Update README.md
Browse files
README.md
CHANGED
@@ -25,12 +25,14 @@ _Fast Inference with Customization:_ Once trained, the ELM model architecture pe
|
|
25 |
## ELM-v0.1 Model Release
|
26 |
This repository contains code to run our ELM models. The current ELM model `elm-v0.1` (named _Rambutan_) was pre-trained (an intermediate checkpoint was used) and then instruction fine-tuned for downstream tasks.
|
27 |
|
28 |
-
|
29 |
- news_classification
|
30 |
- toxicity_detection
|
31 |
- news_content_generation
|
32 |
- news_summarization
|
33 |
|
|
|
|
|
34 |
## Setup ELM
|
35 |
### Download ELM repo
|
36 |
```bash
|
|
|
25 |
## ELM-v0.1 Model Release
|
26 |
This repository contains code to run our ELM models. The current ELM model `elm-v0.1` (named _Rambutan_) was pre-trained (an intermediate checkpoint was used) and then instruction fine-tuned for downstream tasks.
|
27 |
|
28 |
+
ELM models (in the `models` folder) in this repository come in three sizes (`elm-1.0`, `elm-0.75` and `elm-0.25`). **All these different slices are extracted from the same ELM finetuned checkpoint for inference** and supports the following use-case.
|
29 |
- news_classification
|
30 |
- toxicity_detection
|
31 |
- news_content_generation
|
32 |
- news_summarization
|
33 |
|
34 |
+
**NOTE: ELM-v0.1 release is an early version finetuned from an intermediate pretrained checkpoint & without any KV caching, decoding optimizations, or quantization applied.**
|
35 |
+
|
36 |
## Setup ELM
|
37 |
### Download ELM repo
|
38 |
```bash
|