commit b83d3d350f69f78bf8ea6c28fb1f2ecbaf4bbdfb Author: ModelHub XC Date: Thu Jun 4 00:20:19 2026 +0800 初始化项目，由ModelHub XC社区提供模型 Model: bartowski/OpenGVLab_InternVL3_5-4B-GGUF Source: Original Platform diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..f7773b3 --- /dev/null +++ b/.gitattributes @@ -0,0 +1,62 @@ +*.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 +OpenGVLab_InternVL3_5-4B-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q6_K_L.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q5_K_L.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q4_K_L.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q4_1.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q4_0.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-IQ4_NL.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q3_K_XL.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-IQ3_XS.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-IQ3_XXS.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q2_K_L.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-IQ2_M.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-bf16.gguf filter=lfs diff=lfs merge=lfs -text +OpenGVLab_InternVL3_5-4B-imatrix.gguf filter=lfs diff=lfs merge=lfs -text +mmproj-OpenGVLab_InternVL3_5-4B-f16.gguf filter=lfs diff=lfs merge=lfs -text +mmproj-OpenGVLab_InternVL3_5-4B-bf16.gguf filter=lfs diff=lfs merge=lfs -text diff --git a/OpenGVLab_InternVL3_5-4B-IQ2_M.gguf b/OpenGVLab_InternVL3_5-4B-IQ2_M.gguf new file mode 100644 index 0000000..fc70388 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-IQ2_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bf127546725e5514c70f3f75671dedc5cfc4389f08bacb37197d49cace632404 +size 1680110272 diff --git a/OpenGVLab_InternVL3_5-4B-IQ3_M.gguf b/OpenGVLab_InternVL3_5-4B-IQ3_M.gguf new file mode 100644 index 0000000..0a5daaf --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-IQ3_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3fb94a495c4aaf43b0310fcbd3e8fe85a488e07f1542d5ca8b913f21080ad6b4 +size 2130022592 diff --git a/OpenGVLab_InternVL3_5-4B-IQ3_XS.gguf b/OpenGVLab_InternVL3_5-4B-IQ3_XS.gguf new file mode 100644 index 0000000..0dc6463 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-IQ3_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa39f972adfef041143506b1e5b465b52d8a00027cd7f2bd9a07d09cd8d892ce +size 1981501632 diff --git a/OpenGVLab_InternVL3_5-4B-IQ3_XXS.gguf b/OpenGVLab_InternVL3_5-4B-IQ3_XXS.gguf new file mode 100644 index 0000000..b9f5ba4 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-IQ3_XXS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce461603aa42bdf8f53338d715c326d6bc0fee19f6d530cca1ae6561ba1dba6c +size 1837314752 diff --git a/OpenGVLab_InternVL3_5-4B-IQ4_NL.gguf b/OpenGVLab_InternVL3_5-4B-IQ4_NL.gguf new file mode 100644 index 0000000..211aa96 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-IQ4_NL.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:739f16dac7e7f05f620432bf8d495797843302d145fe94c2e9361a1a825065fe +size 2600128192 diff --git a/OpenGVLab_InternVL3_5-4B-IQ4_XS.gguf b/OpenGVLab_InternVL3_5-4B-IQ4_XS.gguf new file mode 100644 index 0000000..f045dbe --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-IQ4_XS.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a4b93b338ec2ef737a968a65f4b9ff6f20604d24f6feb8d139fc180330e0f03 +size 2477381312 diff --git a/OpenGVLab_InternVL3_5-4B-Q2_K.gguf b/OpenGVLab_InternVL3_5-4B-Q2_K.gguf new file mode 100644 index 0000000..fb9aba7 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q2_K.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa5cb96774e811bdc67a13f42d00e1ab5487993a964dcfa58766d299f0319fbc +size 1797122752 diff --git a/OpenGVLab_InternVL3_5-4B-Q2_K_L.gguf b/OpenGVLab_InternVL3_5-4B-Q2_K_L.gguf new file mode 100644 index 0000000..40cf2d5 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q2_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42be0fef12262a4f33bb554c9f725f1cd60ff9681671b1f57174dfff0476e49b +size 2176962752 diff --git a/OpenGVLab_InternVL3_5-4B-Q3_K_L.gguf b/OpenGVLab_InternVL3_5-4B-Q3_K_L.gguf new file mode 100644 index 0000000..b11f120 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q3_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1153547dc5d7e68f9b5bf574dda342fedab3d5b964f4c90fb3be4bc0c3246967 +size 2406912192 diff --git a/OpenGVLab_InternVL3_5-4B-Q3_K_M.gguf b/OpenGVLab_InternVL3_5-4B-Q3_K_M.gguf new file mode 100644 index 0000000..46e22c0 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q3_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05c42d941c43082bf8a4d42e0ae9d347f7a499ca42996237314f24f0904d2896 +size 2242744512 diff --git a/OpenGVLab_InternVL3_5-4B-Q3_K_S.gguf b/OpenGVLab_InternVL3_5-4B-Q3_K_S.gguf new file mode 100644 index 0000000..e55638a --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q3_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a541f59081cd34e339b97a6961e14e98e62b698796b456ca933c33934e82b5d +size 2054123712 diff --git a/OpenGVLab_InternVL3_5-4B-Q3_K_XL.gguf b/OpenGVLab_InternVL3_5-4B-Q3_K_XL.gguf new file mode 100644 index 0000000..4ef91f9 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q3_K_XL.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c57fac71f6a2ce1540449e14de38293daec6d218ba84fa274cce06ad1934454 +size 2747248832 diff --git a/OpenGVLab_InternVL3_5-4B-Q4_0.gguf b/OpenGVLab_InternVL3_5-4B-Q4_0.gguf new file mode 100644 index 0000000..077d85d --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q4_0.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eb10e53ae463d0d0e54b666aefefce35293d1b7696dcecf12c546c0d0c021337 +size 2594557632 diff --git a/OpenGVLab_InternVL3_5-4B-Q4_1.gguf b/OpenGVLab_InternVL3_5-4B-Q4_1.gguf new file mode 100644 index 0000000..a066b8f --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q4_1.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7330efa1818892af481c6a9cb0b279a9d90ff8f24fcb9522072a8c6e76032aa3 +size 2839723712 diff --git a/OpenGVLab_InternVL3_5-4B-Q4_K_L.gguf b/OpenGVLab_InternVL3_5-4B-Q4_K_L.gguf new file mode 100644 index 0000000..ac3c1d3 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q4_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:afa6b06307244f651ebdb2eb527cc85580161d7d0f65e1726e502c6324809d35 +size 3004743872 diff --git a/OpenGVLab_InternVL3_5-4B-Q4_K_M.gguf b/OpenGVLab_InternVL3_5-4B-Q4_K_M.gguf new file mode 100644 index 0000000..7e2549a --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q4_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c1612b6896ad14caa501238e72afa17a600651d0984225e3ff78b39de86099c +size 2716065472 diff --git a/OpenGVLab_InternVL3_5-4B-Q4_K_S.gguf b/OpenGVLab_InternVL3_5-4B-Q4_K_S.gguf new file mode 100644 index 0000000..1b693e0 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q4_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6b2918021b959a51b6a76a42b2c262903f477082d57eec8c8effaf61ca57a7b +size 2602094272 diff --git a/OpenGVLab_InternVL3_5-4B-Q5_K_L.gguf b/OpenGVLab_InternVL3_5-4B-Q5_K_L.gguf new file mode 100644 index 0000000..fab71ee --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q5_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:edc1dc20d9a674c228139238e60840896b5494d05372d70ee1086e3a2bd3962e +size 3396976832 diff --git a/OpenGVLab_InternVL3_5-4B-Q5_K_M.gguf b/OpenGVLab_InternVL3_5-4B-Q5_K_M.gguf new file mode 100644 index 0000000..3b90c07 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q5_K_M.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc820484268ac0060471787e882f159af15f30952dc7adfadc04f08996bc8af4 +size 3156917952 diff --git a/OpenGVLab_InternVL3_5-4B-Q5_K_S.gguf b/OpenGVLab_InternVL3_5-4B-Q5_K_S.gguf new file mode 100644 index 0000000..55962e1 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q5_K_S.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65e21353eb29293d9c4d752db84ba3eaefe6e0e0df3ef2bad7d976079b21b68f +size 3091115712 diff --git a/OpenGVLab_InternVL3_5-4B-Q6_K.gguf b/OpenGVLab_InternVL3_5-4B-Q6_K.gguf new file mode 100644 index 0000000..7071b85 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q6_K.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9d3fac3267f214152762bc74911808b34e1c60c24b4c63d3b76ac1e1dbebbb4 +size 3625323712 diff --git a/OpenGVLab_InternVL3_5-4B-Q6_K_L.gguf b/OpenGVLab_InternVL3_5-4B-Q6_K_L.gguf new file mode 100644 index 0000000..1d3eda7 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q6_K_L.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:157ad29bf7c7a9fa5ea778e207e77e21a12e1644231822741aae3c01604b77f2 +size 3813724352 diff --git a/OpenGVLab_InternVL3_5-4B-Q8_0.gguf b/OpenGVLab_InternVL3_5-4B-Q8_0.gguf new file mode 100644 index 0000000..8d2ef65 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-Q8_0.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ece87031e20486b1a4b86a0ba0f06b8b3b6eed676c8c6842e31041524489992d +size 4693668032 diff --git a/OpenGVLab_InternVL3_5-4B-bf16.gguf b/OpenGVLab_InternVL3_5-4B-bf16.gguf new file mode 100644 index 0000000..edb8494 --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-bf16.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:161ff4491525e55d64ef9191b7226566f74e03c848aed0eb0a39d39aea9a63f0 +size 8829194144 diff --git a/OpenGVLab_InternVL3_5-4B-imatrix.gguf b/OpenGVLab_InternVL3_5-4B-imatrix.gguf new file mode 100644 index 0000000..be18a6d --- /dev/null +++ b/OpenGVLab_InternVL3_5-4B-imatrix.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a967cf43c649250cfa2c5dfc0653abd7bc39ae3862562f32adf4245b82834e3 +size 3872640 diff --git a/README.md b/README.md new file mode 100644 index 0000000..643fe50 --- /dev/null +++ b/README.md @@ -0,0 +1,164 @@ +--- +quantized_by: bartowski +pipeline_tag: image-text-to-text +base_model: OpenGVLab/InternVL3_5-4B +base_model_relation: quantized +--- + +## Llamacpp imatrix Quantizations of InternVL3_5-4B by OpenGVLab + +Using llama.cpp release b6258 for quantization. + +Original model: https://huggingface.co/OpenGVLab/InternVL3_5-4B + +All quants made using imatrix option with dataset from [here](https://gist.github.com/bartowski1182/eb213dccb3571f863da82e99418f81e8) combined with a subset of combined_all_small.parquet from Ed Addario [here](https://huggingface.co/datasets/eaddario/imatrix-calibration/blob/main/combined_all_small.parquet) + +Run them in [LM Studio](https://lmstudio.ai/) + +Run them directly with [llama.cpp](https://github.com/ggml-org/llama.cpp), or any other llama.cpp based project + +## Prompt format + +No prompt format found, check original model page + +## Download a file (not the whole branch) from below: + +| Filename | Quant type | File Size | Split | Description | +| -------- | ---------- | --------- | ----- | ----------- | +| [InternVL3_5-4B-bf16.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-bf16.gguf) | bf16 | 8.83GB | false | Full BF16 weights. | +| [InternVL3_5-4B-Q8_0.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q8_0.gguf) | Q8_0 | 4.69GB | false | Extremely high quality, generally unneeded but max available quant. | +| [InternVL3_5-4B-Q6_K_L.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q6_K_L.gguf) | Q6_K_L | 3.81GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, *recommended*. | +| [InternVL3_5-4B-Q6_K.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q6_K.gguf) | Q6_K | 3.63GB | false | Very high quality, near perfect, *recommended*. | +| [InternVL3_5-4B-Q5_K_L.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q5_K_L.gguf) | Q5_K_L | 3.40GB | false | Uses Q8_0 for embed and output weights. High quality, *recommended*. | +| [InternVL3_5-4B-Q5_K_M.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q5_K_M.gguf) | Q5_K_M | 3.16GB | false | High quality, *recommended*. | +| [InternVL3_5-4B-Q5_K_S.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q5_K_S.gguf) | Q5_K_S | 3.09GB | false | High quality, *recommended*. | +| [InternVL3_5-4B-Q4_K_L.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q4_K_L.gguf) | Q4_K_L | 3.00GB | false | Uses Q8_0 for embed and output weights. Good quality, *recommended*. | +| [InternVL3_5-4B-Q4_1.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q4_1.gguf) | Q4_1 | 2.84GB | false | Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon. | +| [InternVL3_5-4B-Q3_K_XL.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q3_K_XL.gguf) | Q3_K_XL | 2.75GB | false | Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability. | +| [InternVL3_5-4B-Q4_K_M.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q4_K_M.gguf) | Q4_K_M | 2.72GB | false | Good quality, default size for most use cases, *recommended*. | +| [InternVL3_5-4B-Q4_K_S.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q4_K_S.gguf) | Q4_K_S | 2.60GB | false | Slightly lower quality with more space savings, *recommended*. | +| [InternVL3_5-4B-IQ4_NL.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-IQ4_NL.gguf) | IQ4_NL | 2.60GB | false | Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference. | +| [InternVL3_5-4B-Q4_0.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q4_0.gguf) | Q4_0 | 2.59GB | false | Legacy format, offers online repacking for ARM and AVX CPU inference. | +| [InternVL3_5-4B-IQ4_XS.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-IQ4_XS.gguf) | IQ4_XS | 2.48GB | false | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. | +| [InternVL3_5-4B-Q3_K_L.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q3_K_L.gguf) | Q3_K_L | 2.41GB | false | Lower quality but usable, good for low RAM availability. | +| [InternVL3_5-4B-Q3_K_M.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q3_K_M.gguf) | Q3_K_M | 2.24GB | false | Low quality. | +| [InternVL3_5-4B-Q2_K_L.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q2_K_L.gguf) | Q2_K_L | 2.18GB | false | Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable. | +| [InternVL3_5-4B-IQ3_M.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-IQ3_M.gguf) | IQ3_M | 2.13GB | false | Medium-low quality, new method with decent performance comparable to Q3_K_M. | +| [InternVL3_5-4B-Q3_K_S.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q3_K_S.gguf) | Q3_K_S | 2.05GB | false | Low quality, not recommended. | +| [InternVL3_5-4B-IQ3_XS.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-IQ3_XS.gguf) | IQ3_XS | 1.98GB | false | Lower quality, new method with decent performance, slightly better than Q3_K_S. | +| [InternVL3_5-4B-IQ3_XXS.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-IQ3_XXS.gguf) | IQ3_XXS | 1.84GB | false | Lower quality, new method with decent performance, comparable to Q3 quants. | +| [InternVL3_5-4B-Q2_K.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-Q2_K.gguf) | Q2_K | 1.80GB | false | Very low quality but surprisingly usable. | +| [InternVL3_5-4B-IQ2_M.gguf](https://huggingface.co/bartowski/OpenGVLab_InternVL3_5-4B-GGUF/blob/main/OpenGVLab_InternVL3_5-4B-IQ2_M.gguf) | IQ2_M | 1.68GB | 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/OpenGVLab_InternVL3_5-4B-GGUF --include "OpenGVLab_InternVL3_5-4B-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/OpenGVLab_InternVL3_5-4B-GGUF --include "OpenGVLab_InternVL3_5-4B-Q8_0/*" --local-dir ./ +``` + +You can either specify a new local-dir (OpenGVLab_InternVL3_5-4B-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/mmproj-OpenGVLab_InternVL3_5-4B-bf16.gguf b/mmproj-OpenGVLab_InternVL3_5-4B-bf16.gguf new file mode 100644 index 0000000..d749be7 --- /dev/null +++ b/mmproj-OpenGVLab_InternVL3_5-4B-bf16.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c697d9fb3090cde25cd3c24c1cfad6970ef823e6f3fd072174200034f667d143 +size 646227360 diff --git a/mmproj-OpenGVLab_InternVL3_5-4B-f16.gguf b/mmproj-OpenGVLab_InternVL3_5-4B-f16.gguf new file mode 100644 index 0000000..407e2a5 --- /dev/null +++ b/mmproj-OpenGVLab_InternVL3_5-4B-f16.gguf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f9704972fcb9cb0a4f2c0f4eb7fe4f58e53ccd4b06ec17cf7a80271aa963eb7 +size 645023136