eryk-mazus
commited on
Commit
•
4074583
1
Parent(s):
a0d9543
Update README.md
Browse files
README.md
CHANGED
@@ -13,16 +13,15 @@ pipeline_tag: text-generation
|
|
13 |
|
14 |
# polka-1.1B-dpo
|
15 |
|
16 |
-
`eryk-mazus/polka-1.1b-dpo` is the first Polish model trained to act as a helpful, conversational assistant that can be run locally
|
17 |
|
18 |
-
This model is based on [TinyLlama-1.1B](https://huggingface.co/TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T) with an extended tokenizer for more efficient Polish text generation
|
19 |
|
20 |
Context size: 4,096 tokens
|
21 |
|
22 |
-
In addition, we'
|
23 |
* [polka-1.1b](https://huggingface.co/eryk-mazus/polka-1.1b) - our base model with an extended tokenizer and additional pre-training on Polish corpus sampled using [DSIR](https://github.com/p-lambda/dsir)
|
24 |
* [polka-pretrain-en-pl-v1](https://huggingface.co/datasets/eryk-mazus/polka-pretrain-en-pl-v1) - the pre-training dataset
|
25 |
-
* [polka-1.1b-sft](https://huggingface.co/eryk-mazus/polka-1.1b-sft) - SFT version of the base model trained on polish conversations
|
26 |
* [polka-dpo-v1](https://huggingface.co/datasets/eryk-mazus/polka-dpo-v1) - dataset of DPO pairs
|
27 |
|
28 |
## Usage
|
|
|
13 |
|
14 |
# polka-1.1B-dpo
|
15 |
|
16 |
+
`eryk-mazus/polka-1.1b-dpo` is **the first Polish model trained to act as a helpful, conversational assistant that can be run locally.**
|
17 |
|
18 |
+
This model is based on [TinyLlama-1.1B](https://huggingface.co/TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T) with an extended tokenizer for more efficient Polish text generation that pretrained on an additional 6 billion Polish tokens. It was then fine-tuned using synthetically generated and machine-translated multi-turn conversations with the [Direct Preference Optimization (DPO)](https://arxiv.org/abs/2305.18290) performed on top of it.
|
19 |
|
20 |
Context size: 4,096 tokens
|
21 |
|
22 |
+
In addition, we're releasing:
|
23 |
* [polka-1.1b](https://huggingface.co/eryk-mazus/polka-1.1b) - our base model with an extended tokenizer and additional pre-training on Polish corpus sampled using [DSIR](https://github.com/p-lambda/dsir)
|
24 |
* [polka-pretrain-en-pl-v1](https://huggingface.co/datasets/eryk-mazus/polka-pretrain-en-pl-v1) - the pre-training dataset
|
|
|
25 |
* [polka-dpo-v1](https://huggingface.co/datasets/eryk-mazus/polka-dpo-v1) - dataset of DPO pairs
|
26 |
|
27 |
## Usage
|