From 25697e885f904142240ffdb697d7233ffb9f2367 Mon Sep 17 00:00:00 2001 From: ModelHub XC Date: Tue, 21 Jul 2026 18:10:11 +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/mlabonne_Qwen3-4B-abliterated-GGUF Source: Original Platform --- .gitattributes | 60 ++++++++ README.md | 171 +++++++++++++++++++++ mlabonne_Qwen3-4B-abliterated-IQ2_M.gguf | 3 + mlabonne_Qwen3-4B-abliterated-IQ3_M.gguf | 3 + mlabonne_Qwen3-4B-abliterated-IQ3_XS.gguf | 3 + mlabonne_Qwen3-4B-abliterated-IQ3_XXS.gguf | 3 + mlabonne_Qwen3-4B-abliterated-IQ4_NL.gguf | 3 + mlabonne_Qwen3-4B-abliterated-IQ4_XS.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q2_K.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q2_K_L.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q3_K_L.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q3_K_M.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q3_K_S.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q3_K_XL.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q4_0.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q4_1.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q4_K_L.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q4_K_M.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q4_K_S.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q5_K_L.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q5_K_M.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q5_K_S.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q6_K.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q6_K_L.gguf | 3 + mlabonne_Qwen3-4B-abliterated-Q8_0.gguf | 3 + mlabonne_Qwen3-4B-abliterated-bf16.gguf | 3 + mlabonne_Qwen3-4B-abliterated.imatrix | 3 + 27 files changed, 306 insertions(+) create mode 100644 .gitattributes create mode 100644 README.md create mode 100644 mlabonne_Qwen3-4B-abliterated-IQ2_M.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-IQ3_M.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-IQ3_XS.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-IQ3_XXS.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-IQ4_NL.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-IQ4_XS.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q2_K.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q2_K_L.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q3_K_L.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q3_K_M.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q3_K_S.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q3_K_XL.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q4_0.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q4_1.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q4_K_L.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q4_K_M.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q4_K_S.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q5_K_L.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q5_K_M.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q5_K_S.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q6_K.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q6_K_L.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-Q8_0.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated-bf16.gguf create mode 100644 mlabonne_Qwen3-4B-abliterated.imatrix diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..b795563 --- /dev/null +++ b/.gitattributes @@ -0,0 +1,60 @@ +*.7z filter=lfs diff=lfs merge=lfs -text +*.arrow filter=lfs diff=lfs merge=lfs -text +*.bin filter=lfs diff=lfs merge=lfs -text +*.bz2 filter=lfs diff=lfs merge=lfs -text +*.ckpt 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 +*.mlmodel filter=lfs diff=lfs merge=lfs -text +*.model filter=lfs diff=lfs merge=lfs -text +*.msgpack filter=lfs diff=lfs merge=lfs -text +*.npy filter=lfs diff=lfs merge=lfs -text +*.npz filter=lfs diff=lfs merge=lfs -text +*.onnx filter=lfs diff=lfs merge=lfs -text +*.ot filter=lfs diff=lfs merge=lfs -text +*.parquet filter=lfs diff=lfs merge=lfs -text +*.pb filter=lfs diff=lfs merge=lfs -text +*.pickle filter=lfs diff=lfs merge=lfs -text +*.pkl 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 +*.safetensors filter=lfs diff=lfs merge=lfs -text +saved_model/**/* filter=lfs diff=lfs merge=lfs -text +*.tar.* 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 +*.wasm filter=lfs diff=lfs merge=lfs -text +*.xz filter=lfs diff=lfs merge=lfs -text +*.zip filter=lfs diff=lfs merge=lfs -text +*.zst filter=lfs diff=lfs merge=lfs -text +*tfevents* filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q6_K_L.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q5_K_L.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q4_K_L.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q4_1.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q4_0.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-IQ4_NL.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q3_K_XL.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-IQ3_XS.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-IQ3_XXS.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q2_K_L.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-IQ2_M.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated-bf16.gguf filter=lfs diff=lfs merge=lfs -text +mlabonne_Qwen3-4B-abliterated.imatrix filter=lfs diff=lfs merge=lfs -text diff --git a/README.md b/README.md new file mode 100644 index 0000000..1623d8b --- /dev/null +++ b/README.md @@ -0,0 +1,171 @@ +--- +quantized_by: bartowski +pipeline_tag: text-generation +tags: [] +base_model: mlabonne/Qwen3-4B-abliterated +base_model_relation: quantized +--- + +## Llamacpp imatrix Quantizations of Qwen3-4B-abliterated by mlabonne + +Using llama.cpp release b5200 for quantization. + +Original model: https://huggingface.co/mlabonne/Qwen3-4B-abliterated + +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 + +``` +<|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 | +| -------- | ---------- | --------- | ----- | ----------- | +| [Qwen3-4B-abliterated-bf16.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-bf16.gguf) | bf16 | 8.05GB | false | Full BF16 weights. | +| [Qwen3-4B-abliterated-Q8_0.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q8_0.gguf) | Q8_0 | 4.28GB | false | Extremely high quality, generally unneeded but max available quant. | +| [Qwen3-4B-abliterated-Q6_K_L.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q6_K_L.gguf) | Q6_K_L | 3.40GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, *recommended*. | +| [Qwen3-4B-abliterated-Q6_K.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q6_K.gguf) | Q6_K | 3.31GB | false | Very high quality, near perfect, *recommended*. | +| [Qwen3-4B-abliterated-Q5_K_L.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q5_K_L.gguf) | Q5_K_L | 2.98GB | false | Uses Q8_0 for embed and output weights. High quality, *recommended*. | +| [Qwen3-4B-abliterated-Q5_K_M.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q5_K_M.gguf) | Q5_K_M | 2.89GB | false | High quality, *recommended*. | +| [Qwen3-4B-abliterated-Q5_K_S.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q5_K_S.gguf) | Q5_K_S | 2.82GB | false | High quality, *recommended*. | +| [Qwen3-4B-abliterated-Q4_1.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q4_1.gguf) | Q4_1 | 2.60GB | false | Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon. | +| [Qwen3-4B-abliterated-Q4_K_L.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q4_K_L.gguf) | Q4_K_L | 2.59GB | false | Uses Q8_0 for embed and output weights. Good quality, *recommended*. | +| [Qwen3-4B-abliterated-Q4_K_M.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q4_K_M.gguf) | Q4_K_M | 2.50GB | false | Good quality, default size for most use cases, *recommended*. | +| [Qwen3-4B-abliterated-Q4_K_S.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q4_K_S.gguf) | Q4_K_S | 2.38GB | false | Slightly lower quality with more space savings, *recommended*. | +| [Qwen3-4B-abliterated-Q4_0.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q4_0.gguf) | Q4_0 | 2.38GB | false | Legacy format, offers online repacking for ARM and AVX CPU inference. | +| [Qwen3-4B-abliterated-IQ4_NL.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-IQ4_NL.gguf) | IQ4_NL | 2.38GB | false | Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference. | +| [Qwen3-4B-abliterated-Q3_K_XL.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q3_K_XL.gguf) | Q3_K_XL | 2.33GB | false | Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability. | +| [Qwen3-4B-abliterated-IQ4_XS.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-IQ4_XS.gguf) | IQ4_XS | 2.27GB | false | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. | +| [Qwen3-4B-abliterated-Q3_K_L.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q3_K_L.gguf) | Q3_K_L | 2.24GB | false | Lower quality but usable, good for low RAM availability. | +| [Qwen3-4B-abliterated-Q3_K_M.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q3_K_M.gguf) | Q3_K_M | 2.08GB | false | Low quality. | +| [Qwen3-4B-abliterated-IQ3_M.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-IQ3_M.gguf) | IQ3_M | 1.96GB | false | Medium-low quality, new method with decent performance comparable to Q3_K_M. | +| [Qwen3-4B-abliterated-Q3_K_S.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q3_K_S.gguf) | Q3_K_S | 1.89GB | false | Low quality, not recommended. | +| [Qwen3-4B-abliterated-IQ3_XS.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-IQ3_XS.gguf) | IQ3_XS | 1.81GB | false | Lower quality, new method with decent performance, slightly better than Q3_K_S. | +| [Qwen3-4B-abliterated-Q2_K_L.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q2_K_L.gguf) | Q2_K_L | 1.76GB | false | Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable. | +| [Qwen3-4B-abliterated-IQ3_XXS.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-IQ3_XXS.gguf) | IQ3_XXS | 1.67GB | false | Lower quality, new method with decent performance, comparable to Q3 quants. | +| [Qwen3-4B-abliterated-Q2_K.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-Q2_K.gguf) | Q2_K | 1.67GB | false | Very low quality but surprisingly usable. | +| [Qwen3-4B-abliterated-IQ2_M.gguf](https://huggingface.co/bartowski/mlabonne_Qwen3-4B-abliterated-GGUF/blob/main/mlabonne_Qwen3-4B-abliterated-IQ2_M.gguf) | IQ2_M | 1.51GB | 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/mlabonne_Qwen3-4B-abliterated-GGUF --include "mlabonne_Qwen3-4B-abliterated-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/mlabonne_Qwen3-4B-abliterated-GGUF --include "mlabonne_Qwen3-4B-abliterated-Q8_0/*" --local-dir ./ +``` + +You can either specify a new local-dir (mlabonne_Qwen3-4B-abliterated-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/mlabonne_Qwen3-4B-abliterated-IQ2_M.gguf b/mlabonne_Qwen3-4B-abliterated-IQ2_M.gguf new file mode 100644 index 0000000..3263e78 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-IQ2_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e56bb9b5c9a0e293825fe843e52094e6a248e6959637d9a5b06def37a06049f4 +size 1512983584 diff --git a/mlabonne_Qwen3-4B-abliterated-IQ3_M.gguf b/mlabonne_Qwen3-4B-abliterated-IQ3_M.gguf new file mode 100644 index 0000000..4181810 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-IQ3_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d1dd01f1a56c9c3525c42cc5211a39c53549ad00e8f3ddd8c306e9bbfb7778f +size 1962895904 diff --git a/mlabonne_Qwen3-4B-abliterated-IQ3_XS.gguf b/mlabonne_Qwen3-4B-abliterated-IQ3_XS.gguf new file mode 100644 index 0000000..2235d6f --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-IQ3_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e63f5ba905c114f7877c8c5812c9aecc2ffd35b46611dfb6f9914567c038a33 +size 1814374944 diff --git a/mlabonne_Qwen3-4B-abliterated-IQ3_XXS.gguf b/mlabonne_Qwen3-4B-abliterated-IQ3_XXS.gguf new file mode 100644 index 0000000..707c539 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-IQ3_XXS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e69cae63c9c491fa237e4f2eb40b3f34158042c929b659174581008a1428d42 +size 1670188064 diff --git a/mlabonne_Qwen3-4B-abliterated-IQ4_NL.gguf b/mlabonne_Qwen3-4B-abliterated-IQ4_NL.gguf new file mode 100644 index 0000000..fd91f84 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-IQ4_NL.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8908988812de4a912a0f5e735f6b29085c0f0fa9a66b20464e40c23d2e612453 +size 2381343264 diff --git a/mlabonne_Qwen3-4B-abliterated-IQ4_XS.gguf b/mlabonne_Qwen3-4B-abliterated-IQ4_XS.gguf new file mode 100644 index 0000000..cfedf26 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-IQ4_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:83e3db3cb5b00bfc9abe0b0b5cc41c541bb57042a74998914b5924710b3b76de +size 2270751264 diff --git a/mlabonne_Qwen3-4B-abliterated-Q2_K.gguf b/mlabonne_Qwen3-4B-abliterated-Q2_K.gguf new file mode 100644 index 0000000..3595352 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q2_K.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77d85c8807a3b0d85d6452cb61af9f9cb0bc206603d630cc358bbbdfc12a383b +size 1669499424 diff --git a/mlabonne_Qwen3-4B-abliterated-Q2_K_L.gguf b/mlabonne_Qwen3-4B-abliterated-Q2_K_L.gguf new file mode 100644 index 0000000..39309eb --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q2_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f28c0b3e38e9f625cf5fb140f3dd8d8ef2fc0acf20fb2ea3cc98d1678747a555 +size 1763699744 diff --git a/mlabonne_Qwen3-4B-abliterated-Q3_K_L.gguf b/mlabonne_Qwen3-4B-abliterated-Q3_K_L.gguf new file mode 100644 index 0000000..2b4bc7e --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q3_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:869de2dabfb997ead6aad0b28fbda92958b7e5a6ef90459eefe10c4a9ef2c603 +size 2239785504 diff --git a/mlabonne_Qwen3-4B-abliterated-Q3_K_M.gguf b/mlabonne_Qwen3-4B-abliterated-Q3_K_M.gguf new file mode 100644 index 0000000..2a8b6fe --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q3_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ace7f95f32fabc400de47a305037e62072227dd0628ed4e0626d38200e695414 +size 2075617824 diff --git a/mlabonne_Qwen3-4B-abliterated-Q3_K_S.gguf b/mlabonne_Qwen3-4B-abliterated-Q3_K_S.gguf new file mode 100644 index 0000000..2b77292 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q3_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5effa0653204a9d9ea448192458e09a931edc8c60f312ad11dc560b87d58a9e +size 1886997024 diff --git a/mlabonne_Qwen3-4B-abliterated-Q3_K_XL.gguf b/mlabonne_Qwen3-4B-abliterated-Q3_K_XL.gguf new file mode 100644 index 0000000..d7cfd26 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q3_K_XL.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a0607c78d2a95614bedd5cdcd5b68d29fcafe57e08a9dabee410d8d5761cb8b +size 2333985824 diff --git a/mlabonne_Qwen3-4B-abliterated-Q4_0.gguf b/mlabonne_Qwen3-4B-abliterated-Q4_0.gguf new file mode 100644 index 0000000..58c4ece --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q4_0.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7459ac56164a476e21377bb69d8cf0af4c28d91eb746021b5a1d798be9d6d2b6 +size 2375772704 diff --git a/mlabonne_Qwen3-4B-abliterated-Q4_1.gguf b/mlabonne_Qwen3-4B-abliterated-Q4_1.gguf new file mode 100644 index 0000000..439ca66 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q4_1.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64d44b4d8718e46bf7c49264a77e574c9bf6f73a1bc1546861ea9316a3452289 +size 2596629024 diff --git a/mlabonne_Qwen3-4B-abliterated-Q4_K_L.gguf b/mlabonne_Qwen3-4B-abliterated-Q4_K_L.gguf new file mode 100644 index 0000000..38d5d06 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q4_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1051f8f10b73d0c16d07385a9f1dbd156157935191cc093f5228d917574296e +size 2591480864 diff --git a/mlabonne_Qwen3-4B-abliterated-Q4_K_M.gguf b/mlabonne_Qwen3-4B-abliterated-Q4_K_M.gguf new file mode 100644 index 0000000..15efdeb --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q4_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:004f7b8f59ccd5fa42258c52aa2087b89524cced84e955b9c8b115035ca073b2 +size 2497280544 diff --git a/mlabonne_Qwen3-4B-abliterated-Q4_K_S.gguf b/mlabonne_Qwen3-4B-abliterated-Q4_K_S.gguf new file mode 100644 index 0000000..e79f63c --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q4_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e25dad578aa99d43ecfd7a02b047474537682dab9230fa42239ed62b0bfea0b +size 2383309344 diff --git a/mlabonne_Qwen3-4B-abliterated-Q5_K_L.gguf b/mlabonne_Qwen3-4B-abliterated-Q5_K_L.gguf new file mode 100644 index 0000000..3242161 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q5_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e5e57a690ab1805dcfd39e9342481f9d40bd567e07fb5a62ec7bb756e59e6d1 +size 2983713824 diff --git a/mlabonne_Qwen3-4B-abliterated-Q5_K_M.gguf b/mlabonne_Qwen3-4B-abliterated-Q5_K_M.gguf new file mode 100644 index 0000000..eabab8a --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q5_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f2e7c11115afd3ca85d3e5e72b0ed97231994820b9de450a62a0f8707b9e2da +size 2889513504 diff --git a/mlabonne_Qwen3-4B-abliterated-Q5_K_S.gguf b/mlabonne_Qwen3-4B-abliterated-Q5_K_S.gguf new file mode 100644 index 0000000..0fcb580 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q5_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e64a64e9498c1fd706e0d109580987fafeb777a3e4a6c8def1816f1cd5588a92 +size 2823711264 diff --git a/mlabonne_Qwen3-4B-abliterated-Q6_K.gguf b/mlabonne_Qwen3-4B-abliterated-Q6_K.gguf new file mode 100644 index 0000000..4538f49 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q6_K.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abe1412c5b47f268aa08df47e79235ea794325aef83cbe7a7ab5002c99824c35 +size 3306261024 diff --git a/mlabonne_Qwen3-4B-abliterated-Q6_K_L.gguf b/mlabonne_Qwen3-4B-abliterated-Q6_K_L.gguf new file mode 100644 index 0000000..9d8657d --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q6_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c362fce6c71e1b7511f400c4171ece6d72d166aba0ff9cf83f4827260ebb347c +size 3400461344 diff --git a/mlabonne_Qwen3-4B-abliterated-Q8_0.gguf b/mlabonne_Qwen3-4B-abliterated-Q8_0.gguf new file mode 100644 index 0000000..9e00f74 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-Q8_0.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc86a17b751cc6246c4477d01da39d57b7707d4c5c2c9c06fdfaeb5a9c14b5bd +size 4280405024 diff --git a/mlabonne_Qwen3-4B-abliterated-bf16.gguf b/mlabonne_Qwen3-4B-abliterated-bf16.gguf new file mode 100644 index 0000000..3f3e5ce --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated-bf16.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1bd1e80e8e4ec58bb9cdfe891d0a722435c3b18b721ae16eb48856d4b960f2b +size 8051284736 diff --git a/mlabonne_Qwen3-4B-abliterated.imatrix b/mlabonne_Qwen3-4B-abliterated.imatrix new file mode 100644 index 0000000..3595730 --- /dev/null +++ b/mlabonne_Qwen3-4B-abliterated.imatrix @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7dade5790c56a9be76c955466861264c70c1c2198f0e997f1c38e458bd3dba35 +size 3842222