Commit
•
cf7b767
1
Parent(s):
b29249f
Update README.md
Browse files
README.md
CHANGED
@@ -1,62 +1,54 @@
|
|
1 |
---
|
2 |
-
|
3 |
-
library_name: transformers
|
4 |
tags:
|
5 |
-
-
|
6 |
-
-
|
7 |
-
|
|
|
8 |
---
|
9 |
-
# output-model-directory
|
10 |
-
|
11 |
-
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
12 |
|
13 |
-
|
14 |
-
### Merge Method
|
15 |
|
16 |
-
|
|
|
17 |
|
18 |
-
|
|
|
|
|
19 |
|
20 |
-
The following
|
21 |
-
* /workspace/junk/singlestral5
|
22 |
-
* /workspace/junk/singlestral6
|
23 |
-
* /workspace/junk/singlestral1
|
24 |
-
* /workspace/junk/singlestral3
|
25 |
-
* /workspace/junk/singlestral4
|
26 |
-
* /workspace/junk/singlestral7
|
27 |
-
* /workspace/junk/singlestral0
|
28 |
-
* /workspace/junk/singlestral2
|
29 |
|
30 |
-
|
|
|
|
|
|
|
|
|
31 |
|
32 |
-
|
33 |
|
34 |
```yaml
|
35 |
models:
|
36 |
-
- model: /
|
37 |
-
|
38 |
-
|
39 |
-
- model: /
|
40 |
-
|
41 |
-
|
42 |
-
- model: /
|
43 |
-
|
44 |
-
weight: 0.125
|
45 |
-
- model: /workspace/junk/singlestral3
|
46 |
-
parameters:
|
47 |
-
weight: 0.125
|
48 |
-
- model: /workspace/junk/singlestral4
|
49 |
-
parameters:
|
50 |
-
weight: 0.125
|
51 |
-
- model: /workspace/junk/singlestral5
|
52 |
-
parameters:
|
53 |
-
weight: 0.125
|
54 |
-
- model: /workspace/junk/singlestral6
|
55 |
-
parameters:
|
56 |
-
weight: 0.125
|
57 |
-
- model: /workspace/junk/singlestral7
|
58 |
-
parameters:
|
59 |
-
weight: 0.125
|
60 |
merge_method: linear
|
61 |
dtype: float16
|
62 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
license: apache-2.0
|
|
|
3 |
tags:
|
4 |
+
- mixtral
|
5 |
+
- dense
|
6 |
+
- mistral
|
7 |
+
- expert
|
8 |
---
|
|
|
|
|
|
|
9 |
|
10 |
+
# Unmixtraled 22B 8x linear merge
|
|
|
11 |
|
12 |
+
> [!WARNING]
|
13 |
+
> This model outputs gibberish as it was not trained under the dense configuration. Finetuning or merging is needed to make this model useful.
|
14 |
|
15 |
+
This is a 22B Mistral model recycling weights from [mistral-community/Mixtral-8x22B-v0.1](https://huggingface.co/mistral-community/Mixtral-8x22B-v0.1).
|
16 |
+
The model was adapted from a Mixtral architecture to a dense Mistral architecture with the same number of layers, attention heads and hidden dimensions.
|
17 |
+
Embeddings, attention, layer norms and LM head weights were taken directly from the 8x22B model, MLP weights are a linear merge of experts 0 to 7 weights.
|
18 |
|
19 |
+
The following named weight correspondance was used:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
+
| Mistral weight | Mixtral weight |
|
22 |
+
|----------------|------------------------------|
|
23 |
+
| `gate_proj` | `experts.{layer_num}.w1` |
|
24 |
+
| `down_proj` | `experts.{layer_num}.w2` |
|
25 |
+
| `up_proj` | `experts.{layer_num}.w3` |
|
26 |
|
27 |
+
This mergekit configuration was used to merge the experts:
|
28 |
|
29 |
```yaml
|
30 |
models:
|
31 |
+
- model: thomasgauthier/Unmixtraled-22B-v0.1-expert-0
|
32 |
+
- model: thomasgauthier/Unmixtraled-22B-v0.1-expert-1
|
33 |
+
- model: thomasgauthier/Unmixtraled-22B-v0.1-expert-2
|
34 |
+
- model: thomasgauthier/Unmixtraled-22B-v0.1-expert-3
|
35 |
+
- model: thomasgauthier/Unmixtraled-22B-v0.1-expert-4
|
36 |
+
- model: thomasgauthier/Unmixtraled-22B-v0.1-expert-5
|
37 |
+
- model: thomasgauthier/Unmixtraled-22B-v0.1-expert-6
|
38 |
+
- model: thomasgauthier/Unmixtraled-22B-v0.1-expert-7
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
merge_method: linear
|
40 |
dtype: float16
|
41 |
```
|
42 |
+
|
43 |
+
## Unmixtraled models
|
44 |
+
| Expert | Source | Wikitext perplexity |
|
45 |
+
|--------|-----------------|---------------------|
|
46 |
+
| [Unmixtraled-22B-v0.1-expert-0](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-expert-0) | Mixtral 8x22B embed, attn, layernorm, lm_head + expert 0 MLPs | 696.6932983398438 |
|
47 |
+
| [Unmixtraled-22B-v0.1-expert-1](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-expert-1) | Mixtral 8x22B embed, attn, layernorm, lm_head + expert 1 MLPs | 6853.04248046875 |
|
48 |
+
| [Unmixtraled-22B-v0.1-expert-2](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-expert-2) | Mixtral 8x22B embed, attn, layernorm, lm_head + expert 2 MLPs | 4689.181640625 |
|
49 |
+
| [Unmixtraled-22B-v0.1-expert-3](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-expert-3) | Mixtral 8x22B embed, attn, layernorm, lm_head + expert 3 MLPs | 782.3755493164062 |
|
50 |
+
| [Unmixtraled-22B-v0.1-expert-4](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-expert-4) | Mixtral 8x22B embed, attn, layernorm, lm_head + expert 4 MLPs | 2844.943603515625 |
|
51 |
+
| [Unmixtraled-22B-v0.1-expert-5](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-expert-5) | Mixtral 8x22B embed, attn, layernorm, lm_head + expert 5 MLPs | 1099.32373046875 |
|
52 |
+
| [Unmixtraled-22B-v0.1-expert-6](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-expert-6) | Mixtral 8x22B embed, attn, layernorm, lm_head + expert 6 MLPs | 341.5309753417969 |
|
53 |
+
| [Unmixtraled-22B-v0.1-expert-7](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-expert-7) | Mixtral 8x22B embed, attn, layernorm, lm_head + expert 7 MLPs | 2099.63818359375 |
|
54 |
+
| [**Unmixtraled-22B-v0.1-lerp**](https://huggingface.co/thomasgauthier/Unmixtraled-22B-v0.1-lerp) | **Mixtral 8x22B embed, attn, layernorm, lm_head + linear merge of expert 0-7 MLPs** | **1873.9874267578125** |
|