Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,14 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
4 |
|
5 |
# rwkv7-168m-pile
|
@@ -36,7 +45,7 @@ This is RWKV-7 model under flash-linear attention format.
|
|
36 |
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
37 |
Install flash-linear-attention before using this model:
|
38 |
|
39 |
-
```
|
40 |
git clone https://github.com/fla-org/flash-linear-attention
|
41 |
cd flash-linear-attention
|
42 |
pip install -e .
|
@@ -46,7 +55,7 @@ pip install -e .
|
|
46 |
|
47 |
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
48 |
You can use this model just as any other HuggingFace models:
|
49 |
-
```
|
50 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
51 |
model = AutoModelForCausalLM.from_pretrained('fla-hub/rwkv7-168m-pile', trust_remote_code=True)
|
52 |
tokenizer = AutoTokenizer.from_pretrained('fla-hub/rwkv7-168m-pile', trust_remote_code=True)
|
@@ -67,4 +76,4 @@ This model is trained on the Pile with a total of 332 billion tokens.
|
|
67 |
#### Metrics
|
68 |
|
69 |
`lambada_openai`: ppl 14.2 acc 45.6%
|
70 |
-
`piqa`: acc 65.5%
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
datasets:
|
4 |
+
- EleutherAI/the_pile_deduplicated
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
metrics:
|
8 |
+
- accuracy
|
9 |
+
base_model:
|
10 |
+
- BlinkDL/rwkv-7-pile
|
11 |
+
pipeline_tag: text-generation
|
12 |
---
|
13 |
|
14 |
# rwkv7-168m-pile
|
|
|
45 |
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
46 |
Install flash-linear-attention before using this model:
|
47 |
|
48 |
+
```bash
|
49 |
git clone https://github.com/fla-org/flash-linear-attention
|
50 |
cd flash-linear-attention
|
51 |
pip install -e .
|
|
|
55 |
|
56 |
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
57 |
You can use this model just as any other HuggingFace models:
|
58 |
+
```python
|
59 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
60 |
model = AutoModelForCausalLM.from_pretrained('fla-hub/rwkv7-168m-pile', trust_remote_code=True)
|
61 |
tokenizer = AutoTokenizer.from_pretrained('fla-hub/rwkv7-168m-pile', trust_remote_code=True)
|
|
|
76 |
#### Metrics
|
77 |
|
78 |
`lambada_openai`: ppl 14.2 acc 45.6%
|
79 |
+
`piqa`: acc 65.5%
|