From 4955d5b57a7ef2f0cb15f65aad5af4f0624db83a Mon Sep 17 00:00:00 2001 From: ModelHub XC Date: Mon, 20 Jul 2026 00:04:06 +0800 Subject: [PATCH] =?UTF-8?q?=E5=88=9D=E5=A7=8B=E5=8C=96=E9=A1=B9=E7=9B=AE?= =?UTF-8?q?=EF=BC=8C=E7=94=B1ModelHub=20XC=E7=A4=BE=E5=8C=BA=E6=8F=90?= =?UTF-8?q?=E4=BE=9B=E6=A8=A1=E5=9E=8B?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Model: bartowski/nvidia_Nemotron-3-Nano-4B-GGUF Source: Original Platform --- .gitattributes | 73 +++++++++ README.md | 198 +++++++++++++++++++++++++ configuration.json | 1 + nvidia_Nemotron-3-Nano-4B-IQ2_M.gguf | 3 + nvidia_Nemotron-3-Nano-4B-IQ3_M.gguf | 3 + nvidia_Nemotron-3-Nano-4B-IQ3_XS.gguf | 3 + nvidia_Nemotron-3-Nano-4B-IQ3_XXS.gguf | 3 + nvidia_Nemotron-3-Nano-4B-IQ4_NL.gguf | 3 + nvidia_Nemotron-3-Nano-4B-IQ4_XS.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q2_K.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q2_K_L.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q3_K_L.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q3_K_M.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q3_K_S.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q3_K_XL.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q4_0.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q4_1.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q4_K_L.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q4_K_M.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q4_K_S.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q5_K_L.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q5_K_M.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q5_K_S.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q6_K.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q6_K_L.gguf | 3 + nvidia_Nemotron-3-Nano-4B-Q8_0.gguf | 3 + nvidia_Nemotron-3-Nano-4B-bf16.gguf | 3 + nvidia_Nemotron-3-Nano-4B-imatrix.gguf | 3 + 28 files changed, 347 insertions(+) create mode 100644 .gitattributes create mode 100644 README.md create mode 100644 configuration.json create mode 100644 nvidia_Nemotron-3-Nano-4B-IQ2_M.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-IQ3_M.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-IQ3_XS.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-IQ3_XXS.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-IQ4_NL.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-IQ4_XS.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q2_K.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q2_K_L.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q3_K_L.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q3_K_M.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q3_K_S.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q3_K_XL.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q4_0.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q4_1.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q4_K_L.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q4_K_M.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q4_K_S.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q5_K_L.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q5_K_M.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q5_K_S.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q6_K.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q6_K_L.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-Q8_0.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-bf16.gguf create mode 100644 nvidia_Nemotron-3-Nano-4B-imatrix.gguf diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..52c0e62 --- /dev/null +++ b/.gitattributes @@ -0,0 +1,73 @@ +*.7z filter=lfs diff=lfs merge=lfs -text +*.arrow filter=lfs diff=lfs merge=lfs -text +*.bin filter=lfs diff=lfs merge=lfs -text +*.bin.* filter=lfs diff=lfs merge=lfs -text +*.bz2 filter=lfs diff=lfs merge=lfs -text +*.ftz filter=lfs diff=lfs merge=lfs -text +*.gz filter=lfs diff=lfs merge=lfs -text +*.h5 filter=lfs diff=lfs merge=lfs -text +*.joblib filter=lfs diff=lfs merge=lfs -text +*.lfs.* filter=lfs diff=lfs merge=lfs -text +*.model filter=lfs diff=lfs merge=lfs -text +*.msgpack filter=lfs diff=lfs merge=lfs -text +*.onnx filter=lfs diff=lfs merge=lfs -text + +*.parquet filter=lfs diff=lfs merge=lfs -text +*.pb filter=lfs diff=lfs merge=lfs -text +*.pt filter=lfs diff=lfs merge=lfs -text +*.pth filter=lfs diff=lfs merge=lfs -text +*.rar filter=lfs diff=lfs merge=lfs -text +saved_model/**/* filter=lfs diff=lfs merge=lfs -text +*.tar.* filter=lfs diff=lfs merge=lfs -text +*.tflite filter=lfs diff=lfs merge=lfs -text +*.tgz filter=lfs diff=lfs merge=lfs -text +*.xz filter=lfs diff=lfs merge=lfs -text +*.zip filter=lfs diff=lfs merge=lfs -text +*.zstandard filter=lfs diff=lfs merge=lfs -text +*.tfevents* filter=lfs diff=lfs merge=lfs -text +*.db* filter=lfs diff=lfs merge=lfs -text +*.ark* filter=lfs diff=lfs merge=lfs -text +**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text +**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text +**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text +*.safetensors filter=lfs diff=lfs merge=lfs -text +*.ckpt filter=lfs diff=lfs merge=lfs -text + +*.ggml filter=lfs diff=lfs merge=lfs -text +*.llamafile* filter=lfs diff=lfs merge=lfs -text +*.pt2 filter=lfs diff=lfs merge=lfs -text +*.mlmodel filter=lfs diff=lfs merge=lfs -text +*.npy filter=lfs diff=lfs merge=lfs -text +*.npz filter=lfs diff=lfs merge=lfs -text +*.pickle filter=lfs diff=lfs merge=lfs -text +*.pkl 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 + +nvidia_Nemotron-3-Nano-4B-IQ4_NL.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-IQ3_XXS.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-bf16.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q5_K_L.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-imatrix.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-IQ3_XS.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-IQ2_M.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q2_K_L.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q4_K_L.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q6_K_L.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q4_1.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q3_K_XL.gguf filter=lfs diff=lfs merge=lfs -text +nvidia_Nemotron-3-Nano-4B-Q4_0.gguf filter=lfs diff=lfs merge=lfs -text \ No newline at end of file diff --git a/README.md b/README.md new file mode 100644 index 0000000..e1f5d0d --- /dev/null +++ b/README.md @@ -0,0 +1,198 @@ +--- +quantized_by: bartowski +pipeline_tag: text-generation +base_model: nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16 +license: other +track_downloads: true +license_name: nvidia-nemotron-open-model-license +language: +- en +license_link: https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-nemotron-open-model-license/ +tags: +- nvidia +base_model_relation: quantized +datasets: +- nvidia/Nemotron-CC-v2 +- nvidia/Nemotron-Post-Training-Dataset-v2 +- nvidia/Nemotron-Science-v1 +- nvidia/Nemotron-Instruction-Following-Chat-v1 +- nvidia/Nemotron-Agentic-v1 +- nvidia/Nemotron-Competitive-Programming-v1 +- nvidia/Nemotron-Math-Proofs-v1 +- nvidia/Nemotron-RL-Agentic-Conversational-Tool-Use-Pivot-v1 +- nvidia/Nemotron-RL-instruction_following +- nvidia/Nemotron-RL-agent-calendar_scheduling +- nvidia/Nemotron-RL-instruction_following-structured_outputs +--- + +## Llamacpp imatrix Quantizations of Nemotron-3-Nano-4B by nvidia + +Using llama.cpp release b8388 for quantization. + +Original model: https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16 + +All quants made using imatrix option with dataset from [here](https://gist.github.com/bartowski1182/82ae9b520227f57d79ba04add13d0d0d) + +Run them in your choice of tools: + +- [llama.cpp](https://github.com/ggml-org/llama.cpp) +- [LM Studio](https://lmstudio.ai/) +- [koboldcpp](https://github.com/LostRuins/koboldcpp) +- [Jan AI](https://www.jan.ai/) +- [Text Generation Web UI](https://github.com/oobabooga/text-generation-webui) +- [LoLLMs](https://github.com/ParisNeo/lollms) + +Note: if it's a newly supported model, you may need to wait for an update from the developers. + +## Prompt format + +``` +<|im_start|>system +{system_prompt}<|im_end|> +<|im_start|>user +{prompt}<|im_end|> +<|im_start|>assistant + +``` + +## Download a file (not the whole branch) from below: + +| Filename | Quant type | File Size | Split | Description | +| -------- | ---------- | --------- | ----- | ----------- | +| [Nemotron-3-Nano-4B-bf16.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-bf16.gguf) | bf16 | 7.96GB | false | Full BF16 weights. | +| [Nemotron-3-Nano-4B-Q8_0.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q8_0.gguf) | Q8_0 | 4.23GB | false | Extremely high quality, generally unneeded but max available quant. | +| [Nemotron-3-Nano-4B-Q6_K_L.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q6_K_L.gguf) | Q6_K_L | 4.01GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, *recommended*. | +| [Nemotron-3-Nano-4B-Q6_K.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q6_K.gguf) | Q6_K | 4.01GB | false | Very high quality, near perfect, *recommended*. | +| [Nemotron-3-Nano-4B-Q5_K_L.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q5_K_L.gguf) | Q5_K_L | 3.33GB | false | Uses Q8_0 for embed and output weights. High quality, *recommended*. | +| [Nemotron-3-Nano-4B-Q5_K_M.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q5_K_M.gguf) | Q5_K_M | 3.21GB | false | High quality, *recommended*. | +| [Nemotron-3-Nano-4B-Q5_K_S.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q5_K_S.gguf) | Q5_K_S | 3.14GB | false | High quality, *recommended*. | +| [Nemotron-3-Nano-4B-Q4_K_L.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q4_K_L.gguf) | Q4_K_L | 3.13GB | false | Uses Q8_0 for embed and output weights. Good quality, *recommended*. | +| [Nemotron-3-Nano-4B-Q4_K_M.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q4_K_M.gguf) | Q4_K_M | 2.98GB | false | Good quality, default size for most use cases, *recommended*. | +| [Nemotron-3-Nano-4B-Q4_K_S.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q4_K_S.gguf) | Q4_K_S | 2.90GB | false | Slightly lower quality with more space savings, *recommended*. | +| [Nemotron-3-Nano-4B-Q3_K_XL.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q3_K_XL.gguf) | Q3_K_XL | 2.85GB | false | Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability. | +| [Nemotron-3-Nano-4B-Q4_1.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q4_1.gguf) | Q4_1 | 2.80GB | false | Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon. | +| [Nemotron-3-Nano-4B-Q3_K_L.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q3_K_L.gguf) | Q3_K_L | 2.64GB | false | Lower quality but usable, good for low RAM availability. | +| [Nemotron-3-Nano-4B-Q2_K_L.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q2_K_L.gguf) | Q2_K_L | 2.64GB | false | Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable. | +| [Nemotron-3-Nano-4B-Q4_0.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q4_0.gguf) | Q4_0 | 2.60GB | false | Legacy format, offers online repacking for ARM and AVX CPU inference. | +| [Nemotron-3-Nano-4B-IQ4_NL.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-IQ4_NL.gguf) | IQ4_NL | 2.59GB | false | Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference. | +| [Nemotron-3-Nano-4B-Q3_K_M.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q3_K_M.gguf) | Q3_K_M | 2.57GB | false | Low quality. | +| [Nemotron-3-Nano-4B-IQ4_XS.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-IQ4_XS.gguf) | IQ4_XS | 2.56GB | false | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. | +| [Nemotron-3-Nano-4B-IQ3_M.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-IQ3_M.gguf) | IQ3_M | 2.49GB | false | Medium-low quality, new method with decent performance comparable to Q3_K_M. | +| [Nemotron-3-Nano-4B-Q3_K_S.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q3_K_S.gguf) | Q3_K_S | 2.46GB | false | Low quality, not recommended. | +| [Nemotron-3-Nano-4B-IQ3_XS.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-IQ3_XS.gguf) | IQ3_XS | 2.46GB | false | Lower quality, new method with decent performance, slightly better than Q3_K_S. | +| [Nemotron-3-Nano-4B-Q2_K.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-Q2_K.gguf) | Q2_K | 2.43GB | false | Very low quality but surprisingly usable. | +| [Nemotron-3-Nano-4B-IQ3_XXS.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-IQ3_XXS.gguf) | IQ3_XXS | 2.41GB | false | Lower quality, new method with decent performance, comparable to Q3 quants. | +| [Nemotron-3-Nano-4B-IQ2_M.gguf](https://huggingface.co/bartowski/nvidia_Nemotron-3-Nano-4B-GGUF/blob/main/nvidia_Nemotron-3-Nano-4B-IQ2_M.gguf) | IQ2_M | 2.29GB | false | Relatively low quality, uses SOTA techniques to be surprisingly 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/nvidia_Nemotron-3-Nano-4B-GGUF --include "nvidia_Nemotron-3-Nano-4B-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/nvidia_Nemotron-3-Nano-4B-GGUF --include "nvidia_Nemotron-3-Nano-4B-Q8_0/*" --local-dir ./ +``` + +You can either specify a new local-dir (nvidia_Nemotron-3-Nano-4B-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/ggml-org/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/ggml-org/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/ggml-org/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/ggml-org/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/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 diff --git a/nvidia_Nemotron-3-Nano-4B-IQ2_M.gguf b/nvidia_Nemotron-3-Nano-4B-IQ2_M.gguf new file mode 100644 index 0000000..35bcf52 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-IQ2_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:779f5f85317d93a96e6bbc552510338d36db14b5fd19290b30681dbb6b30ba2f +size 2286132032 diff --git a/nvidia_Nemotron-3-Nano-4B-IQ3_M.gguf b/nvidia_Nemotron-3-Nano-4B-IQ3_M.gguf new file mode 100644 index 0000000..3db0389 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-IQ3_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9aff3d1e814b5aa812176bc395d8e46ccbde2cc883586a81fa5c680d89475081 +size 2494092736 diff --git a/nvidia_Nemotron-3-Nano-4B-IQ3_XS.gguf b/nvidia_Nemotron-3-Nano-4B-IQ3_XS.gguf new file mode 100644 index 0000000..c1f79b9 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-IQ3_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:975f82fb02817e2556663d96dd94ef56c41fd6725a80e6a14503a449ee1b28ea +size 2457031488 diff --git a/nvidia_Nemotron-3-Nano-4B-IQ3_XXS.gguf b/nvidia_Nemotron-3-Nano-4B-IQ3_XXS.gguf new file mode 100644 index 0000000..381e573 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-IQ3_XXS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8c980c434442f6f2dfccc22ed49108654a058b86fcb74fde74f94d32f43eff2 +size 2413265472 diff --git a/nvidia_Nemotron-3-Nano-4B-IQ4_NL.gguf b/nvidia_Nemotron-3-Nano-4B-IQ4_NL.gguf new file mode 100644 index 0000000..3855a29 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-IQ4_NL.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fad849dd2af1290703a151ea2a748def729f14dfe8e95a812a3e39bc1b1c45fc +size 2589565120 diff --git a/nvidia_Nemotron-3-Nano-4B-IQ4_XS.gguf b/nvidia_Nemotron-3-Nano-4B-IQ4_XS.gguf new file mode 100644 index 0000000..0dd2339 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-IQ4_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4b2f6f9a779cb88a6c25cf84c757b87494d644bd6fef756c925d5bc918b5e62 +size 2558380736 diff --git a/nvidia_Nemotron-3-Nano-4B-Q2_K.gguf b/nvidia_Nemotron-3-Nano-4B-Q2_K.gguf new file mode 100644 index 0000000..e218fba --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q2_K.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6fc515a828e039ca6fac00bc9ea1028e2f0bde988f7bef56ffb616c76dad209c +size 2430124608 diff --git a/nvidia_Nemotron-3-Nano-4B-Q2_K_L.gguf b/nvidia_Nemotron-3-Nano-4B-Q2_K_L.gguf new file mode 100644 index 0000000..9295101 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q2_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78a6127f6168fd1cb671b749f35afa6b37157b663cf83361b13963c03763efaa +size 2635645504 diff --git a/nvidia_Nemotron-3-Nano-4B-Q3_K_L.gguf b/nvidia_Nemotron-3-Nano-4B-Q3_K_L.gguf new file mode 100644 index 0000000..140c632 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q3_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f74167adeb141253039d3a9701aea9662fc99086030217119febe55eddacf78 +size 2640380864 diff --git a/nvidia_Nemotron-3-Nano-4B-Q3_K_M.gguf b/nvidia_Nemotron-3-Nano-4B-Q3_K_M.gguf new file mode 100644 index 0000000..aa6acfb --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q3_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cff955a0edf5e86270fba86379a0f43705a3cc63b5cb6c4f07afa29e23e26878 +size 2566220736 diff --git a/nvidia_Nemotron-3-Nano-4B-Q3_K_S.gguf b/nvidia_Nemotron-3-Nano-4B-Q3_K_S.gguf new file mode 100644 index 0000000..86609a0 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q3_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86ead1bb834a102e61a32027452b39c74539589a88756f45f3552eecc260af75 +size 2457031488 diff --git a/nvidia_Nemotron-3-Nano-4B-Q3_K_XL.gguf b/nvidia_Nemotron-3-Nano-4B-Q3_K_XL.gguf new file mode 100644 index 0000000..ff376b8 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q3_K_XL.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85df4bab62edbfe9e6e91a6840408c8731a0fb42d1b17aa831c715bc09da009a +size 2845901760 diff --git a/nvidia_Nemotron-3-Nano-4B-Q4_0.gguf b/nvidia_Nemotron-3-Nano-4B-Q4_0.gguf new file mode 100644 index 0000000..087b022 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q4_0.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:592acff23f60e19d7d6c2a820cc0801fd9d9ac590a6bf1744597f85f7c1d487b +size 2601858240 diff --git a/nvidia_Nemotron-3-Nano-4B-Q4_1.gguf b/nvidia_Nemotron-3-Nano-4B-Q4_1.gguf new file mode 100644 index 0000000..4e4e725 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q4_1.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d62b48ab27334688b5aacca02a87c96feeb2c2315dfd3d451a01513181aa063 +size 2795079744 diff --git a/nvidia_Nemotron-3-Nano-4B-Q4_K_L.gguf b/nvidia_Nemotron-3-Nano-4B-Q4_K_L.gguf new file mode 100644 index 0000000..b605da5 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q4_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa2a1ed17628ce04db1ed60386baf8c95b604a85974ce4511a8b8fe381e4cf4d +size 3132513344 diff --git a/nvidia_Nemotron-3-Nano-4B-Q4_K_M.gguf b/nvidia_Nemotron-3-Nano-4B-Q4_K_M.gguf new file mode 100644 index 0000000..63c8cdc --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q4_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ad076af2037b37c7a654cd413de80518dacc6a1ef98476e51b4b3d84ef59583 +size 2978372672 diff --git a/nvidia_Nemotron-3-Nano-4B-Q4_K_S.gguf b/nvidia_Nemotron-3-Nano-4B-Q4_K_S.gguf new file mode 100644 index 0000000..9f1ff2b --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q4_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:303578df56149440ef5eb2c7aa06e8abca119ef2bc6861ffbd6e79074f8d75d1 +size 2900844480 diff --git a/nvidia_Nemotron-3-Nano-4B-Q5_K_L.gguf b/nvidia_Nemotron-3-Nano-4B-Q5_K_L.gguf new file mode 100644 index 0000000..5f673e8 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q5_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d793f12955c162eb8e8e6ec9536e1b3edf59ef283598c89ed58d5abafa2a055a +size 3333763008 diff --git a/nvidia_Nemotron-3-Nano-4B-Q5_K_M.gguf b/nvidia_Nemotron-3-Nano-4B-Q5_K_M.gguf new file mode 100644 index 0000000..7bd2682 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q5_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5596e389ec93fde6a96325b62df86692219a3ee5b3c54c9fee17fe4f8be60313 +size 3205312448 diff --git a/nvidia_Nemotron-3-Nano-4B-Q5_K_S.gguf b/nvidia_Nemotron-3-Nano-4B-Q5_K_S.gguf new file mode 100644 index 0000000..c01783d --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q5_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d7d94439d6fb629745cbd11dd5211af593afdcc338fe94decca9a47e89f3541 +size 3143740224 diff --git a/nvidia_Nemotron-3-Nano-4B-Q6_K.gguf b/nvidia_Nemotron-3-Nano-4B-Q6_K.gguf new file mode 100644 index 0000000..92eef35 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q6_K.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe93d2f6dd28d1182e329c4bfa333f675b7ab11d08249d3f36c6245dca698327 +size 4007557696 diff --git a/nvidia_Nemotron-3-Nano-4B-Q6_K_L.gguf b/nvidia_Nemotron-3-Nano-4B-Q6_K_L.gguf new file mode 100644 index 0000000..92eef35 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q6_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe93d2f6dd28d1182e329c4bfa333f675b7ab11d08249d3f36c6245dca698327 +size 4007557696 diff --git a/nvidia_Nemotron-3-Nano-4B-Q8_0.gguf b/nvidia_Nemotron-3-Nano-4B-Q8_0.gguf new file mode 100644 index 0000000..730a8a2 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-Q8_0.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e023262fd2fd5298e28e4f9aee668f976fc383ec5f49536a7f2a97bdf8db5b8 +size 4233682112 diff --git a/nvidia_Nemotron-3-Nano-4B-bf16.gguf b/nvidia_Nemotron-3-Nano-4B-bf16.gguf new file mode 100644 index 0000000..1eebb4f --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-bf16.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1393772e520fba2d0217a44d436f224419ed1c619de673451e218d2a7e86bde1 +size 7957650464 diff --git a/nvidia_Nemotron-3-Nano-4B-imatrix.gguf b/nvidia_Nemotron-3-Nano-4B-imatrix.gguf new file mode 100644 index 0000000..2fbb833 --- /dev/null +++ b/nvidia_Nemotron-3-Nano-4B-imatrix.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e093640cc7120aac090670e8175a534cc828d4c293ca74e886812a42eb1c288b +size 2221472