RichardErkhov
commited on
Commit
•
5b6b7f5
1
Parent(s):
63c1dce
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,91 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Quantization made by Richard Erkhov.
|
2 |
+
|
3 |
+
[Github](https://github.com/RichardErkhov)
|
4 |
+
|
5 |
+
[Discord](https://discord.gg/pvy7H8DZMG)
|
6 |
+
|
7 |
+
[Request more models](https://github.com/RichardErkhov/quant_request)
|
8 |
+
|
9 |
+
|
10 |
+
Chronorctypus-Limarobormes-13b - GGUF
|
11 |
+
- Model creator: https://huggingface.co/chargoddard/
|
12 |
+
- Original model: https://huggingface.co/chargoddard/Chronorctypus-Limarobormes-13b/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [Chronorctypus-Limarobormes-13b.Q2_K.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q2_K.gguf) | Q2_K | 4.52GB |
|
18 |
+
| [Chronorctypus-Limarobormes-13b.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.IQ3_XS.gguf) | IQ3_XS | 4.99GB |
|
19 |
+
| [Chronorctypus-Limarobormes-13b.IQ3_S.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.IQ3_S.gguf) | IQ3_S | 5.27GB |
|
20 |
+
| [Chronorctypus-Limarobormes-13b.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q3_K_S.gguf) | Q3_K_S | 5.27GB |
|
21 |
+
| [Chronorctypus-Limarobormes-13b.IQ3_M.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.IQ3_M.gguf) | IQ3_M | 5.57GB |
|
22 |
+
| [Chronorctypus-Limarobormes-13b.Q3_K.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q3_K.gguf) | Q3_K | 5.9GB |
|
23 |
+
| [Chronorctypus-Limarobormes-13b.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q3_K_M.gguf) | Q3_K_M | 5.9GB |
|
24 |
+
| [Chronorctypus-Limarobormes-13b.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q3_K_L.gguf) | Q3_K_L | 6.45GB |
|
25 |
+
| [Chronorctypus-Limarobormes-13b.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.IQ4_XS.gguf) | IQ4_XS | 6.54GB |
|
26 |
+
| [Chronorctypus-Limarobormes-13b.Q4_0.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q4_0.gguf) | Q4_0 | 6.86GB |
|
27 |
+
| [Chronorctypus-Limarobormes-13b.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.IQ4_NL.gguf) | IQ4_NL | 6.9GB |
|
28 |
+
| [Chronorctypus-Limarobormes-13b.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q4_K_S.gguf) | Q4_K_S | 6.91GB |
|
29 |
+
| [Chronorctypus-Limarobormes-13b.Q4_K.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q4_K.gguf) | Q4_K | 7.33GB |
|
30 |
+
| [Chronorctypus-Limarobormes-13b.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q4_K_M.gguf) | Q4_K_M | 7.33GB |
|
31 |
+
| [Chronorctypus-Limarobormes-13b.Q4_1.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q4_1.gguf) | Q4_1 | 7.61GB |
|
32 |
+
| [Chronorctypus-Limarobormes-13b.Q5_0.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q5_0.gguf) | Q5_0 | 8.36GB |
|
33 |
+
| [Chronorctypus-Limarobormes-13b.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q5_K_S.gguf) | Q5_K_S | 8.36GB |
|
34 |
+
| [Chronorctypus-Limarobormes-13b.Q5_K.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q5_K.gguf) | Q5_K | 8.6GB |
|
35 |
+
| [Chronorctypus-Limarobormes-13b.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q5_K_M.gguf) | Q5_K_M | 8.6GB |
|
36 |
+
| [Chronorctypus-Limarobormes-13b.Q5_1.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q5_1.gguf) | Q5_1 | 9.1GB |
|
37 |
+
| [Chronorctypus-Limarobormes-13b.Q6_K.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q6_K.gguf) | Q6_K | 9.95GB |
|
38 |
+
| [Chronorctypus-Limarobormes-13b.Q8_0.gguf](https://huggingface.co/RichardErkhov/chargoddard_-_Chronorctypus-Limarobormes-13b-gguf/blob/main/Chronorctypus-Limarobormes-13b.Q8_0.gguf) | Q8_0 | 12.88GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
---
|
45 |
+
tags:
|
46 |
+
- llama
|
47 |
+
- merge
|
48 |
+
---
|
49 |
+
Five different instruction-tuned models (which I'm sure are intuitively obvious from the name) merged using the methodology described in [Resolving Interference When Merging Models](https://arxiv.org/abs/2306.01708).
|
50 |
+
|
51 |
+
In theory this should retain more of the capabilites of the constituent models than a straight linear merge would. In my testing, it feels quite capable.
|
52 |
+
|
53 |
+
Base model used for the merge: [TheBloke/Llama-2-13B-fp16](https://huggingface.co/TheBloke/Llama-2-13B-fp16)
|
54 |
+
|
55 |
+
Models merged in:
|
56 |
+
* [OpenOrca-Platypus2-13B](https://huggingface.co/Open-Orca/OpenOrca-Platypus2-13B)
|
57 |
+
* [limarp-13b-merged](https://huggingface.co/Oniichat/limarp-13b-merged)
|
58 |
+
* [Nous-Hermes-Llama2-13b](https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b)
|
59 |
+
* [chronos-13b-v2](https://huggingface.co/elinas/chronos-13b-v2)
|
60 |
+
* [airoboros-l2-13b-gpt4-1.4.1](https://huggingface.co/jondurbin/airoboros-l2-13b-gpt4-1.4.1)
|
61 |
+
|
62 |
+
Works quite well with Alpaca-style prompts:
|
63 |
+
```
|
64 |
+
### Instruction:
|
65 |
+
|
66 |
+
...
|
67 |
+
|
68 |
+
### Response:
|
69 |
+
```
|
70 |
+
|
71 |
+
The script I used to perform the merge is available [here](https://github.com/cg123/ties-merge).
|
72 |
+
|
73 |
+
The command that produced this model:
|
74 |
+
```
|
75 |
+
python ties_merge.py TheBloke/Llama-2-13B-fp16 ./Chronorctypus-Limarobormes-13b --merge elinas/chronos-13b-v2 --merge Open-Orca/OpenOrca-Platypus2-13B --merge Oniichat/limarp-13b-merged --merge jondurbin/airoboros-l2-13b-gpt4-1.4.1 --merge NousResearch/Nous-Hermes-Llama2-13b --cuda
|
76 |
+
```
|
77 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
78 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_chargoddard__Chronorctypus-Limarobormes-13b)
|
79 |
+
|
80 |
+
| Metric | Value |
|
81 |
+
|-----------------------|---------------------------|
|
82 |
+
| Avg. | 49.88 |
|
83 |
+
| ARC (25-shot) | 59.9 |
|
84 |
+
| HellaSwag (10-shot) | 82.75 |
|
85 |
+
| MMLU (5-shot) | 58.45 |
|
86 |
+
| TruthfulQA (0-shot) | 51.9 |
|
87 |
+
| Winogrande (5-shot) | 74.43 |
|
88 |
+
| GSM8K (5-shot) | 3.87 |
|
89 |
+
| DROP (3-shot) | 17.89 |
|
90 |
+
|
91 |
+
|