Nemesia-Qwen-2.5-7B-v1.0

EDIT: MAY NOT WORK FOR GGUFs

I don't know if its an issue with me or the model, but I can't seem to make quants of this model. I consistently get llama_model_quantize: failed to quantize: tensor 'blk.24.attn_norm.weight' has invalid data. My whole setup has so many stds and idiosynchasies that it may just be my system, but I tried redoing the whole thing and same stuff happened. At this point it may be an issue with the NuSLERP method or one of the models I'm using in the merge. Not sure gang. I will try swapping out a model or two in the merge and trying again to upload as a v2.0.

EDIT (again): DO NOT USE THIS MODEL

I tried it four more times, swapping out models, swapping base models and models entirely, swapping params, git pulling llamacpp and mergekit, nothing. Errors when making quants, every one. I'm declaring this a lost cause. I'm leaving this up in case someone gets it working.

info

Merge using the brand new NuSLERP method. Fresh out of the oven. Performance not guaranteed.

Uses the slightly-unstable EVA and two other finetunes I found. I also turned on both the NuSLERP exclusive mergekit options for fun.

Named after the nemesia, a temperate shrubby flower. I tried to pick a flower that sounded kind of like NuSLERP. It doesn't, but the name still has the '''essence''' of NuSLERP I guess? (it doesn't.) Very pretty flower nonetheless

mergekit

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the NuSLERP merge method using EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1 as a base.

Models Merged

The following models were included in the merge:

• allura-org/Teleut-7b
• FourOhFour/Vapor_v2_7B

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: allura-org/Teleut-7b
    parameters:
      weight: 0.6
  - model: FourOhFour/Vapor_v2_7B
    parameters:
      weight: 0.2
  - model: EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1
    parameters:
      weight: 1.0
merge_method: nuslerp
base_model: EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1
parameters:
  normalize: true
  int8_mask: true
  nuslerp_flatten: false
  nuslerp_row_wise: true
dtype: float16