From 27433a3a075f049af510524bc666ac1a06a159a6 Mon Sep 17 00:00:00 2001 From: Cherrytest Date: Fri, 9 May 2025 19:08:58 +0000 Subject: [PATCH] Upload folder using ModelScope SDK --- .gitattributes | 4 +- README.md | 204 ++++++++++++++---- ...phin-Mistral-24B-Venice-Edition-IQ2_M.gguf | 3 + ...phin-Mistral-24B-Venice-Edition-IQ2_S.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-IQ2_XS.gguf | 3 + ...phin-Mistral-24B-Venice-Edition-IQ3_M.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-IQ3_XS.gguf | 3 + ...in-Mistral-24B-Venice-Edition-IQ3_XXS.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-IQ4_NL.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-IQ4_XS.gguf | 3 + ...lphin-Mistral-24B-Venice-Edition-Q2_K.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q2_K_L.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q3_K_L.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q3_K_M.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q3_K_S.gguf | 3 + ...in-Mistral-24B-Venice-Edition-Q3_K_XL.gguf | 3 + ...lphin-Mistral-24B-Venice-Edition-Q4_0.gguf | 3 + ...lphin-Mistral-24B-Venice-Edition-Q4_1.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q4_K_L.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q4_K_M.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q4_K_S.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q5_K_L.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q5_K_M.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q5_K_S.gguf | 3 + ...lphin-Mistral-24B-Venice-Edition-Q6_K.gguf | 3 + ...hin-Mistral-24B-Venice-Edition-Q6_K_L.gguf | 3 + ...lphin-Mistral-24B-Venice-Edition-Q8_0.gguf | 3 + ...lphin-Mistral-24B-Venice-Edition-bf16.gguf | 3 + ...Dolphin-Mistral-24B-Venice-Edition.imatrix | 3 + configuration.json | 1 + 30 files changed, 248 insertions(+), 42 deletions(-) create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_M.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_S.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_XS.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_M.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XS.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XXS.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_NL.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_XS.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K_L.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_L.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_M.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_S.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_XL.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_0.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_1.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_L.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_M.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_S.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_L.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_M.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_S.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K_L.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q8_0.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-bf16.gguf create mode 100644 cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition.imatrix create mode 100644 configuration.json diff --git a/.gitattributes b/.gitattributes index 53d7257..9b52413 100644 --- a/.gitattributes +++ b/.gitattributes @@ -44,4 +44,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text *.tar filter=lfs diff=lfs merge=lfs -text *.wasm filter=lfs diff=lfs merge=lfs -text *.zst filter=lfs diff=lfs merge=lfs -text -*tfevents* filter=lfs diff=lfs merge=lfs -text \ No newline at end of file +*tfevents* filter=lfs diff=lfs merge=lfs -text + +cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition.imatrix filter=lfs diff=lfs merge=lfs -text \ No newline at end of file diff --git a/README.md b/README.md index c4189fe..a8332d5 100644 --- a/README.md +++ b/README.md @@ -1,47 +1,169 @@ --- -license: Apache License 2.0 - -#model-type: -##如 gpt、phi、llama、chatglm、baichuan 等 -#- gpt - -#domain: -##如 nlp、cv、audio、multi-modal -#- nlp - -#language: -##语言代码列表 https://help.aliyun.com/document_detail/215387.html?spm=a2c4g.11186623.0.0.9f8d7467kni6Aa -#- cn - -#metrics: -##如 CIDEr、Blue、ROUGE 等 -#- CIDEr - -#tags: -##各种自定义，包括 pretrained、fine-tuned、instruction-tuned、RL-tuned 等训练方法和其他 -#- pretrained - -#tools: -##如 vllm、fastchat、llamacpp、AdaSeq 等 -#- vllm +quantized_by: bartowski +pipeline_tag: text-generation +license: apache-2.0 +base_model_relation: quantized +base_model: cognitivecomputations/Dolphin-Mistral-24B-Venice-Edition --- -### 当前模型的贡献者未提供更加详细的模型介绍。模型文件和权重，可浏览“模型文件”页面获取。 -#### 您可以通过如下git clone命令，或者ModelScope SDK来下载模型 -SDK下载 -```bash -#安装ModelScope -pip install modelscope +## Llamacpp imatrix Quantizations of Dolphin-Mistral-24B-Venice-Edition by cognitivecomputations + +Using llama.cpp release b5317 for quantization. + +Original model: https://huggingface.co/cognitivecomputations/Dolphin-Mistral-24B-Venice-Edition + +All quants made using imatrix option with dataset from [here](https://gist.github.com/bartowski1182/eb213dccb3571f863da82e99418f81e8) + +Run them in [LM Studio](https://lmstudio.ai/) + +Run them directly with [llama.cpp](https://github.com/ggerganov/llama.cpp), or any other llama.cpp based project + +## Prompt format + ``` -```python -#SDK模型下载 -from modelscope import snapshot_download -model_dir = snapshot_download('bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF') -``` -Git下载 -``` -#Git模型下载 -git clone https://www.modelscope.cn/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF.git +[SYSTEM_PROMPT]{system_prompt}[/SYSTEM_PROMPT][INST]{prompt}[/INST] ``` -

如果您是本模型的贡献者，我们邀请您根据模型贡献文档，及时完善模型卡片内容。

\ No newline at end of file +## Download a file (not the whole branch) from below: + +| Filename | Quant type | File Size | Split | Description | +| -------- | ---------- | --------- | ----- | ----------- | +| [Dolphin-Mistral-24B-Venice-Edition-bf16.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-bf16.gguf) | bf16 | 47.15GB | false | Full BF16 weights. | +| [Dolphin-Mistral-24B-Venice-Edition-Q8_0.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q8_0.gguf) | Q8_0 | 25.05GB | false | Extremely high quality, generally unneeded but max available quant. | +| [Dolphin-Mistral-24B-Venice-Edition-Q6_K_L.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K_L.gguf) | Q6_K_L | 19.67GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q6_K.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K.gguf) | Q6_K | 19.35GB | false | Very high quality, near perfect, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q5_K_L.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_L.gguf) | Q5_K_L | 17.18GB | false | Uses Q8_0 for embed and output weights. High quality, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q5_K_M.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_M.gguf) | Q5_K_M | 16.76GB | false | High quality, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q5_K_S.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_S.gguf) | Q5_K_S | 16.30GB | false | High quality, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q4_1.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_1.gguf) | Q4_1 | 14.87GB | false | Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon. | +| [Dolphin-Mistral-24B-Venice-Edition-Q4_K_L.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_L.gguf) | Q4_K_L | 14.83GB | false | Uses Q8_0 for embed and output weights. Good quality, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q4_K_M.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_M.gguf) | Q4_K_M | 14.33GB | false | Good quality, default size for most use cases, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q4_K_S.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_S.gguf) | Q4_K_S | 13.55GB | false | Slightly lower quality with more space savings, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q4_0.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_0.gguf) | Q4_0 | 13.49GB | false | Legacy format, offers online repacking for ARM and AVX CPU inference. | +| [Dolphin-Mistral-24B-Venice-Edition-IQ4_NL.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_NL.gguf) | IQ4_NL | 13.47GB | false | Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference. | +| [Dolphin-Mistral-24B-Venice-Edition-Q3_K_XL.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_XL.gguf) | Q3_K_XL | 12.99GB | false | Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability. | +| [Dolphin-Mistral-24B-Venice-Edition-IQ4_XS.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_XS.gguf) | IQ4_XS | 12.76GB | false | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. | +| [Dolphin-Mistral-24B-Venice-Edition-Q3_K_L.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_L.gguf) | Q3_K_L | 12.40GB | false | Lower quality but usable, good for low RAM availability. | +| [Dolphin-Mistral-24B-Venice-Edition-Q3_K_M.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_M.gguf) | Q3_K_M | 11.47GB | false | Low quality. | +| [Dolphin-Mistral-24B-Venice-Edition-IQ3_M.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_M.gguf) | IQ3_M | 10.65GB | false | Medium-low quality, new method with decent performance comparable to Q3_K_M. | +| [Dolphin-Mistral-24B-Venice-Edition-Q3_K_S.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_S.gguf) | Q3_K_S | 10.40GB | false | Low quality, not recommended. | +| [Dolphin-Mistral-24B-Venice-Edition-IQ3_XS.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XS.gguf) | IQ3_XS | 9.91GB | false | Lower quality, new method with decent performance, slightly better than Q3_K_S. | +| [Dolphin-Mistral-24B-Venice-Edition-Q2_K_L.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K_L.gguf) | Q2_K_L | 9.55GB | false | Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable. | +| [Dolphin-Mistral-24B-Venice-Edition-IQ3_XXS.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XXS.gguf) | IQ3_XXS | 9.28GB | false | Lower quality, new method with decent performance, comparable to Q3 quants. | +| [Dolphin-Mistral-24B-Venice-Edition-Q2_K.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K.gguf) | Q2_K | 8.89GB | false | Very low quality but surprisingly usable. | +| [Dolphin-Mistral-24B-Venice-Edition-IQ2_M.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_M.gguf) | IQ2_M | 8.11GB | false | Relatively low quality, uses SOTA techniques to be surprisingly usable. | +| [Dolphin-Mistral-24B-Venice-Edition-IQ2_S.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_S.gguf) | IQ2_S | 7.48GB | false | Low quality, uses SOTA techniques to be usable. | +| [Dolphin-Mistral-24B-Venice-Edition-IQ2_XS.gguf](https://huggingface.co/bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF/blob/main/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_XS.gguf) | IQ2_XS | 7.21GB | false | Low quality, uses SOTA techniques to be usable. | + +## Embed/output weights + +Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to. + +## Downloading using huggingface-cli + +
+ Click to view download instructions + +First, make sure you have hugginface-cli installed: + +``` +pip install -U "huggingface_hub[cli]" +``` + +Then, you can target the specific file you want: + +``` +huggingface-cli download bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF --include "cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_M.gguf" --local-dir ./ +``` + +If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run: + +``` +huggingface-cli download bartowski/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-GGUF --include "cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q8_0/*" --local-dir ./ +``` + +You can either specify a new local-dir (cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q8_0) or download them all in place (./) + +
+ +## ARM/AVX information + +Previously, you would download Q4_0_4_4/4_8/8_8, and these would have their weights interleaved in memory in order to improve performance on ARM and AVX machines by loading up more data in one pass. + +Now, however, there is something called "online repacking" for weights. details in [this PR](https://github.com/ggerganov/llama.cpp/pull/9921). If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly. + +As of llama.cpp build [b4282](https://github.com/ggerganov/llama.cpp/releases/tag/b4282) you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0. + +Additionally, if you want to get slightly better quality for , you can use IQ4_NL thanks to [this PR](https://github.com/ggerganov/llama.cpp/pull/10541) which will also repack the weights for ARM, though only the 4_4 for now. The loading time may be slower but it will result in an overall speed incrase. + +
+ Click to view Q4_0_X_X information (deprecated + +I'm keeping this section to show the potential theoretical uplift in performance from using the Q4_0 with online repacking. + +
+ Click to view benchmarks on an AVX2 system (EPYC7702) + +| model | size | params | backend | threads | test | t/s | % (vs Q4_0) | +| ------------------------------ | ---------: | ---------: | ---------- | ------: | ------------: | -------------------: |-------------: | +| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp512 | 204.03 ± 1.03 | 100% | +| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp1024 | 282.92 ± 0.19 | 100% | +| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp2048 | 259.49 ± 0.44 | 100% | +| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg128 | 39.12 ± 0.27 | 100% | +| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg256 | 39.31 ± 0.69 | 100% | +| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg512 | 40.52 ± 0.03 | 100% | +| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp512 | 301.02 ± 1.74 | 147% | +| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp1024 | 287.23 ± 0.20 | 101% | +| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp2048 | 262.77 ± 1.81 | 101% | +| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg128 | 18.80 ± 0.99 | 48% | +| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg256 | 24.46 ± 3.04 | 83% | +| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg512 | 36.32 ± 3.59 | 90% | +| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | pp512 | 271.71 ± 3.53 | 133% | +| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | pp1024 | 279.86 ± 45.63 | 100% | +| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | pp2048 | 320.77 ± 5.00 | 124% | +| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | tg128 | 43.51 ± 0.05 | 111% | +| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | tg256 | 43.35 ± 0.09 | 110% | +| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | tg512 | 42.60 ± 0.31 | 105% | + +Q4_0_8_8 offers a nice bump to prompt processing and a small bump to text generation + +
+ +
+ +## Which file should I choose? + +
+ Click here for details + +A great write up with charts showing various performances is provided by Artefact2 [here](https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9) + +The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have. + +If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM. + +If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total. + +Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'. + +If you don't want to think too much, grab one of the K-quants. These are in format 'QX_K_X', like Q5_K_M. + +If you want to get more into the weeds, you can check out this extremely useful feature chart: + +[llama.cpp feature matrix](https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix) + +But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQX_X, like IQ3_M. These are newer and offer better performance for their size. + +These I-quants can also be used on CPU, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide. + +
+ +## Credits + +Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset. + +Thank you ZeroWw for the inspiration to experiment with embed/output. + +Thank you to LM Studio for sponsoring my work. + +Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_M.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_M.gguf new file mode 100644 index 0000000..97ed73d --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff1e50497f7553d512062b5fb9d7103390db71378987bb2e0c36ba1e34536442 +size 135 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_S.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_S.gguf new file mode 100644 index 0000000..7989c7e --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6609bf173b3880e70f1e19ce527eed11feace341805d96276c4c4fbd26a1d241 +size 135 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_XS.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_XS.gguf new file mode 100644 index 0000000..f2a9f2a --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ2_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b5c9e4aec32a45f5c2bf6e8ec1bf3a5d05bc0ccd98c905bab53fe55f7bddf6f +size 135 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_M.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_M.gguf new file mode 100644 index 0000000..5bdac07 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42ad614d2cb3a3338152efc5fc55b818a12794da793da6e472cc2ebb9765c8c3 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XS.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XS.gguf new file mode 100644 index 0000000..38ebaf5 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61cb0002effe0f7b093b1dc703966041c4c68bf038c7f2226b66ab50e34406dd +size 135 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XXS.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XXS.gguf new file mode 100644 index 0000000..d8eb9cc --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ3_XXS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05b7be24569d699ad0c70e93c3e23164bdd1eab7e29053bded9f69f7db65a724 +size 135 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_NL.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_NL.gguf new file mode 100644 index 0000000..e1dd9ab --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_NL.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:295702aae686fefca822cc865aa189815a67f0b9792c5b57907f581e41807f55 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_XS.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_XS.gguf new file mode 100644 index 0000000..a4227d8 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-IQ4_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:feb86c8076142eb5bb305b69b2d786afa807e16aa6ec36beabe443c43feedf21 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K.gguf new file mode 100644 index 0000000..c7827d1 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b4536f6016565153401167606edd0409b01f855499e737f38fc11a5a9b83b95 +size 135 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K_L.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K_L.gguf new file mode 100644 index 0000000..1f9b939 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q2_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:059adeff1d8098043d9b195da00115ab42a5adba0fc6346b57c50da260ff6211 +size 135 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_L.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_L.gguf new file mode 100644 index 0000000..a04e8cc --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:357d7e32e31fae4afdf3ac3f75e1d80042e8712cb5023ccb96f4e2c874375c70 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_M.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_M.gguf new file mode 100644 index 0000000..0213363 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a7868bcfb9c301addc205d5a5314a67f66c08ba11799617d9cf0bbaa4c410b89 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_S.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_S.gguf new file mode 100644 index 0000000..a8f328b --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc3f6356d4852015b93b92eff08c774d20381424560a7da8b88107d18f6706ed +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_XL.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_XL.gguf new file mode 100644 index 0000000..3cd511c --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q3_K_XL.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6948e8bb46767aa53daee72995307165c89525950d92d546fb4a874d9f6f9e8 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_0.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_0.gguf new file mode 100644 index 0000000..621211f --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_0.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8476825721c02e9893e022629cf499dbfa5ef788ae8635f3183c6e4a0b17dc49 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_1.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_1.gguf new file mode 100644 index 0000000..fbf379c --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_1.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d5547e22f94e130ecf3b7363d41dbc071a37901d016f4db5f30e6417c71a3d36 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_L.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_L.gguf new file mode 100644 index 0000000..dabebc0 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fcaef56f43fd37e49bb88cae8b119a95d28b44c1ad740df7c64055d88db17219 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_M.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_M.gguf new file mode 100644 index 0000000..0eba99b --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45944ddcb053bc9e9cc48a669f742f363d50073051b0cb4c325515c409090170 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_S.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_S.gguf new file mode 100644 index 0000000..6a78e3a --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q4_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c9feb658ceb2134dc0ababb037090b30b26d0bb34c52e02167fddb25451de04 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_L.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_L.gguf new file mode 100644 index 0000000..9ae841c --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7528d9b57b3daffa02e04b09fa7e6e97b5afaa0670293542155573ec23a33c88 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_M.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_M.gguf new file mode 100644 index 0000000..0ab4ec6 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b0c727313ba401ba62ca1a413ba20c386ae31be19c37aef08c20c8ec595cd76 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_S.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_S.gguf new file mode 100644 index 0000000..2098e8b --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q5_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d3209a62d4e5810680c162eb65eb274a44c8c651ea1ba9c7bb409293c2c040a +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K.gguf new file mode 100644 index 0000000..0ca7739 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1955632a171a08d3625cb3a9fb67b1ec3e4fc7d322391680976e5eb6a217205a +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K_L.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K_L.gguf new file mode 100644 index 0000000..b0ca716 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q6_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:faebd4c4563279ae89e0bfb2522c1b8940899d00529af1ba2c35123f3da69a21 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q8_0.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q8_0.gguf new file mode 100644 index 0000000..6b3456d --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-Q8_0.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e597da9c8592e89cf89b909460a4c08bca3cf6efc4142ac6ab8471c5e980d1df +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-bf16.gguf b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-bf16.gguf new file mode 100644 index 0000000..a4ae1da --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition-bf16.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62dbc69f552cc89a4caf62d983d870cff212971306b0748cb0a49599d6aa2540 +size 136 diff --git a/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition.imatrix b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition.imatrix new file mode 100644 index 0000000..5918524 --- /dev/null +++ b/cognitivecomputations_Dolphin-Mistral-24B-Venice-Edition.imatrix @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:157f0c4925057f39215e862271f5d8ece19d69800d898e11d2e82326a6a47ec3 +size 10003538 diff --git a/configuration.json b/configuration.json new file mode 100644 index 0000000..bbeeda1 --- /dev/null +++ b/configuration.json @@ -0,0 +1 @@ +{"framework": "pytorch", "task": "text-generation", "allow_remote": true} \ No newline at end of file