptorVersion": "0.0.1", - "datePublished": "2023-06-03T05:34:04.000Z", - "name": "Nous-Hermes-13b", - "description": "Nous-Hermes-13b is a state-of-the-art language model fine-tuned on over 300,000 instructions. This model was fine-tuned by Nous Research, with Teknium and Karan4D leading the fine tuning process and dataset curation, Redmond AI sponsoring the compute, and several other contributors. The result is an enhanced Llama 13b model that rivals GPT-3.5-turbo in performance across a variety of tasks. This model stands out for its long responses, low hallucination rate, and absence of OpenAI censorship mechanisms. The fine-tuning process was performed with a 2000 sequence length on an 8x a100 80GB DGX machine for over 50 hours.", + "datePublished": "2023-06-25T11:37:35.000Z", + "name": "Airboros-13b-gpt4-1.4", + "description": "This qlora fine-tuned 13b parameter LlaMa model uses synthetic training data created via github.com/jondurbin/airoboros. It extends version 1.1, introducing thousands of new training data and an update for \"PLAINFORMAT\" to print code without backticks or explanations. The dataset, available online, focuses on coding, math/reasoning, trivia, role playing, multiple choice, fill-in-the-blank, context-obedient question answering, theory of mind, and general topics. The model was fine-tuned with a qlora fork, updated to use a modified vicuna template compatible with 7b/13b versions. The format involves a preamble/system prompt, followed by \"USER: [prompt] ASSISTANT: \", with the prompt allowing for multiple lines and spaces.", "author": { - "name": "Nous Research", - "url": "https://nousresearch.com", - "blurb": "Nous Research is dedicated to advancing the field of natural language processing, in collaboration with the open-source community, through bleeding-edge research and a commitment to symbiotic development." + "name": "Jon Durbin", + "url": "https://github.com/jondurbin", + "blurb": "Jon Durbin is a Computer Scientist and the author of the Airboros (7B, 13B, 33B, 65B) qlora fine-tuned LlaMa family of models." }, "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/NousResearch/Nous-Hermes-13b", - "downloadUrl": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML" + "canonicalUrl": "https://huggingface.co/jondurbin/airoboros-13b-gpt4-1.4", + "downloadUrl": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "nous-hermes-13b.ggmlv3.q3_K_S.bin" + "name": "airoboros-13b-gpt4-1.4.ggmlv3.q4_K_S.bin" }, "most_capable": { - "name": "nous-hermes-13b.ggmlv3.q6_K.bin" + "name": "airoboros-13b-gpt4-1.4.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "nous-hermes-13b.ggmlv3.q3_K_S.bin", - "url": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML/resolve/main/nous-hermes-13b.ggmlv3.q3_K_S.bin", - "sizeBytes": 5594695104, + "name": "airoboros-13b-gpt4-1.4.ggmlv3.q4_K_S.bin", + "url": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML/resolve/main/airoboros-13b-gpt4-1.4.ggmlv3.q4_K_S.bin", + "sizeBytes": 7365545088, "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "591a49f1ef4dbc2cf43943c5ec9bd617e6086264e30f66718bc764bc55286b5e", + "sha256checksum": "dd5d8019e73de1e99089e9d82dde6a08173818dc0afeb1e95c0bb7ec77891eaf", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/Nous-Hermes-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML" + "respository": "TheBloke/airoboros-13B-gpt4-1.4-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML" }, { - "name": "nous-hermes-13b.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML/resolve/main/nous-hermes-13b.ggmlv3.q6_K.bin", - "sizeBytes": 10678859104, + "name": "airoboros-13b-gpt4-1.4.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML/resolve/main/airoboros-13b-gpt4-1.4.ggmlv3.q6_K.bin", + "sizeBytes": 10678850688, "quantization": "Q6_K", "format": "ggml", - "sha256checksum": "efe8ffe14aa97c3c5f45f2bc8e80a02933c5e907813deb685c93341bf671f44e", + "sha256checksum": "5a72a053eb02c9e6e4fa1ee24ed73c89da4877468923309fdec088d3a3fbb5ff", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/Nous-Hermes-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML" + "respository": "TheBloke/airoboros-13B-gpt4-1.4-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML" } ] } @@ -118,234 +118,230 @@ }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-07-18T21:08:14.000Z", - "name": "Llama-2-7B-Chat-GGML", - "description": "This is the 7B model from the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Meta's fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. Llama-2-Chat models outperform open-source chat models on most benchmarks we tested, and in Meta's human evaluations for helpfulness and safety, are on par with some popular closed-source models like ChatGPT and PaLM.", + "datePublished": "2023-07-23T12:47:44.000Z", + "name": "Dolphin-Llama-13B", + "description": "This model, which is based on llama1, is designed specifically for non-commercial use and has undergone a thorough training process using an open-source implementation of Microsoft's Orca dataset. The utilized dataset, post rigorous cleaning and deduplication procedures, encompasses 842,610 instructional instances augmented with GPT-4 completions and an additional 2,625,353 instructional instances augmented with GPT-3.5 completions. The unique aspect of this model is its uncensored nature, having been meticulously filtered to remove any instances of alignment and bias. However, it is advised for users to implement their own alignment layer before deploying this model as a service. The model was trained using the comprehensive flan5m (gpt3.5 completions) dataset for 3 epochs and the flan1m (gpt4 completions) dataset for 2.5 epochs, with distinct learning rates set to preemptively avoid overfitting. This extensive training took approximately 600 hours on a setup of 8x H100s.", "author": { - "name": "Meta AI", - "url": "https://ai.meta.com", - "blurb": "Pushing the boundaries of AI through research, infrastructure and product innovation." + "name": "Eric Hartford", + "url": "https://twitter.com/erhartford", + "blurb": "Eric Hartford is a software engineer and entrepreneur. He specializes in alignment, uncensored models, intersection of AI and society, curating datasets and training chat and instruction tuned models." }, "numParameters": "7B", "resources": { - "canonicalUrl": "https://huggingface.co/meta-llama/Llama-2-7b-chat-hf", - "paperUrl": "https://ai.meta.com/research/publications/llama-2-open-foundation-and-fine-tuned-chat-models/", - "downloadUrl": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML" + "canonicalUrl": "https://huggingface.co/ehartford/dolphin-llama-13b", + "downloadUrl": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "llama-2-7b-chat.ggmlv3.q4_K_S.bin" + "name": "dolphin-llama-13b.ggmlv3.q4_K_S.bin" }, "most_capable": { - "name": "llama-2-7b-chat.ggmlv3.q6_K.bin" + "name": "dolphin-llama-13b.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "llama-2-7b-chat.ggmlv3.q4_K_S.bin", - "url": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/resolve/main/llama-2-7b-chat.ggmlv3.q4_K_S.bin", - "sizeBytes": 3825517184, + "name": "dolphin-llama-13b.ggmlv3.q4_K_S.bin", + "url": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML/resolve/main/dolphin-llama-13b.ggmlv3.q4_K_S.bin", + "sizeBytes": 3791725184, "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "32b758bf5e4f16fb5944b75d577fbca18c11c57000b41c6cc04bb281632d58f3", + "sha256checksum": "3a4e3eff4a0bfa1dc8a0b257ae36abbc532a4e3e09e1e27cd82958ff8addd173", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/Llama-2-7B-Chat-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML" + "respository": "TheBloke/Dolphin-Llama-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML" }, { - "name": "llama-2-7b-chat.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/resolve/main/llama-2-7b-chat.ggmlv3.q6_K.bin", - "sizeBytes": 5528904320, + "name": "dolphin-llama-13b.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML/resolve/main/dolphin-llama-13b.ggmlv3.q6_K.bin", + "sizeBytes": 10678850688, "quantization": "Q6_K", "format": "ggml", - "sha256checksum": "24a2097aba9bc63395654515618fb2ceeaea64452147ee5299990b636e4c00ce", + "sha256checksum": "8258f8c014a939ce93519bca658eb714202dacbee8f5df05b2dee0f02472460d", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/Llama-2-7B-Chat-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML" + "respository": "TheBloke/Dolphin-Llama-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-06-12T11:16:39.000Z", - "name": "Manticore 13B Chat Pyg Guanaco", - "description": "The model, augmented with Guanaco qLora, shows broad capabilities compared to other models like Wizard or Manticore. It excels in in-context learning and reasoning but may have weaknesses in coding. The model follows instructions, works as a chatbot, and produces intelligent responses. It accepts various prompting styles, including the ###-Variant. The model is generally unrestricted and doesn't berate users. Recommended settings include low temperature, low diversity, and slight repetition penalty.", + "datePublished": "2023-10-22T03:04:42", + "name": "falcon-40b-instruct", + "description": "Falcon-40B-Instruct, based on Falcon-40B, has been fine-tuned on chat and instruct datasets. It offers outstanding performance and is an excellent choice for chat and instruct applications. Falcon-7B-Instruct is part of the Falcon family of language models, known for their exceptional capabilities and openness.", "author": { - "name": "Open Access AI Collective", - "url": "https://huggingface.co/openaccess-ai-collective/", - "blurb": "" + "name": "Maddes8cht", + "url": "https://huggingface.co/maddes8cht", + "blurb": "Maddes8cht Passionate about Open Source and AI. On Hugging Face he is advocating for real open source AI models with OSI compliant licenses" }, - "numParameters": "13B", + "numParameters": "40B", "resources": { - "canonicalUrl": "https://huggingface.co/openaccess-ai-collective/manticore-13b-chat-pyg", - "downloadUrl": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML" + "canonicalUrl": "https://huggingface.co/tiiuae/falcon-40b-instruct", + "downloadUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" }, "trainedFor": "chat", - "arch": "llama", + "arch": "falcon", "files": { "highlighted": { "economical": { - "name": "Manticore-13B-Chat-Pyg-Guanaco-GGML-q3_K_M.bin" + "name": "tiiuae-falcon-40b-Q4_K_M.gguf" }, "most_capable": { - "name": "Manticore-13B-Chat-Pyg-Guanaco-GGML-q5_K_S.bin" + "name": "tiiuae-falcon-40b-Q6_K.gguf" } }, "all": [ { - "name": "Manticore-13B-Chat-Pyg-Guanaco-GGML-q3_K_M.bin", - "url": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML/resolve/main/Manticore-13B-Chat-Pyg-Guanaco-GGML-q3_K_M.bin", - "sizeBytes": 6249231488, - "quantization": "Q3_K_M", - "format": "ggml", - "sha256checksum": "ea266b52d6080e7df1b8d98ea290d0b4e261dc9cfe5fc9169abcef0c154831e5", + "name": "tiiuae-falcon-40b-instruct-Q2_K.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf/blob/main/tiiuae-falcon-40b-instruct-Q2_K.gguf", + "sizeBytes": 17400024800, + "quantization": "Q2_K", + "format": "gguf", + "sha256checksum": "93551752b87a1b265c43dbbf3b1cdf47d4ff4d92122fa44ea12e3b0c6f7d1981", "publisher": { - "name": "mindrage", - "socialUrl": "https://github.com/mindrages" + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" }, - "respository": "mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML", - "repositoryUrl": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML" + "respository": "maddes8cht/tiiuae-falcon-40b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" }, { - "name": "Manticore-13B-Chat-Pyg-Guanaco-GGML-q5_K_S.bin", - "url": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML/resolve/main/Manticore-13B-Chat-Pyg-Guanaco-GGML-q5_K_S.bin", - "sizeBytes": 8950236288, - "quantization": "Q5_K_S", - "format": "ggml", - "sha256checksum": "00a47eeb6364e1e022a45cfe1232a37fa78e9f040242dba78d34d6df383e32d1", + "name": "tiiuae-falcon-40b-instruct-Q4_K_M.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf/blob/main/tiiuae-falcon-40b-instruct-Q4_K_M.gguf", + "sizeBytes": 25452629728, + "quantization": "Q4_K_M", + "format": "gguf", + "sha256checksum": "ff140febf094721f656568250eb6f288ac8dfa49b5b4483139078761a9515cbb", "publisher": { - "name": "mindrage", - "socialUrl": "https://github.com/mindrages" + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" }, - "respository": "mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML", - "repositoryUrl": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML" - } - ] - } - }, - { - "_descriptorVersion": "0.0.1", - "datePublished": "2023-09-27T16:12:57", - "name": "Mistral 7B Instruct v0.1", - "description": "The Mistral-7B-Instruct-v0.1 is a Large Language Model (LLM) developed by Mistral AI. This LLM is an instruct fine-tuned version of a generative text model, leveraging a variety of publicly available conversation datasets. The model's architecture is based on a transformer model, featuring Grouped-Query Attention, Sliding-Window Attention, and a Byte-fallback BPE tokenizer. To utilize the instruction fine-tuning capabilities, prompts should be enclosed within [INST] and [/INST] tokens. The initial instruction should commence with a beginning-of-sentence id, whereas subsequent instructions should not. The generation process by the assistant will terminate with the end-of-sentence token id. For detailed information about this model, refer to the release blog posts by Mistral AI.", - "author": { - "name": "Mistral AI", - "url": "https://mistral.ai/", - "blurb": "Mistral AI's mission is to spearhead the revolution of open models." - }, - "numParameters": "7B", - "resources": { - "canonicalUrl": "https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1", - "paperUrl": "https://mistral.ai/news/announcing-mistral-7b/", - "downloadUrl": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF" - }, - "trainedFor": "chat", - "arch": "mistral", - "files": { - "highlighted": { - "economical": { - "name": "mistral-7b-instruct-v0.1.Q4_K_S.gguf" + "respository": "maddes8cht/tiiuae-falcon-40b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" }, - "most_capable": { - "name": "mistral-7b-instruct-v0.1.Q6_K.gguf" - } - }, - "all": [ { - "name": "mistral-7b-instruct-v0.1.Q4_K_S.gguf", - "url": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_K_S.gguf", - "sizeBytes": 4140373664, - "quantization": "Q4_K_S", + "name": "tiiuae-falcon-40b-instruct-Q5_K_M.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf/blob/main/tiiuae-falcon-40b-instruct-Q5_K_M.gguf", + "sizeBytes": 30642294496, + "quantization": "Q5_K_M", "format": "gguf", - "sha256checksum": "f1b7f1885029080be49aff49c83f87333449ef727089546e0d887e2f17f0d02e", + "sha256checksum": "fe39d1fc51d35039da996e3d2ed262617d1a70b3d9d0a7762571c749103cba39", "publisher": { - "name": "TheBloke", - "socialUrl": "https://twitter.com/TheBlokeAI" + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" }, - "respository": "TheBloke/Mistral-7B-Instruct-v0.1-GGUF", - "repositoryUrl": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF" + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" }, { - "name": "mistral-7b-instruct-v0.1.Q6_K.gguf", - "url": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q6_K.gguf", - "sizeBytes": 5942064800, + "name": "tiiuae-falcon-40b-instruct-Q6_K.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q6_K.gguf", + "sizeBytes": 34456456928, "quantization": "Q6_K", "format": "gguf", - "sha256checksum": "dfb053cb8d5f56abde8f56899ffe0d23e1285a423df0b65ea3f3adbb263b22c2", + "sha256checksum": "5c2a9c95dfcb72d34acb09ac304625847a1f3de307984f559e71f7414345a4e9", "publisher": { - "name": "TheBloke", - "socialUrl": "https://twitter.com/TheBlokeAI" + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" }, - "respository": "TheBloke/Mistral-7B-Instruct-v0.1-GGUF", - "repositoryUrl": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF" + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-07-23T12:47:44.000Z", - "name": "Dolphin-Llama-13B", - "description": "This model, which is based on llama1, is designed specifically for non-commercial use and has undergone a thorough training process using an open-source implementation of Microsoft's Orca dataset. The utilized dataset, post rigorous cleaning and deduplication procedures, encompasses 842,610 instructional instances augmented with GPT-4 completions and an additional 2,625,353 instructional instances augmented with GPT-3.5 completions. The unique aspect of this model is its uncensored nature, having been meticulously filtered to remove any instances of alignment and bias. However, it is advised for users to implement their own alignment layer before deploying this model as a service. The model was trained using the comprehensive flan5m (gpt3.5 completions) dataset for 3 epochs and the flan1m (gpt4 completions) dataset for 2.5 epochs, with distinct learning rates set to preemptively avoid overfitting. This extensive training took approximately 600 hours on a setup of 8x H100s.", + "datePublished": "2023-10-31T16:01:50", + "name": "falcon-7b-instruct", + "description": "Falcon-7B-Instruct, based on Falcon-7B, has been fine-tuned on chat and instruct datasets. It offers outstanding performance and is an excellent choice for chat and instruct applications. Falcon-7B-Instruct is part of the Falcon family of language models, known for their exceptional capabilities and openness.", "author": { - "name": "Eric Hartford", - "url": "https://twitter.com/erhartford", - "blurb": "Eric Hartford is a software engineer and entrepreneur. He specializes in alignment, uncensored models, intersection of AI and society, curating datasets and training chat and instruction tuned models." + "name": "Maddes8cht", + "url": "https://huggingface.co/maddes8cht", + "blurb": "Maddes8cht Passionate about Open Source and AI. On Hugging Face he is advocating for real open source AI models with OSI compliant licenses" }, "numParameters": "7B", "resources": { - "canonicalUrl": "https://huggingface.co/ehartford/dolphin-llama-13b", - "downloadUrl": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML" + "canonicalUrl": "https://huggingface.co/tiiuae/falcon-7b-instruct", + "downloadUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" }, "trainedFor": "chat", - "arch": "llama", + "arch": "falcon", "files": { "highlighted": { "economical": { - "name": "dolphin-llama-13b.ggmlv3.q4_K_S.bin" + "name": "tiiuae-falcon-7b-instruct-Q4_K_M" }, "most_capable": { - "name": "dolphin-llama-13b.ggmlv3.q6_K.bin" + "name": "tiiuae-falcon-7b-Q6_K.gguf" } }, "all": [ { - "name": "dolphin-llama-13b.ggmlv3.q4_K_S.bin", - "url": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML/resolve/main/dolphin-llama-13b.ggmlv3.q4_K_S.bin", - "sizeBytes": 3791725184, - "quantization": "Q4_K_S", - "format": "ggml", - "sha256checksum": "3a4e3eff4a0bfa1dc8a0b257ae36abbc532a4e3e09e1e27cd82958ff8addd173", + "name": "tiiuae-falcon-7b-instruct-Q2_K.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q2_K.gguf", + "sizeBytes": 4025162688, + "quantization": "Q2_K", + "format": "gguf", + "sha256checksum": "c0e87b27def08c0a9dc8fa9bfd70cd7bc373d328943530984165ac4b20d2202c", "publisher": { - "name": "TheBloke", - "socialUrl": "https://twitter.com/TheBlokeAI" + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" }, - "respository": "TheBloke/Dolphin-Llama-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML" + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" }, { - "name": "dolphin-llama-13b.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML/resolve/main/dolphin-llama-13b.ggmlv3.q6_K.bin", - "sizeBytes": 10678850688, + "name": "tiiuae-falcon-7b-instruct-Q4_K_M.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q4_K_M.gguf", + "sizeBytes": 4975385792, + "quantization": "Q4_K_M", + "format": "gguf", + "sha256checksum": "6f6c886ed07d4f6133a4dd7f8b799764ed719c50973d96816b2ac5ff3ada9913", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" + }, + { + "name": "tiiuae-falcon-7b-instruct-Q5_K_M.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q5_K_M.gguf", + "sizeBytes": 5731380160, + "quantization": "Q5_K_M", + "format": "gguf", + "sha256checksum": "3a4666a200024584f667b4dcf2761463046924f9b0708ba13315831416d8548a", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" + }, + { + "name": "tiiuae-falcon-7b-instruct-Q6_K.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q6_K.gguf", + "sizeBytes": 7031618496, "quantization": "Q6_K", - "format": "ggml", - "sha256checksum": "8258f8c014a939ce93519bca658eb714202dacbee8f5df05b2dee0f02472460d", + "format": "gguf", + "sha256checksum": "e6b23eb4193b2d8c7effb4870d2db7a3faa6ed2c8f2c075f7ced51591938818a", "publisher": { - "name": "TheBloke", - "socialUrl": "https://twitter.com/TheBlokeAI" + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" }, - "respository": "TheBloke/Dolphin-Llama-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Dolphin-Llama-13B-GGML" + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" } ] } @@ -411,243 +407,359 @@ }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-05-04T23:15:35.000Z", - "name": "MPT-7B-StoryWriter", - "description": "MPT-7B-StoryWriter-65k+ is a model designed to read and write fictional stories with super long context lengths. It was built by finetuning MPT-7B with a context length of 65k tokens on a filtered fiction subset of the books3 dataset. At inference time, thanks to ALiBi, MPT-7B-StoryWriter-65k+ can extrapolate even beyond 65k tokens. We demonstrate generations as long as 84k tokens on a single node of 8 A100-80GB GPUs in (https://www.mosaicml.com/blog/mpt-7b). This model was trained by MosaicML and follows a modified decoder-only transformer architecture. License: Apache 2.0", + "datePublished": "2023-07-18T21:08:14.000Z", + "name": "Llama-2-7B-Chat-GGML", + "description": "This is the 7B model from the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Meta's fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. Llama-2-Chat models outperform open-source chat models on most benchmarks we tested, and in Meta's human evaluations for helpfulness and safety, are on par with some popular closed-source models like ChatGPT and PaLM.", "author": { - "name": "MosaicML", - "url": "https://www.mosaicml.com/", - "blurb": "MosaicML is the generative AI platform that empowers enterprises to build their own AI. Our research and engineering teams use cutting-edge scientific research to develop products that make it fast, cost-effective, and easy to train today's most popular machine learning models. MosaicML enables developers to maintain full control over the AI models they build, with model ownership and data privacy built into the platform's design." + "name": "Meta AI", + "url": "https://ai.meta.com", + "blurb": "Pushing the boundaries of AI through research, infrastructure and product innovation." }, "numParameters": "7B", "resources": { - "canonicalUrl": "https://huggingface.co/mosaicml/mpt-7b-storywriter", - "downloadUrl": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML" + "canonicalUrl": "https://huggingface.co/meta-llama/Llama-2-7b-chat-hf", + "paperUrl": "https://ai.meta.com/research/publications/llama-2-open-foundation-and-fine-tuned-chat-models/", + "downloadUrl": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML" }, - "trainedFor": "other", - "arch": "mpt", + "trainedFor": "chat", + "arch": "llama", "files": { "highlighted": { "economical": { - "name": "mpt-7b-storywriter.ggmlv3.q4_0.bin" + "name": "llama-2-7b-chat.ggmlv3.q4_K_S.bin" }, "most_capable": { - "name": "mpt-7b-storywriter.ggmlv3.q5_1.bin" + "name": "llama-2-7b-chat.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "mpt-7b-storywriter.ggmlv3.q4_0.bin", - "url": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML/resolve/main/mpt-7b-storywriter.ggmlv3.q4_0.bin", - "sizeBytes": 3741665280, - "quantization": "q4_0", + "name": "llama-2-7b-chat.ggmlv3.q4_K_S.bin", + "url": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/resolve/main/llama-2-7b-chat.ggmlv3.q4_K_S.bin", + "sizeBytes": 3825517184, + "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "357a536464982987e49fb2660fe3f3f53226eaa047f42b31f04d21629aab94fb", + "sha256checksum": "32b758bf5e4f16fb5944b75d577fbca18c11c57000b41c6cc04bb281632d58f3", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/MPT-7B-Storywriter-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML" + "respository": "TheBloke/Llama-2-7B-Chat-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML" }, { - "name": "mpt-7b-storywriter.ggmlv3.q5_1.bin", - "url": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML/resolve/main/mpt-7b-storywriter.ggmlv3.q5_1.bin", - "sizeBytes": 4988356608, - "quantization": "q5_1", + "name": "llama-2-7b-chat.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/resolve/main/llama-2-7b-chat.ggmlv3.q6_K.bin", + "sizeBytes": 5528904320, + "quantization": "Q6_K", "format": "ggml", - "sha256checksum": "3b7dd7aa7508cc8cb4e262fe4b93214826f38d18d04059075e05837457f54025", + "sha256checksum": "24a2097aba9bc63395654515618fb2ceeaea64452147ee5299990b636e4c00ce", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/MPT-7B-Storywriter-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML" + "respository": "TheBloke/Llama-2-7B-Chat-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-07-27T07:00:55.000Z", - "name": "Stable Beluga 13B", - "description": "Stable Beluga 13B is a language model developed by Stability AI, derived from the Llama2 13B model and tailored for English language tasks. It has undergone fine-tuning on a proprietary Orca-style dataset. The training procedure employs a supervised learning approach, with the model trained in mixed-precision (BF16) and optimized using the AdamW optimizer. To utilize Stable Beluga 13B, a specific interaction pattern involving a system prompt (\"### System:\"), user input (\"### User:\"), and the model's response is followed (\"### Assistant\"). This language model operates under a non-commercial community license. More details about the license can be found in the Hugging Face model card.", + "datePublished": "2023-06-12T11:16:39.000Z", + "name": "Manticore 13B Chat Pyg Guanaco", + "description": "The model, augmented with Guanaco qLora, shows broad capabilities compared to other models like Wizard or Manticore. It excels in in-context learning and reasoning but may have weaknesses in coding. The model follows instructions, works as a chatbot, and produces intelligent responses. It accepts various prompting styles, including the ###-Variant. The model is generally unrestricted and doesn't berate users. Recommended settings include low temperature, low diversity, and slight repetition penalty.", "author": { - "name": "StabilityAI", - "url": "https://stability.ai/", - "blurb": "Stability's goal is to maximize the accessibility of modern AI to inspire global creativity and innovation." + "name": "Open Access AI Collective", + "url": "https://huggingface.co/openaccess-ai-collective/", + "blurb": "" }, "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/stabilityai/StableBeluga-13B", - "downloadUrl": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML" + "canonicalUrl": "https://huggingface.co/openaccess-ai-collective/manticore-13b-chat-pyg", + "downloadUrl": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "stablebeluga-13b.ggmlv3.q4_K_S.bin" + "name": "Manticore-13B-Chat-Pyg-Guanaco-GGML-q3_K_M.bin" }, "most_capable": { - "name": "stablebeluga-13b.ggmlv3.q6_K.bin" + "name": "Manticore-13B-Chat-Pyg-Guanaco-GGML-q5_K_S.bin" } }, "all": [ { - "name": "stablebeluga-13b.ggmlv3.q4_K_S.bin", - "url": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML/resolve/main/stablebeluga-13b.ggmlv3.q4_K_S.bin", - "sizeBytes": 7365545088, - "quantization": "Q4_K_S", + "name": "Manticore-13B-Chat-Pyg-Guanaco-GGML-q3_K_M.bin", + "url": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML/resolve/main/Manticore-13B-Chat-Pyg-Guanaco-GGML-q3_K_M.bin", + "sizeBytes": 6249231488, + "quantization": "Q3_K_M", "format": "ggml", - "sha256checksum": "89262c80eb0899fbcad0538d4e10efe82a7e0d7d0371b6025d487d66a264338e", + "sha256checksum": "ea266b52d6080e7df1b8d98ea290d0b4e261dc9cfe5fc9169abcef0c154831e5", "publisher": { - "name": "TheBloke", - "socialUrl": "https://twitter.com/TheBlokeAI" + "name": "mindrage", + "socialUrl": "https://github.com/mindrages" }, - "respository": "TheBloke/StableBeluga-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML" + "respository": "mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML", + "repositoryUrl": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML" }, { - "name": "stablebeluga-13b.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML/resolve/main/stablebeluga-13b.ggmlv3.q6_K.bin", - "sizeBytes": 10678850688, - "quantization": "Q6_K", + "name": "Manticore-13B-Chat-Pyg-Guanaco-GGML-q5_K_S.bin", + "url": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML/resolve/main/Manticore-13B-Chat-Pyg-Guanaco-GGML-q5_K_S.bin", + "sizeBytes": 8950236288, + "quantization": "Q5_K_S", "format": "ggml", - "sha256checksum": "59e1f74d30bdce1995c9c94901e75bc32e34e31dd8e8f1ab1cda6ea0f18bb54d", + "sha256checksum": "00a47eeb6364e1e022a45cfe1232a37fa78e9f040242dba78d34d6df383e32d1", "publisher": { - "name": "TheBloke", - "socialUrl": "https://twitter.com/TheBlokeAI" + "name": "mindrage", + "socialUrl": "https://github.com/mindrages" }, - "respository": "TheBloke/StableBeluga-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML" + "respository": "mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML", + "repositoryUrl": "https://huggingface.co/mindrage/Manticore-13B-Chat-Pyg-Guanaco-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-08-07T21:51:07", - "name": "Vicuna 13B v1.5 16K", - "description": "Vicuna is a state-of-the-art chat assistant developed by LMSYS, leveraging the transformative power of an auto-regressive language model based on the transformer architecture. Trained via fine-tuning the Llama 2 model, Vicuna is used primarily for research in large language models and chatbots, catering to the needs of AI enthusiasts, machine learning researchers, and hobbyists. The model operates under the Llama 2 Community License Agreement. The latest version, Vicuna v1.5 (16k), has been fine-tuned using supervised instruction and linear RoPE scaling, with training data comprising around 125K conversations collected from ShareGPT.com, packed into sequences of 16K tokens each. A comprehensive explanation of the training details can be found in the appendix to the linked paper titled \"Training Details of Vicuna Models.\"", + "datePublished": "2023-05-19T11:16:39.000Z", + "name": "Manticore 13B", + "description": "Manticore 13B is a refined version of the Llama 13B model, having been fine-tuned on a variety of datasets. These include ShareGPT, which is based on a cleaned and de-duplicated subset, WizardLM, and Wizard-Vicuna. It also incorporates a subset of QingyiSi/Alpaca-CoT, specifically for roleplay and CoT. Other datasets used in the fine-tuning process are GPT4-LLM-Cleaned and GPTeacher-General-Instruct. The model also utilizes ARC-Easy & ARC-Challenge, both of which have been augmented for detailed responses. The mmlu dataset, also augmented for detailed responses, includes subsets such as abstract_algebra, conceptual_physics, formal_logic, high_school_physics, and logical_fallacies. A 5K row subset of hellaswag has been used for instructing the model to generate concise responses. Additionally, metaeval/ScienceQA_text_only has been used for concise response instruction, and openai/summarize_from_feedback has been used for tl;dr summarization instruction.", "author": { - "name": "LMSYS Org", - "url": "https://lmsys.org/", - "blurb": "Large Model Systems Organization (LMSYS Org) is an open research organization founded by students and faculty from UC Berkeley in collaboration with UCSD and CMU." + "name": "Open Access AI Collective", + "url": "https://huggingface.co/openaccess-ai-collective/", + "blurb": "" }, "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/lmsys/vicuna-13b-v1.5-16k", - "downloadUrl": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML", - "paperUrl": "https://arxiv.org/abs/2306.05685" + "canonicalUrl": "https://huggingface.co/openaccess-ai-collective/manticore-13b", + "downloadUrl": "https://huggingface.co/TheBloke/Manticore-13B-GGML" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "vicuna-13b-v1.5-16k.ggmlv3.q4_K_S.bin" + "name": "Manticore-13B.ggmlv3.q4_K_S.bin" }, "most_capable": { - "name": "vicuna-13b-v1.5-16k.ggmlv3.q6_K.bin" + "name": "Manticore-13B.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "vicuna-13b-v1.5-16k.ggmlv3.q4_K_S.bin", - "url": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML/resolve/main/vicuna-13b-v1.5-16k.ggmlv3.q4_K_S.bin", - "sizeBytes": 7365545088, + "name": "Manticore-13B.ggmlv3.q4_K_S.bin", + "url": "https://huggingface.co/TheBloke/Manticore-13B-GGML/resolve/main/Manticore-13B.ggmlv3.q4_K_S.bin", + "sizeBytes": 7323305088, "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "c343200172b677488a7e59a2ef74642b63b2578fde3c976911751181d29ce201", + "sha256checksum": "84599645aeda2cd7c97a3a59f3210fb3e559cb72b4f3c5d5288924fa9e80b737", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/vicuna-13B-v1.5-16K-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML" + "respository": "TheBloke/Manticore-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Manticore-13B-GGML" }, { - "name": "vicuna-13b-v1.5-16k.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML/resolve/main/vicuna-13b-v1.5-16k.ggmlv3.q6_K.bin", + "name": "Manticore-13B.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/Manticore-13B-GGML/resolve/main/Manticore-13B.ggmlv3.q6_K.bin", "sizeBytes": 10678850688, "quantization": "Q6_K", "format": "ggml", - "sha256checksum": "949e2003a47552d2931ebce29b94ad59482561a36429fb1c00d13db7feb87fd9", + "sha256checksum": "1be08ec3dcfbe7c28bf524061cd65fa5a5b7dc4525dee99a0f2297a23a77778e", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/vicuna-13B-v1.5-16K-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML" + "respository": "TheBloke/Manticore-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Manticore-13B-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-10-26T11:25:50", - "name": "Zephyr 7B \u03b2", - "description": "The Zephyr-7B-\u03b2 is the second model in the Zephyr series, designed to function as an assistant. It is a fine-tuned version of the mistralai/Mistral-7B-v0.1 model, leveraging a 7B parameter GPT-like architecture. The model has been trained on a combination of synthetic datasets and publicly available data using Direct Preference Optimization (DPO), a technique that improved its performance on the MT Bench. An important aspect to note is that the in-built alignment of the training datasets was deliberately omitted during the training process, a decision that, while enhancing the model's helpfulness, also makes it prone to generating potentially problematic outputs when prompted. Therefore, it is advised to use the model strictly for research and educational purposes. The model primarily supports the English language and is licensed under the MIT License. Additional details can be found in the associated technical report.", + "datePublished": "2023-09-27T16:12:57", + "name": "Mistral 7B Instruct v0.1", + "description": "The Mistral-7B-Instruct-v0.1 is a Large Language Model (LLM) developed by Mistral AI. This LLM is an instruct fine-tuned version of a generative text model, leveraging a variety of publicly available conversation datasets. The model's architecture is based on a transformer model, featuring Grouped-Query Attention, Sliding-Window Attention, and a Byte-fallback BPE tokenizer. To utilize the instruction fine-tuning capabilities, prompts should be enclosed within [INST] and [/INST] tokens. The initial instruction should commence with a beginning-of-sentence id, whereas subsequent instructions should not. The generation process by the assistant will terminate with the end-of-sentence token id. For detailed information about this model, refer to the release blog posts by Mistral AI.", "author": { - "name": "Hugging Face H4", - "url": "https://huggingface.co/HuggingFaceH4", - "blurb": "Hugging Face H4 team, focused on aligning language models to be helpful, honest, harmless, and huggy \ud83e\udd17" + "name": "Mistral AI", + "url": "https://mistral.ai/", + "blurb": "Mistral AI's mission is to spearhead the revolution of open models." }, "numParameters": "7B", "resources": { - "canonicalUrl": "https://huggingface.co/HuggingFaceH4/zephyr-7b-beta", - "paperUrl": "https://arxiv.org/abs/2310.16944", - "downloadUrl": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF" + "canonicalUrl": "https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1", + "paperUrl": "https://mistral.ai/news/announcing-mistral-7b/", + "downloadUrl": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF" }, "trainedFor": "chat", "arch": "mistral", "files": { "highlighted": { "economical": { - "name": "zephyr-7b-beta.Q4_K_S.gguf" + "name": "mistral-7b-instruct-v0.1.Q4_K_S.gguf" }, "most_capable": { - "name": "zephyr-7b-beta.Q6_K.gguf" + "name": "mistral-7b-instruct-v0.1.Q6_K.gguf" } }, "all": [ { - "name": "zephyr-7b-beta.Q4_K_S.gguf", - "url": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF/resolve/main/zephyr-7b-beta.Q4_K_S.gguf", - "sizeBytes": 4140373696, + "name": "mistral-7b-instruct-v0.1.Q4_K_S.gguf", + "url": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_K_S.gguf", + "sizeBytes": 4140373664, "quantization": "Q4_K_S", "format": "gguf", - "sha256checksum": "cafa0b85b2efc15ca33023f3b87f8d0c44ddcace16b3fb608280e0eb8f425cb1", + "sha256checksum": "f1b7f1885029080be49aff49c83f87333449ef727089546e0d887e2f17f0d02e", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/zephyr-7B-beta-GGUF", - "repositoryUrl": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF" + "respository": "TheBloke/Mistral-7B-Instruct-v0.1-GGUF", + "repositoryUrl": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF" }, { - "name": "zephyr-7b-beta.Q6_K.gguf", - "url": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF/resolve/main/zephyr-7b-beta.Q6_K.gguf", - "sizeBytes": 5942064832, + "name": "mistral-7b-instruct-v0.1.Q6_K.gguf", + "url": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q6_K.gguf", + "sizeBytes": 5942064800, "quantization": "Q6_K", "format": "gguf", - "sha256checksum": "39b52e291eea6040de078283ee5316ff2a317e2b6f59be56724d9b29bada6cfe", + "sha256checksum": "dfb053cb8d5f56abde8f56899ffe0d23e1285a423df0b65ea3f3adbb263b22c2", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/zephyr-7B-beta-GGUF", - "repositoryUrl": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF" + "respository": "TheBloke/Mistral-7B-Instruct-v0.1-GGUF", + "repositoryUrl": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-07-20T09:26:31.000Z", - "name": "Redmond-Puffin-13B-V1.3", - "description": "Redmond-Puffin-13B is one of the worlds first llama-2 based, fine-tuned language models, leveraging a hand curated set of 3K high quality examples, many of which take full advantage of the 4096 context length of Llama 2. This model was trained for multiple epochs on a dataset of curated GPT-4 examples, most of which are long context conversations between a real human and GPT-4. Additional data came from carefully curated sub sections of datasets such as CamelAI's Physics, Chemistry, Biology and Math. Puffin 13B v1.3 was fine-tuned by Nous Research, with LDJ leading the training and dataset curation, along with significant dataset formation contributions by J-Supha.", + "datePublished": "2023-05-04T23:15:35.000Z", + "name": "MPT-7B-StoryWriter", + "description": "MPT-7B-StoryWriter-65k+ is a model designed to read and write fictional stories with super long context lengths. It was built by finetuning MPT-7B with a context length of 65k tokens on a filtered fiction subset of the books3 dataset. At inference time, thanks to ALiBi, MPT-7B-StoryWriter-65k+ can extrapolate even beyond 65k tokens. We demonstrate generations as long as 84k tokens on a single node of 8 A100-80GB GPUs in (https://www.mosaicml.com/blog/mpt-7b). This model was trained by MosaicML and follows a modified decoder-only transformer architecture. License: Apache 2.0", + "author": { + "name": "MosaicML", + "url": "https://www.mosaicml.com/", + "blurb": "MosaicML is the generative AI platform that empowers enterprises to build their own AI. Our research and engineering teams use cutting-edge scientific research to develop products that make it fast, cost-effective, and easy to train today's most popular machine learning models. MosaicML enables developers to maintain full control over the AI models they build, with model ownership and data privacy built into the platform's design." + }, + "numParameters": "7B", + "resources": { + "canonicalUrl": "https://huggingface.co/mosaicml/mpt-7b-storywriter", + "downloadUrl": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML" + }, + "trainedFor": "other", + "arch": "mpt", + "files": { + "highlighted": { + "economical": { + "name": "mpt-7b-storywriter.ggmlv3.q4_0.bin" + }, + "most_capable": { + "name": "mpt-7b-storywriter.ggmlv3.q5_1.bin" + } + }, + "all": [ + { + "name": "mpt-7b-storywriter.ggmlv3.q4_0.bin", + "url": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML/resolve/main/mpt-7b-storywriter.ggmlv3.q4_0.bin", + "sizeBytes": 3741665280, + "quantization": "q4_0", + "format": "ggml", + "sha256checksum": "357a536464982987e49fb2660fe3f3f53226eaa047f42b31f04d21629aab94fb", + "publisher": { + "name": "TheBloke", + "socialUrl": "https://twitter.com/TheBlokeAI" + }, + "respository": "TheBloke/MPT-7B-Storywriter-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML" + }, + { + "name": "mpt-7b-storywriter.ggmlv3.q5_1.bin", + "url": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML/resolve/main/mpt-7b-storywriter.ggmlv3.q5_1.bin", + "sizeBytes": 4988356608, + "quantization": "q5_1", + "format": "ggml", + "sha256checksum": "3b7dd7aa7508cc8cb4e262fe4b93214826f38d18d04059075e05837457f54025", + "publisher": { + "name": "TheBloke", + "socialUrl": "https://twitter.com/TheBlokeAI" + }, + "respository": "TheBloke/MPT-7B-Storywriter-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML" + } + ] + } + }, + { + "_descriptorVersion": "0.0.1", + "datePublished": "2023-08-10T20:51:49", + "name": "MythoMax L2 13B", + "description": "MythoMax-L2 is a creative writing model that builds upon Llama2, MythoLogic-L2, and Huginn. It can craft captivating texts for various domains and audiences, using MythoLogic-L2\u00e2\u20ac\u2122s vast knowledge base and Huginn\u00e2\u20ac\u2122s expressive language skills. Specialized at creative writing, it is also proficient at roleplaying, and is able to adapt to different contexts and characters with flexibility and realism. This model is an experimental merge of several other models (MythoMix, Mythologic-L2, and Huggin).", + "author": { + "name": "Gryphe Padar", + "url": "https://huggingface.co/Gryphe", + "blurb": "Author behind various Mytho merges, most prominent being MythoMax, MythoLogic, and MythoBoros" + }, + "numParameters": "13B", + "resources": { + "canonicalUrl": "https://huggingface.co/Gryphe/MythoMax-L2-13b", + "downloadUrl": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML" + }, + "trainedFor": "chat", + "arch": "llama", + "files": { + "highlighted": { + "economical": { + "name": "mythomax-l2-13b.ggmlv3.q4_K_S.bin" + }, + "most_capable": { + "name": "mythomax-l2-13b.ggmlv3.q6_K.bin" + } + }, + "all": [ + { + "name": "mythomax-l2-13b.ggmlv3.q4_K_S.bin", + "url": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML/resolve/main/mythomax-l2-13b.ggmlv3.q4_K_S.bin", + "sizeBytes": 7365545088, + "quantization": "Q4_K_S", + "format": "ggml", + "sha256checksum": "7b71bd39663bfde58c374319e61dbf446fefcd68393f175a628fba8d0361aec7", + "publisher": { + "name": "TheBloke", + "socialUrl": "https://twitter.com/TheBlokeAI" + }, + "respository": "TheBloke/MythoMax-L2-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML" + }, + { + "name": "mythomax-l2-13b.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML/resolve/main/mythomax-l2-13b.ggmlv3.q6_K.bin", + "sizeBytes": 10678850688, + "quantization": "Q6_K", + "format": "ggml", + "sha256checksum": "37a1ea28a3a08721d2956a8bb87a06127c6036742267174b79d9c7b83133dba0", + "publisher": { + "name": "TheBloke", + "socialUrl": "https://twitter.com/TheBlokeAI" + }, + "respository": "TheBloke/MythoMax-L2-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML" + } + ] + } + }, + { + "_descriptorVersion": "0.0.1", + "datePublished": "2023-06-03T05:34:04.000Z", + "name": "Nous-Hermes-13b", + "description": "Nous-Hermes-13b is a state-of-the-art language model fine-tuned on over 300,000 instructions. This model was fine-tuned by Nous Research, with Teknium and Karan4D leading the fine tuning process and dataset curation, Redmond AI sponsoring the compute, and several other contributors. The result is an enhanced Llama 13b model that rivals GPT-3.5-turbo in performance across a variety of tasks. This model stands out for its long responses, low hallucination rate, and absence of OpenAI censorship mechanisms. The fine-tuning process was performed with a 2000 sequence length on an 8x a100 80GB DGX machine for over 50 hours.", "author": { "name": "Nous Research", "url": "https://nousresearch.com", @@ -655,206 +767,206 @@ }, "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/NousResearch/Redmond-Puffin-13B", - "downloadUrl": "https://huggingface.co/NousResearch/Redmond-Puffin-13B-GGML" + "canonicalUrl": "https://huggingface.co/NousResearch/Nous-Hermes-13b", + "downloadUrl": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "GGML-Redmond-Puffin-v1.3-13B-Q4_K_M.bin" + "name": "nous-hermes-13b.ggmlv3.q3_K_S.bin" + }, + "most_capable": { + "name": "nous-hermes-13b.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "GGML-Redmond-Puffin-v1.3-13B-Q4_K_M.bin", - "url": "https://huggingface.co/NousResearch/Redmond-Puffin-13B-GGML/resolve/main/GGML-Redmond-Puffin-v1.3-13B-Q4_K_M.bin", - "sizeBytes": 8059366944, - "quantization": "Q4_K_M", + "name": "nous-hermes-13b.ggmlv3.q3_K_S.bin", + "url": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML/resolve/main/nous-hermes-13b.ggmlv3.q3_K_S.bin", + "sizeBytes": 5594695104, + "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "7f48f820d47d91b43fe46f96a54529c64a5292a60f625deba8b6b11194e84a98", + "sha256checksum": "591a49f1ef4dbc2cf43943c5ec9bd617e6086264e30f66718bc764bc55286b5e", "publisher": { - "name": "Nous Research", - "socialUrl": "https://nousresearch.com" + "name": "TheBloke", + "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "NousResearch/Redmond-Puffin-13B-GGML", - "repositoryUrl": "https://huggingface.co/NousResearch/Redmond-Puffin-13B-GGML" + "respository": "TheBloke/Nous-Hermes-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML" + }, + { + "name": "nous-hermes-13b.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML/resolve/main/nous-hermes-13b.ggmlv3.q6_K.bin", + "sizeBytes": 10678859104, + "quantization": "Q6_K", + "format": "ggml", + "sha256checksum": "efe8ffe14aa97c3c5f45f2bc8e80a02933c5e907813deb685c93341bf671f44e", + "publisher": { + "name": "TheBloke", + "socialUrl": "https://twitter.com/TheBlokeAI" + }, + "respository": "TheBloke/Nous-Hermes-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-06-25T11:37:35.000Z", - "name": "Airboros-13b-gpt4-1.4", - "description": "This qlora fine-tuned 13b parameter LlaMa model uses synthetic training data created via github.com/jondurbin/airoboros. It extends version 1.1, introducing thousands of new training data and an update for \"PLAINFORMAT\" to print code without backticks or explanations. The dataset, available online, focuses on coding, math/reasoning, trivia, role playing, multiple choice, fill-in-the-blank, context-obedient question answering, theory of mind, and general topics. The model was fine-tuned with a qlora fork, updated to use a modified vicuna template compatible with 7b/13b versions. The format involves a preamble/system prompt, followed by \"USER: [prompt] ASSISTANT: \", with the prompt allowing for multiple lines and spaces.", + "datePublished": "2023-07-22T02:20:14.000Z", + "name": "Nous-Hermes-Llama2-13b", + "description": "The Nous-Hermes-Llama2-13b is a state-of-the-art language model developed by Nous Research in collaboration with Teknium, Emozilla, and Redmond AI, fine-tuned on over 300,000 instructions. It uses the same dataset as its predecessor, Hermes on Llama-1, to maintain consistency while enhancing capability. This model is known for generating long responses with a lower hallucination rate, free from OpenAI censorship mechanisms. It was fine-tuned using a sequence length of 4096 on an 8x a100 80GB DGX machine. The training dataset is composed of synthetic GPT-4 outputs, including data from diverse sources such as GPTeacher, roleplay v1&2, code instruct datasets, and unpublished Nous Instruct & PDACTL, among others. The model utilizes the Alpaca prompt format for instructions and responses.", "author": { - "name": "Jon Durbin", - "url": "https://github.com/jondurbin", - "blurb": "Jon Durbin is a Computer Scientist and the author of the Airboros (7B, 13B, 33B, 65B) qlora fine-tuned LlaMa family of models." + "name": "Nous Research", + "url": "https://nousresearch.com", + "blurb": "Nous Research is dedicated to advancing the field of natural language processing, in collaboration with the open-source community, through bleeding-edge research and a commitment to symbiotic development." }, "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/jondurbin/airoboros-13b-gpt4-1.4", - "downloadUrl": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML" + "canonicalUrl": "https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b", + "downloadUrl": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "airoboros-13b-gpt4-1.4.ggmlv3.q4_K_S.bin" + "name": "nous-hermes-llama2-13b.ggmlv3.q3_K_S.bin" }, "most_capable": { - "name": "airoboros-13b-gpt4-1.4.ggmlv3.q6_K.bin" + "name": "nous-hermes-llama2-13b.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "airoboros-13b-gpt4-1.4.ggmlv3.q4_K_S.bin", - "url": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML/resolve/main/airoboros-13b-gpt4-1.4.ggmlv3.q4_K_S.bin", - "sizeBytes": 7365545088, + "name": "nous-hermes-llama2-13b.ggmlv3.q3_K_S.bin", + "url": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML/resolve/main/nous-hermes-llama2-13b.ggmlv3.q3_K_S.bin", + "sizeBytes": 5874151072, "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "dd5d8019e73de1e99089e9d82dde6a08173818dc0afeb1e95c0bb7ec77891eaf", + "sha256checksum": "9e76993f3e8a804b3ee012bc461fbd01e9ff71fbdf8ac9f6f13ccadd5ff51b4e", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/airoboros-13B-gpt4-1.4-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML" + "respository": "TheBloke/Nous-Hermes-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML" }, { - "name": "airoboros-13b-gpt4-1.4.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML/resolve/main/airoboros-13b-gpt4-1.4.ggmlv3.q6_K.bin", - "sizeBytes": 10678850688, + "name": "nous-hermes-llama2-13b.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML/resolve/main/nous-hermes-llama2-13b.ggmlv3.q6_K.bin", + "sizeBytes": 10830311072, "quantization": "Q6_K", "format": "ggml", - "sha256checksum": "5a72a053eb02c9e6e4fa1ee24ed73c89da4877468923309fdec088d3a3fbb5ff", + "sha256checksum": "02f696d0df53174cf9a3aba62d22a1b17882389a5381e68b2c7395845f87d3f5", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/airoboros-13B-gpt4-1.4-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/airoboros-13B-gpt4-1.4-GGML" + "respository": "TheBloke/Nous-Hermes-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-05-19T11:16:39.000Z", - "name": "Manticore 13B", - "description": "Manticore 13B is a refined version of the Llama 13B model, having been fine-tuned on a variety of datasets. These include ShareGPT, which is based on a cleaned and de-duplicated subset, WizardLM, and Wizard-Vicuna. It also incorporates a subset of QingyiSi/Alpaca-CoT, specifically for roleplay and CoT. Other datasets used in the fine-tuning process are GPT4-LLM-Cleaned and GPTeacher-General-Instruct. The model also utilizes ARC-Easy & ARC-Challenge, both of which have been augmented for detailed responses. The mmlu dataset, also augmented for detailed responses, includes subsets such as abstract_algebra, conceptual_physics, formal_logic, high_school_physics, and logical_fallacies. A 5K row subset of hellaswag has been used for instructing the model to generate concise responses. Additionally, metaeval/ScienceQA_text_only has been used for concise response instruction, and openai/summarize_from_feedback has been used for tl;dr summarization instruction.", + "datePublished": "2023-08-11T20:24:21", + "name": "OpenOrca-Platypus2-13B", + "description": "OpenOrca-Platypus2-13B is a powerful language model developed by merging garage-bAInd/Platypus2-13B and Open-Orca/OpenOrcaxOpenChat-Preview2-13B. It's an auto-regressive model based on the Lllama 2 transformer architecture, trained predominantly in English. The model was tested via Language Model Evaluation Harness and performed impressively on the HuggingFace Leaderboard with an average score of 64.56 across various shot scenarios. It also exceeded the base Preview2 model's performance on AGIEval and BigBench-Hard evaluations. Cole Hunter & Ariel Lee trained Platypus2-13B using the STEM and logic-based dataset garage-bAInd/Open-Platypus, while Open-Orca trained the OpenOrcaxOpenChat-Preview2-13B model using a refined subset of the GPT-4 data from the OpenOrca dataset. The training licenses for the base weights differ, with Platypus2-13B operating under a Non-Commercial Creative Commons license (CC BY-NC-4.0) and OpenOrcaxOpenChat-Preview2-13B under a Llama 2 Commercial license. Detailed training procedures and benchmark results are available in the Platypus GitHub repo.", "author": { - "name": "Open Access AI Collective", - "url": "https://huggingface.co/openaccess-ai-collective/", - "blurb": "" + "name": "Open Orca", + "url": "https://huggingface.co/Open-Orca", + "blurb": "A collaboration between Alignment Lab AI and the Platypus / garage-bAInd team" }, "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/openaccess-ai-collective/manticore-13b", - "downloadUrl": "https://huggingface.co/TheBloke/Manticore-13B-GGML" + "canonicalUrl": "https://huggingface.co/Open-Orca/OpenOrca-Platypus2-13B", + "downloadUrl": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML", + "paperUrl": "https://arxiv.org/abs/2308.07317" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "Manticore-13B.ggmlv3.q4_K_S.bin" + "name": "openorca-platypus2-13b.ggmlv3.q4_K_S.bin" }, "most_capable": { - "name": "Manticore-13B.ggmlv3.q6_K.bin" + "name": "openorca-platypus2-13b.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "Manticore-13B.ggmlv3.q4_K_S.bin", - "url": "https://huggingface.co/TheBloke/Manticore-13B-GGML/resolve/main/Manticore-13B.ggmlv3.q4_K_S.bin", - "sizeBytes": 7323305088, + "name": "openorca-platypus2-13b.ggmlv3.q4_K_S.bin", + "url": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML/resolve/main/openorca-platypus2-13b.ggmlv3.q4_K_S.bin", + "sizeBytes": 7558851360, "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "84599645aeda2cd7c97a3a59f3210fb3e559cb72b4f3c5d5288924fa9e80b737", + "sha256checksum": "55318c58c22c4e4a8dd5504919cefde5f843e848b6c96cd06a3ad69f2f41365b", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/Manticore-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Manticore-13B-GGML" + "respository": "TheBloke/OpenOrca-Platypus2-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML" }, { - "name": "Manticore-13B.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/Manticore-13B-GGML/resolve/main/Manticore-13B.ggmlv3.q6_K.bin", - "sizeBytes": 10678850688, + "name": "openorca-platypus2-13b.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML/resolve/main/openorca-platypus2-13b.ggmlv3.q6_K.bin", + "sizeBytes": 10829916960, "quantization": "Q6_K", "format": "ggml", - "sha256checksum": "1be08ec3dcfbe7c28bf524061cd65fa5a5b7dc4525dee99a0f2297a23a77778e", + "sha256checksum": "0487eb1a91683020cb8dc59b096b9f7e086735efdf1cf5f5ec84c842ec0a6799", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/Manticore-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Manticore-13B-GGML" + "respository": "TheBloke/OpenOrca-Platypus2-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-06-14T11:50:53.000Z", - "name": "WizardCoder-15B-V1.0", - "description": "WizardCoder: Empowering Code Large Language Models with Evol-Instruct. To develop our WizardCoder model, we begin by adapting the Evol-Instruct method specifically for coding tasks. This involves tailoring the prompt to the domain of code-related instructions. Subsequently, we fine-tune the Code LLM, StarCoder, utilizing the newly created instruction-following training set.", + "datePublished": "2023-07-20T09:26:31.000Z", + "name": "Redmond-Puffin-13B-V1.3", + "description": "Redmond-Puffin-13B is one of the worlds first llama-2 based, fine-tuned language models, leveraging a hand curated set of 3K high quality examples, many of which take full advantage of the 4096 context length of Llama 2. This model was trained for multiple epochs on a dataset of curated GPT-4 examples, most of which are long context conversations between a real human and GPT-4. Additional data came from carefully curated sub sections of datasets such as CamelAI's Physics, Chemistry, Biology and Math. Puffin 13B v1.3 was fine-tuned by Nous Research, with LDJ leading the training and dataset curation, along with significant dataset formation contributions by J-Supha.", "author": { - "name": "WizardLM", - "url": "https://huggingface.co/WizardLM", - "blurb": "WizardLM: An Instruction-following LLM Using Evol-Instruct" + "name": "Nous Research", + "url": "https://nousresearch.com", + "blurb": "Nous Research is dedicated to advancing the field of natural language processing, in collaboration with the open-source community, through bleeding-edge research and a commitment to symbiotic development." }, - "numParameters": "15B", + "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/WizardLM/WizardCoder-15B-V1.0", - "downloadUrl": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML", - "paperUrl": "https://arxiv.org/abs/2306.08568" + "canonicalUrl": "https://huggingface.co/NousResearch/Redmond-Puffin-13B", + "downloadUrl": "https://huggingface.co/NousResearch/Redmond-Puffin-13B-GGML" }, - "trainedFor": "instruct", - "arch": "starcoder", + "trainedFor": "chat", + "arch": "llama", "files": { "highlighted": { "economical": { - "name": "WizardCoder-15B-1.0.ggmlv3.q4_0.bin" - }, - "most_capable": { - "name": "WizardCoder-15B-1.0.ggmlv3.q5_1.bin" + "name": "GGML-Redmond-Puffin-v1.3-13B-Q4_K_M.bin" } }, "all": [ { - "name": "WizardCoder-15B-1.0.ggmlv3.q4_0.bin", - "url": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML/resolve/main/WizardCoder-15B-1.0.ggmlv3.q4_0.bin", - "sizeBytes": 10746570393, - "quantization": "q4_0", - "format": "ggml", - "sha256checksum": "b70164bc0b58a472c0987905133735ab3b27e2c439dedf8174a43951c51c3229", - "publisher": { - "name": "TheBloke", - "socialUrl": "https://twitter.com/TheBlokeAI" - }, - "respository": "TheBloke/WizardCoder-15B-1.0-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML" - }, - { - "name": "WizardCoder-15B-1.0.ggmlv3.q5_1.bin", - "url": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML/resolve/main/WizardCoder-15B-1.0.ggmlv3.q5_1.bin", - "sizeBytes": 14257205145, - "quantization": "q5_1", + "name": "GGML-Redmond-Puffin-v1.3-13B-Q4_K_M.bin", + "url": "https://huggingface.co/NousResearch/Redmond-Puffin-13B-GGML/resolve/main/GGML-Redmond-Puffin-v1.3-13B-Q4_K_M.bin", + "sizeBytes": 8059366944, + "quantization": "Q4_K_M", "format": "ggml", - "sha256checksum": "1219d9fc6d51901d9a1e58e3cb7f03818d02a1d0ab2d070b4cbabdefeb7d0363", + "sha256checksum": "7f48f820d47d91b43fe46f96a54529c64a5292a60f625deba8b6b11194e84a98", "publisher": { - "name": "TheBloke", - "socialUrl": "https://twitter.com/TheBlokeAI" + "name": "Nous Research", + "socialUrl": "https://nousresearch.com" }, - "respository": "TheBloke/WizardCoder-15B-1.0-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML" + "respository": "NousResearch/Redmond-Puffin-13B-GGML", + "repositoryUrl": "https://huggingface.co/NousResearch/Redmond-Puffin-13B-GGML" } ] } @@ -919,175 +1031,235 @@ }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-08-10T20:51:49", - "name": "MythoMax L2 13B", - "description": "MythoMax-L2 is a creative writing model that builds upon Llama2, MythoLogic-L2, and Huginn. It can craft captivating texts for various domains and audiences, using MythoLogic-L2\u2019s vast knowledge base and Huginn\u2019s expressive language skills. Specialized at creative writing, it is also proficient at roleplaying, and is able to adapt to different contexts and characters with flexibility and realism. This model is an experimental merge of several other models (MythoMix, Mythologic-L2, and Huggin).", + "datePublished": "2023-07-27T07:00:55.000Z", + "name": "Stable Beluga 13B", + "description": "Stable Beluga 13B is a language model developed by Stability AI, derived from the Llama2 13B model and tailored for English language tasks. It has undergone fine-tuning on a proprietary Orca-style dataset. The training procedure employs a supervised learning approach, with the model trained in mixed-precision (BF16) and optimized using the AdamW optimizer. To utilize Stable Beluga 13B, a specific interaction pattern involving a system prompt (\"### System:\"), user input (\"### User:\"), and the model's response is followed (\"### Assistant\"). This language model operates under a non-commercial community license. More details about the license can be found in the Hugging Face model card.", "author": { - "name": "Gryphe Padar", - "url": "https://huggingface.co/Gryphe", - "blurb": "Author behind various Mytho merges, most prominent being MythoMax, MythoLogic, and MythoBoros" + "name": "StabilityAI", + "url": "https://stability.ai/", + "blurb": "Stability's goal is to maximize the accessibility of modern AI to inspire global creativity and innovation." }, "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/Gryphe/MythoMax-L2-13b", - "downloadUrl": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML" + "canonicalUrl": "https://huggingface.co/stabilityai/StableBeluga-13B", + "downloadUrl": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "mythomax-l2-13b.ggmlv3.q4_K_S.bin" + "name": "stablebeluga-13b.ggmlv3.q4_K_S.bin" }, "most_capable": { - "name": "mythomax-l2-13b.ggmlv3.q6_K.bin" + "name": "stablebeluga-13b.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "mythomax-l2-13b.ggmlv3.q4_K_S.bin", - "url": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML/resolve/main/mythomax-l2-13b.ggmlv3.q4_K_S.bin", + "name": "stablebeluga-13b.ggmlv3.q4_K_S.bin", + "url": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML/resolve/main/stablebeluga-13b.ggmlv3.q4_K_S.bin", "sizeBytes": 7365545088, "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "7b71bd39663bfde58c374319e61dbf446fefcd68393f175a628fba8d0361aec7", + "sha256checksum": "89262c80eb0899fbcad0538d4e10efe82a7e0d7d0371b6025d487d66a264338e", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/MythoMax-L2-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML" + "respository": "TheBloke/StableBeluga-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML" }, { - "name": "mythomax-l2-13b.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML/resolve/main/mythomax-l2-13b.ggmlv3.q6_K.bin", + "name": "stablebeluga-13b.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML/resolve/main/stablebeluga-13b.ggmlv3.q6_K.bin", "sizeBytes": 10678850688, "quantization": "Q6_K", "format": "ggml", - "sha256checksum": "37a1ea28a3a08721d2956a8bb87a06127c6036742267174b79d9c7b83133dba0", + "sha256checksum": "59e1f74d30bdce1995c9c94901e75bc32e34e31dd8e8f1ab1cda6ea0f18bb54d", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/MythoMax-L2-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/MythoMax-L2-13B-GGML" + "respository": "TheBloke/StableBeluga-13B-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/StableBeluga-13B-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-07-22T02:20:14.000Z", - "name": "Nous-Hermes-Llama2-13b", - "description": "The Nous-Hermes-Llama2-13b is a state-of-the-art language model developed by Nous Research in collaboration with Teknium, Emozilla, and Redmond AI, fine-tuned on over 300,000 instructions. It uses the same dataset as its predecessor, Hermes on Llama-1, to maintain consistency while enhancing capability. This model is known for generating long responses with a lower hallucination rate, free from OpenAI censorship mechanisms. It was fine-tuned using a sequence length of 4096 on an 8x a100 80GB DGX machine. The training dataset is composed of synthetic GPT-4 outputs, including data from diverse sources such as GPTeacher, roleplay v1&2, code instruct datasets, and unpublished Nous Instruct & PDACTL, among others. The model utilizes the Alpaca prompt format for instructions and responses.", + "datePublished": "2023-08-07T21:51:07", + "name": "Vicuna 13B v1.5 16K", + "description": "Vicuna is a state-of-the-art chat assistant developed by LMSYS, leveraging the transformative power of an auto-regressive language model based on the transformer architecture. Trained via fine-tuning the Llama 2 model, Vicuna is used primarily for research in large language models and chatbots, catering to the needs of AI enthusiasts, machine learning researchers, and hobbyists. The model operates under the Llama 2 Community License Agreement. The latest version, Vicuna v1.5 (16k), has been fine-tuned using supervised instruction and linear RoPE scaling, with training data comprising around 125K conversations collected from ShareGPT.com, packed into sequences of 16K tokens each. A comprehensive explanation of the training details can be found in the appendix to the linked paper titled \"Training Details of Vicuna Models.\"", "author": { - "name": "Nous Research", - "url": "https://nousresearch.com", - "blurb": "Nous Research is dedicated to advancing the field of natural language processing, in collaboration with the open-source community, through bleeding-edge research and a commitment to symbiotic development." + "name": "LMSYS Org", + "url": "https://lmsys.org/", + "blurb": "Large Model Systems Organization (LMSYS Org) is an open research organization founded by students and faculty from UC Berkeley in collaboration with UCSD and CMU." }, "numParameters": "13B", "resources": { - "canonicalUrl": "https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b", - "downloadUrl": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML" + "canonicalUrl": "https://huggingface.co/lmsys/vicuna-13b-v1.5-16k", + "downloadUrl": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML", + "paperUrl": "https://arxiv.org/abs/2306.05685" }, "trainedFor": "chat", "arch": "llama", "files": { "highlighted": { "economical": { - "name": "nous-hermes-llama2-13b.ggmlv3.q3_K_S.bin" + "name": "vicuna-13b-v1.5-16k.ggmlv3.q4_K_S.bin" }, "most_capable": { - "name": "nous-hermes-llama2-13b.ggmlv3.q6_K.bin" + "name": "vicuna-13b-v1.5-16k.ggmlv3.q6_K.bin" } }, "all": [ { - "name": "nous-hermes-llama2-13b.ggmlv3.q3_K_S.bin", - "url": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML/resolve/main/nous-hermes-llama2-13b.ggmlv3.q3_K_S.bin", - "sizeBytes": 5874151072, + "name": "vicuna-13b-v1.5-16k.ggmlv3.q4_K_S.bin", + "url": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML/resolve/main/vicuna-13b-v1.5-16k.ggmlv3.q4_K_S.bin", + "sizeBytes": 7365545088, "quantization": "Q4_K_S", "format": "ggml", - "sha256checksum": "9e76993f3e8a804b3ee012bc461fbd01e9ff71fbdf8ac9f6f13ccadd5ff51b4e", + "sha256checksum": "c343200172b677488a7e59a2ef74642b63b2578fde3c976911751181d29ce201", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/Nous-Hermes-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML" + "respository": "TheBloke/vicuna-13B-v1.5-16K-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML" }, { - "name": "nous-hermes-llama2-13b.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML/resolve/main/nous-hermes-llama2-13b.ggmlv3.q6_K.bin", - "sizeBytes": 10830311072, + "name": "vicuna-13b-v1.5-16k.ggmlv3.q6_K.bin", + "url": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML/resolve/main/vicuna-13b-v1.5-16k.ggmlv3.q6_K.bin", + "sizeBytes": 10678850688, "quantization": "Q6_K", "format": "ggml", - "sha256checksum": "02f696d0df53174cf9a3aba62d22a1b17882389a5381e68b2c7395845f87d3f5", + "sha256checksum": "949e2003a47552d2931ebce29b94ad59482561a36429fb1c00d13db7feb87fd9", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/Nous-Hermes-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/Nous-Hermes-Llama2-GGML" + "respository": "TheBloke/vicuna-13B-v1.5-16K-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GGML" } ] } }, { "_descriptorVersion": "0.0.1", - "datePublished": "2023-08-11T20:24:21", - "name": "OpenOrca-Platypus2-13B", - "description": "OpenOrca-Platypus2-13B is a powerful language model developed by merging garage-bAInd/Platypus2-13B and Open-Orca/OpenOrcaxOpenChat-Preview2-13B. It's an auto-regressive model based on the Lllama 2 transformer architecture, trained predominantly in English. The model was tested via Language Model Evaluation Harness and performed impressively on the HuggingFace Leaderboard with an average score of 64.56 across various shot scenarios. It also exceeded the base Preview2 model's performance on AGIEval and BigBench-Hard evaluations. Cole Hunter & Ariel Lee trained Platypus2-13B using the STEM and logic-based dataset garage-bAInd/Open-Platypus, while Open-Orca trained the OpenOrcaxOpenChat-Preview2-13B model using a refined subset of the GPT-4 data from the OpenOrca dataset. The training licenses for the base weights differ, with Platypus2-13B operating under a Non-Commercial Creative Commons license (CC BY-NC-4.0) and OpenOrcaxOpenChat-Preview2-13B under a Llama 2 Commercial license. Detailed training procedures and benchmark results are available in the Platypus GitHub repo.", + "datePublished": "2023-06-14T11:50:53.000Z", + "name": "WizardCoder-15B-V1.0", + "description": "WizardCoder: Empowering Code Large Language Models with Evol-Instruct. To develop our WizardCoder model, we begin by adapting the Evol-Instruct method specifically for coding tasks. This involves tailoring the prompt to the domain of code-related instructions. Subsequently, we fine-tune the Code LLM, StarCoder, utilizing the newly created instruction-following training set.", "author": { - "name": "Open Orca", - "url": "https://huggingface.co/Open-Orca", - "blurb": "A collaboration between Alignment Lab AI and the Platypus / garage-bAInd team" + "name": "WizardLM", + "url": "https://huggingface.co/WizardLM", + "blurb": "WizardLM: An Instruction-following LLM Using Evol-Instruct" }, - "numParameters": "13B", + "numParameters": "15B", "resources": { - "canonicalUrl": "https://huggingface.co/Open-Orca/OpenOrca-Platypus2-13B", - "downloadUrl": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML", - "paperUrl": "https://arxiv.org/abs/2308.07317" + "canonicalUrl": "https://huggingface.co/WizardLM/WizardCoder-15B-V1.0", + "downloadUrl": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML", + "paperUrl": "https://arxiv.org/abs/2306.08568" + }, + "trainedFor": "instruct", + "arch": "starcoder", + "files": { + "highlighted": { + "economical": { + "name": "WizardCoder-15B-1.0.ggmlv3.q4_0.bin" + }, + "most_capable": { + "name": "WizardCoder-15B-1.0.ggmlv3.q5_1.bin" + } + }, + "all": [ + { + "name": "WizardCoder-15B-1.0.ggmlv3.q4_0.bin", + "url": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML/resolve/main/WizardCoder-15B-1.0.ggmlv3.q4_0.bin", + "sizeBytes": 10746570393, + "quantization": "q4_0", + "format": "ggml", + "sha256checksum": "b70164bc0b58a472c0987905133735ab3b27e2c439dedf8174a43951c51c3229", + "publisher": { + "name": "TheBloke", + "socialUrl": "https://twitter.com/TheBlokeAI" + }, + "respository": "TheBloke/WizardCoder-15B-1.0-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML" + }, + { + "name": "WizardCoder-15B-1.0.ggmlv3.q5_1.bin", + "url": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML/resolve/main/WizardCoder-15B-1.0.ggmlv3.q5_1.bin", + "sizeBytes": 14257205145, + "quantization": "q5_1", + "format": "ggml", + "sha256checksum": "1219d9fc6d51901d9a1e58e3cb7f03818d02a1d0ab2d070b4cbabdefeb7d0363", + "publisher": { + "name": "TheBloke", + "socialUrl": "https://twitter.com/TheBlokeAI" + }, + "respository": "TheBloke/WizardCoder-15B-1.0-GGML", + "repositoryUrl": "https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML" + } + ] + } + }, + { + "_descriptorVersion": "0.0.1", + "datePublished": "2023-10-26T11:25:50", + "name": "Zephyr 7B \u00ce\u00b2", + "description": "The Zephyr-7B-\u00ce\u00b2 is the second model in the Zephyr series, designed to function as an assistant. It is a fine-tuned version of the mistralai/Mistral-7B-v0.1 model, leveraging a 7B parameter GPT-like architecture. The model has been trained on a combination of synthetic datasets and publicly available data using Direct Preference Optimization (DPO), a technique that improved its performance on the MT Bench. An important aspect to note is that the in-built alignment of the training datasets was deliberately omitted during the training process, a decision that, while enhancing the model's helpfulness, also makes it prone to generating potentially problematic outputs when prompted. Therefore, it is advised to use the model strictly for research and educational purposes. The model primarily supports the English language and is licensed under the MIT License. Additional details can be found in the associated technical report.", + "author": { + "name": "Hugging Face H4", + "url": "https://huggingface.co/HuggingFaceH4", + "blurb": "Hugging Face H4 team, focused on aligning language models to be helpful, honest, harmless, and huggy \u00f0\u0178\u00a4\u2014" + }, + "numParameters": "7B", + "resources": { + "canonicalUrl": "https://huggingface.co/HuggingFaceH4/zephyr-7b-beta", + "paperUrl": "https://arxiv.org/abs/2310.16944", + "downloadUrl": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF" }, "trainedFor": "chat", - "arch": "llama", + "arch": "mistral", "files": { "highlighted": { "economical": { - "name": "openorca-platypus2-13b.ggmlv3.q4_K_S.bin" + "name": "zephyr-7b-beta.Q4_K_S.gguf" }, "most_capable": { - "name": "openorca-platypus2-13b.ggmlv3.q6_K.bin" + "name": "zephyr-7b-beta.Q6_K.gguf" } }, "all": [ { - "name": "openorca-platypus2-13b.ggmlv3.q4_K_S.bin", - "url": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML/resolve/main/openorca-platypus2-13b.ggmlv3.q4_K_S.bin", - "sizeBytes": 7558851360, + "name": "zephyr-7b-beta.Q4_K_S.gguf", + "url": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF/resolve/main/zephyr-7b-beta.Q4_K_S.gguf", + "sizeBytes": 4140373696, "quantization": "Q4_K_S", - "format": "ggml", - "sha256checksum": "55318c58c22c4e4a8dd5504919cefde5f843e848b6c96cd06a3ad69f2f41365b", + "format": "gguf", + "sha256checksum": "cafa0b85b2efc15ca33023f3b87f8d0c44ddcace16b3fb608280e0eb8f425cb1", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/OpenOrca-Platypus2-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML" + "respository": "TheBloke/zephyr-7B-beta-GGUF", + "repositoryUrl": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF" }, { - "name": "openorca-platypus2-13b.ggmlv3.q6_K.bin", - "url": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML/resolve/main/openorca-platypus2-13b.ggmlv3.q6_K.bin", - "sizeBytes": 10829916960, + "name": "zephyr-7b-beta.Q6_K.gguf", + "url": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF/resolve/main/zephyr-7b-beta.Q6_K.gguf", + "sizeBytes": 5942064832, "quantization": "Q6_K", - "format": "ggml", - "sha256checksum": "0487eb1a91683020cb8dc59b096b9f7e086735efdf1cf5f5ec84c842ec0a6799", + "format": "gguf", + "sha256checksum": "39b52e291eea6040de078283ee5316ff2a317e2b6f59be56724d9b29bada6cfe", "publisher": { "name": "TheBloke", "socialUrl": "https://twitter.com/TheBlokeAI" }, - "respository": "TheBloke/OpenOrca-Platypus2-13B-GGML", - "repositoryUrl": "https://huggingface.co/TheBloke/OpenOrca-Platypus2-13B-GGML" + "respository": "TheBloke/zephyr-7B-beta-GGUF", + "repositoryUrl": "https://huggingface.co/TheBloke/zephyr-7B-beta-GGUF" } ] } diff --git a/models/falcon-40b-instruct.json b/models/falcon-40b-instruct.json new file mode 100644 index 0000000..c17b805 --- /dev/null +++ b/models/falcon-40b-instruct.json @@ -0,0 +1,86 @@ +{ + "_descriptorVersion": "0.0.1", + "datePublished": "2023-10-22T03:04:42", + "name": "falcon-40b-instruct", + "description": "Falcon-40B-Instruct, based on Falcon-40B, has been fine-tuned on chat and instruct datasets. It offers outstanding performance and is an excellent choice for chat and instruct applications. Falcon-7B-Instruct is part of the Falcon family of language models, known for their exceptional capabilities and openness.", + "author": { + "name": "Maddes8cht", + "url": "https://huggingface.co/maddes8cht", + "blurb": "Maddes8cht Passionate about Open Source and AI. On Hugging Face he is advocating for real open source AI models with OSI compliant licenses" + }, + "numParameters": "40B", + "resources": { + "canonicalUrl": "https://huggingface.co/tiiuae/falcon-40b-instruct", + "downloadUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" + }, + "trainedFor": "chat", + "arch": "falcon", + "files": { + "highlighted": { + "economical": { + "name": "tiiuae-falcon-40b-Q4_K_M.gguf" + }, + "most_capable": { + "name": "tiiuae-falcon-40b-Q6_K.gguf" + } + }, + "all": [ + { + "name": "tiiuae-falcon-40b-instruct-Q2_K.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf/blob/main/tiiuae-falcon-40b-instruct-Q2_K.gguf", + "sizeBytes": 17400024800, + "quantization": "Q2_K", + "format": "gguf", + "sha256checksum": "93551752b87a1b265c43dbbf3b1cdf47d4ff4d92122fa44ea12e3b0c6f7d1981", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-40b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" + }, + { + "name": "tiiuae-falcon-40b-instruct-Q4_K_M.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf/blob/main/tiiuae-falcon-40b-instruct-Q4_K_M.gguf", + "sizeBytes": 25452629728, + "quantization": "Q4_K_M", + "format": "gguf", + "sha256checksum": "ff140febf094721f656568250eb6f288ac8dfa49b5b4483139078761a9515cbb", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-40b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" + }, + { + "name": "tiiuae-falcon-40b-instruct-Q5_K_M.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf/blob/main/tiiuae-falcon-40b-instruct-Q5_K_M.gguf", + "sizeBytes": 30642294496, + "quantization": "Q5_K_M", + "format": "gguf", + "sha256checksum": "fe39d1fc51d35039da996e3d2ed262617d1a70b3d9d0a7762571c749103cba39", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" + }, + { + "name": "tiiuae-falcon-40b-instruct-Q6_K.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q6_K.gguf", + "sizeBytes": 34456456928, + "quantization": "Q6_K", + "format": "gguf", + "sha256checksum": "5c2a9c95dfcb72d34acb09ac304625847a1f3de307984f559e71f7414345a4e9", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-40b-instruct-gguf" + } + ] + } +} \ No newline at end of file diff --git a/models/falcon-7b-instruct.json b/models/falcon-7b-instruct.json new file mode 100644 index 0000000..4d92a94 --- /dev/null +++ b/models/falcon-7b-instruct.json @@ -0,0 +1,86 @@ +{ + "_descriptorVersion": "0.0.1", + "datePublished": "2023-10-31T16:01:50", + "name": "falcon-7b-instruct", + "description": "Falcon-7B-Instruct, based on Falcon-7B, has been fine-tuned on chat and instruct datasets. It offers outstanding performance and is an excellent choice for chat and instruct applications. Falcon-7B-Instruct is part of the Falcon family of language models, known for their exceptional capabilities and openness.", + "author": { + "name": "Maddes8cht", + "url": "https://huggingface.co/maddes8cht", + "blurb": "Maddes8cht Passionate about Open Source and AI. On Hugging Face he is advocating for real open source AI models with OSI compliant licenses" + }, + "numParameters": "7B", + "resources": { + "canonicalUrl": "https://huggingface.co/tiiuae/falcon-7b-instruct", + "downloadUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" + }, + "trainedFor": "chat", + "arch": "falcon", + "files": { + "highlighted": { + "economical": { + "name": "tiiuae-falcon-7b-instruct-Q4_K_M" + }, + "most_capable": { + "name": "tiiuae-falcon-7b-Q6_K.gguf" + } + }, + "all": [ + { + "name": "tiiuae-falcon-7b-instruct-Q2_K.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q2_K.gguf", + "sizeBytes": 4025162688, + "quantization": "Q2_K", + "format": "gguf", + "sha256checksum": "c0e87b27def08c0a9dc8fa9bfd70cd7bc373d328943530984165ac4b20d2202c", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" + }, + { + "name": "tiiuae-falcon-7b-instruct-Q4_K_M.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q4_K_M.gguf", + "sizeBytes": 4975385792, + "quantization": "Q4_K_M", + "format": "gguf", + "sha256checksum": "6f6c886ed07d4f6133a4dd7f8b799764ed719c50973d96816b2ac5ff3ada9913", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" + }, + { + "name": "tiiuae-falcon-7b-instruct-Q5_K_M.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q5_K_M.gguf", + "sizeBytes": 5731380160, + "quantization": "Q5_K_M", + "format": "gguf", + "sha256checksum": "3a4666a200024584f667b4dcf2761463046924f9b0708ba13315831416d8548a", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" + }, + { + "name": "tiiuae-falcon-7b-instruct-Q6_K.gguf", + "url": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf/blob/main/tiiuae-falcon-7b-instruct-Q6_K.gguf", + "sizeBytes": 7031618496, + "quantization": "Q6_K", + "format": "gguf", + "sha256checksum": "e6b23eb4193b2d8c7effb4870d2db7a3faa6ed2c8f2c075f7ced51591938818a", + "publisher": { + "name": "maddes8cht", + "socialUrl": "https://twitter.com/maddes1966" + }, + "respository": "maddes8cht/tiiuae-falcon-7b-instruct-gguf", + "repositoryUrl": "https://huggingface.co/maddes8cht/tiiuae-falcon-7b-instruct-gguf" + } + ] + } +} \ No newline at end of file diff --git a/schema.json b/schema.json index cb0f197..985bbda 100644 --- a/schema.json +++ b/schema.json @@ -51,7 +51,7 @@ }, "numParameters": { "type": "string", - "enum": ["3B", "7B", "13B", "15B", "30B", "65B", "unknown"] + "enum": ["3B", "7B", "13B", "15B", "30B", "40B", "65B", "unknown"] }, "trainedFor": { "type": "string", @@ -118,7 +118,7 @@ }, "quantization": { "type": "string", - "enum": ["q4_0", "q4_1", "q5_0", "q5_1", "q8_0", "Q2_K", "Q3_K_S", "Q3_K_M", "Q4_K_S", "Q4_K_M", "Q5_K_S", "Q5_K_M", "Q6_K", "unknown"] + "enum": ["q4_0", "q4_1", "q5_0", "q5_1", "q8_0", "Q2_K", "Q3_K_S", "Q3_K_M", "Q3_K_L", "Q4_K_S", "Q4_K_M", "Q5_K_S", "Q5_K_M", "Q6_K", "unknown"] }, "format": { "type": "string"