Update README.md
Browse files
README.md
CHANGED
|
@@ -24,25 +24,25 @@ EQ-Bench v2_de: 69.93 (171/171).
|
|
| 24 |
|
| 25 |
The merge tree involves the following models:
|
| 26 |
|
| 27 |
-
NousResearch/Hermes-3-Llama-3.1-8B
|
| 28 |
-
Undi95/Meta-Llama-3.1-8B-Claude
|
| 29 |
-
Dampfinchen/Llama-3.1-8B-Ultra-Instruct
|
| 30 |
-
VAGOsolutions/Llama-3.1-SauerkrautLM-8b-Instruct
|
| 31 |
-
akjindal53244/Llama-3.1-Storm-8B
|
| 32 |
-
nbeerbower/llama3.1-gutenberg-8B
|
| 33 |
-
Undi95/Meta-Llama-3.1-8B-Claude
|
| 34 |
-
DiscoResearch/Llama3-DiscoLeo-Instruct-8B-v0.1
|
| 35 |
-
nbeerbower/llama-3-wissenschaft-8B-v2
|
| 36 |
-
Azure99/blossom-v5-llama3-8b
|
| 37 |
-
VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
|
| 38 |
-
princeton-nlp/Llama-3-Instruct-8B-SimPO
|
| 39 |
-
Locutusque/llama-3-neural-chat-v1-8b
|
| 40 |
-
Locutusque/Llama-3-Orca-1.0-8B
|
| 41 |
-
DiscoResearch/Llama3_DiscoLM_German_8b_v0.1_experimental
|
| 42 |
-
seedboxai/Llama-3-Kafka-8B-v0.2
|
| 43 |
-
VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
|
| 44 |
-
nbeerbower/llama-3-wissenschaft-8B-v2
|
| 45 |
-
mlabonne/Daredevil-8B-abliterated-dpomix
|
| 46 |
|
| 47 |
There have been a number of steps involved, among which, slep merging of only middle layers compensating for tokenizer / chat template differences. An illustration below.
|
| 48 |
|
|
@@ -89,4 +89,5 @@ dtype: float16
|
|
| 89 |
|
| 90 |
## 💻 Usage
|
| 91 |
|
| 92 |
-
Use with llama3 chat template as common. Here are GGUF quants for use with e.g. ollama: cstr/llama3.1-8b-spaetzle-v90-GGUF
|
|
|
|
|
|
| 24 |
|
| 25 |
The merge tree involves the following models:
|
| 26 |
|
| 27 |
+
- NousResearch/Hermes-3-Llama-3.1-8B
|
| 28 |
+
- Undi95/Meta-Llama-3.1-8B-Claude
|
| 29 |
+
- Dampfinchen/Llama-3.1-8B-Ultra-Instruct
|
| 30 |
+
- VAGOsolutions/Llama-3.1-SauerkrautLM-8b-Instruct
|
| 31 |
+
- akjindal53244/Llama-3.1-Storm-8B
|
| 32 |
+
- nbeerbower/llama3.1-gutenberg-8B
|
| 33 |
+
- Undi95/Meta-Llama-3.1-8B-Claude
|
| 34 |
+
- DiscoResearch/Llama3-DiscoLeo-Instruct-8B-v0.1
|
| 35 |
+
- nbeerbower/llama-3-wissenschaft-8B-v2
|
| 36 |
+
- Azure99/blossom-v5-llama3-8b
|
| 37 |
+
- VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
|
| 38 |
+
- princeton-nlp/Llama-3-Instruct-8B-SimPO
|
| 39 |
+
- Locutusque/llama-3-neural-chat-v1-8b
|
| 40 |
+
- Locutusque/Llama-3-Orca-1.0-8B
|
| 41 |
+
- DiscoResearch/Llama3_DiscoLM_German_8b_v0.1_experimental
|
| 42 |
+
- seedboxai/Llama-3-Kafka-8B-v0.2
|
| 43 |
+
- VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
|
| 44 |
+
- nbeerbower/llama-3-wissenschaft-8B-v2
|
| 45 |
+
- mlabonne/Daredevil-8B-abliterated-dpomix
|
| 46 |
|
| 47 |
There have been a number of steps involved, among which, slep merging of only middle layers compensating for tokenizer / chat template differences. An illustration below.
|
| 48 |
|
|
|
|
| 89 |
|
| 90 |
## 💻 Usage
|
| 91 |
|
| 92 |
+
Use with llama3 chat template as common. Here are GGUF quants for use with llama.cpp & wrappers as e.g. ollama: [cstr/llama3.1-8b-spaetzle-v90-GGUF](https://huggingface.co/cstr/llama3.1-8b-spaetzle-v90-GGUF)
|
| 93 |
+
|