base_model: ibivibiv/hydra-moe-120b
language:
- en
library_name: transformers
license: apache-2.0
no_imarix: https://github.com/ggerganov/llama.cpp/issues/6597
quantized_by: mradermacher
tags:
- moe
- moerge
About
weighted/imatrix quants of https://huggingface.co/ibivibiv/hydra-moe-120b
No more quants will be forthcoming, as llama.cpp segfaults.
static quants are available at https://huggingface.co/mradermacher/hydra-moe-120b-GGUF
Usage
If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files.
Provided Quants
(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
| Link | Type | Size/GB | Notes |
|---|---|---|---|
| GGUF | i1-IQ2_M | 37.3 | |
| GGUF | i1-Q2_K | 41.6 | IQ3_XXS probably better |
| GGUF | i1-IQ3_XXS | 43.8 | lower quality |
| GGUF | i1-IQ3_XS | 46.5 | |
| GGUF | i1-Q3_K_S | 49.1 | IQ3_XS probably better |
| GGUF | i1-IQ3_S | 49.2 | beats Q3_K* |
| PART 1 PART 2 | i1-Q3_K_M | 54.5 | IQ3_S probably better |
| PART 1 PART 2 | i1-Q3_K_L | 59.1 | IQ3_M probably better |
| PART 1 PART 2 | i1-IQ4_XS | 60.7 | |
| PART 1 PART 2 | i1-Q4_0 | 64.4 | fast, low quality |
| PART 1 PART 2 | i1-Q4_K_S | 64.7 | optimal size/speed/quality |
| PART 1 PART 2 | i1-Q4_K_M | 68.8 | fast, recommended |
| PART 1 PART 2 | i1-Q5_K_S | 78.3 | |
| PART 1 PART 2 | i1-Q5_K_M | 80.7 | |
| PART 1 PART 2 | i1-Q6_K | 93.3 | practically like static Q6_K |
Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):
And here are Artefact2's thoughts on the matter: https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
FAQ / Model Request
See https://huggingface.co/mradermacher/model_requests for some answers to questions you might have and/or if you want some other model quantized.
Thanks
I thank my company, nethype GmbH, for letting me use its servers and providing upgrades to my workstation to enable this work in my free time.