Qwen2.5-VL-7B-Instruct - DASH-Q 2-bit GGUF
llama.cpp GGUF files of Qwen/Qwen2.5-VL-7B-Instruct quantized with DASH-Q, at several 2-bit-class sizes. Every file uses only standard llama.cpp tensor types (no tensor above 4 bits) and loads in any recent llama.cpp build.
| File | Type | Size | Bits / weight |
|---|---|---|---|
Qwen2.5-VL-7B-Instruct-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 2.40 GB | 2.53 |
Qwen2.5-VL-7B-Instruct-DASHQ-IQ2_XS.gguf |
IQ2_XS | 2.71 GB | 2.85 |
Qwen2.5-VL-7B-Instruct-DASHQ-IQ2_M.gguf |
IQ2_M | 2.86 GB | 3.01 |
Qwen2.5-VL-7B-Instruct-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 3.11 GB | 3.27 |
Text only: the vision tower is not included.
Perplexity (lower is better)
| Type | Model | Size | WikiText-2 | C4 |
|---|---|---|---|---|
| IQ2_XXS | llama.cpp IQ2_XXS (imatrix) | 2.27 GB | 16.45 | 20.68 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 2.40 GB | 13.05 | 17.98 |
| IQ2_XXS | DASH-Q IQ2_XXS | 2.40 GB | 12.17 | 17.05 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 2.47 GB | 13.87 | 16.52 |
| IQ2_XS | DASH-Q IQ2_XS | 2.71 GB | 10.13 | 14.71 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 2.78 GB | 11.95 | 15.51 |
| IQ2_M | unsloth UD-IQ2_M | 2.86 GB | 10.75 | 14.58 |
| IQ2_M | DASH-Q IQ2_M | 2.86 GB | 10.00 | 14.16 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 3.02 GB | 10.00 | 14.33 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 3.24 GB | 10.36 | 13.95 |
| Q2_K_XL | DASH-Q Q2_K_XL | 3.11 GB | 9.43 | 13.76 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Qwen2.5-VL-7B-Instruct-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (Qwen/Qwen2.5-VL-7B-Instruct).