24 GB VRAM
Collection
Quants that run fast on single 3090/4090 card with 24GB of VRAM and 4096 context length
•
18 items
•
Updated
•
5
This is a merge of pre-trained language models created using mergekit.
This model was merged using the passthrough merge method.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
dtype: float32
merge_method: passthrough
slices:
- sources:
- layer_range: [0, 20]
model: "jondurbin_bagel-34b-v0.2"
parameters:
scale:
- filter: q_proj
value: 0.7071067812
- filter: k_proj
value: 0.7071067812
- value: 1
- sources:
- layer_range: [10, 30]
model: "jondurbin_bagel-34b-v0.2"
parameters:
scale:
- filter: q_proj
value: 0.7071067812
- filter: k_proj
value: 0.7071067812
- value: 1
- sources:
- layer_range: [20, 40]
model: "jondurbin_bagel-34b-v0.2"
parameters:
scale:
- filter: q_proj
value: 0.7071067812
- filter: k_proj
value: 0.7071067812
- value: 1
- sources:
- layer_range: [30, 50]
model: "jondurbin_bagel-34b-v0.2"
parameters:
scale:
- filter: q_proj
value: 0.7071067812
- filter: k_proj
value: 0.7071067812
- value: 1
- sources:
- layer_range: [40, 60]
model: "jondurbin_bagel-34b-v0.2"
parameters:
scale:
- filter: q_proj
value: 0.7071067812
- filter: k_proj
value: 0.7071067812
- value: 1
name: 2xbagel_fp32
---
dtype: bfloat16
merge_method: passthrough
slices:
- sources:
- layer_range: [0, 100]
model: 2xbagel_fp32
name: bagel_new