Update README.md

README.md

@@ -1,249 +1,293 @@
 ---
 library_name: transformers
-tags: []
+license: gemma
+language:
+- sl
+- en
+- hr
+- sr
+- bs
+base_model:
+- google/gemma-2-9b
+pipeline_tag: text-generation
 ---
 
-# Model Card for Model ID
+# Model Card for GaMS-9B
 
-GaMS-9B after the second round of training - high quality Slovene corpora. The training was done using NeMo 2.0. To fill the batches, the EOS token was added.
+GaMS-2B, GaMS-9B and GaMS-27B represent new improved and larger models of the GaMS (Generative Model for Slovene) familly. The models are based on Google's Gemma 2 familly and continually pretrained on Slovene, English and some portion of Croatian, Serbian and Bosnian corpora.
 
-Starting model: GaMS-Beta/GaMS-Parallel-2.0
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/652d40a78fa1fbb0aae165bb/94gX0PG8zRB_Zg31K2y_i.png)
 
-## Data
-
-| Corpus | Language | # Tokens | Percentage |
-| :----- | :------- | :------: | :--------: |
-| KAS	 | Slovene  | 2.77 B   | 20.34 %    |
-| Metafida	| Slovene | 4.66 B | 34.18 %    |
-| Wikipedia-En | English | 5.45 B |	39.99 % |
-| Wikipedia-Sl | Slovene | 0.16 B |	1.19 %  |
-| Wikipedia-Hr | Croatian | 0.15 B | 1.13 % |
-| Wikipedia-Bs | Bosnian | 0.07 B | 0.50 %  |
-| Wikipedia-Sr-Latin | Serbian | 0.36 B | 2.68 % |
-| Total | | 13.62 B | |
-
-## Slovenian-LLM-Eval results
-
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/652d40a78fa1fbb0aae165bb/YszAknPaoxiBR0c_7Gu3U.png)
-
-## Slobench Results
-
-The reported results were obtained using guided decoding.
-
-### 0-shot results
-
-|   | Model                                  | BoolQ_accuracy    | MultiRC_exact_match   | MultiRC_per_question_f1   | MultiRC_f1_over_all_answers   | WSC_accuracy      | COPA_accuracy     | RTE_accuracy      | CB_accuracy       | CB_f1             | NLI_accuracy      | NLI_precision_entailment   | NLI_recall_entailment   | NLI_f1_entailment   | NLI_precision_neutral   | NLI_recall_neutral   | NLI_f1_neutral    | NLI_precision_contradiction   | NLI_recall_contradiction   | NLI_f1_contradiction   |
-|--:|:---------------------------------------|:------------------|:----------------------|:--------------------------|:------------------------------|:------------------|:------------------|:------------------|:------------------|:------------------|:------------------|:---------------------------|:------------------------|:--------------------|:------------------------|:---------------------|:------------------|:------------------------------|:---------------------------|:-----------------------|
-| 0 | /models/hf_models/GaMS-9B-Parallel-2.0 | 0.76 [0.74, 0.77] | 0.21 [0.19, 0.24]     | 0.56 [0.54, 0.58]         | 0.55 [0.53, 0.57]             | 0.38 [0.28, 0.47] | 0.6 [0.5, 0.7]    | 0.6 [0.54, 0.66]  | 0.68 [0.55, 0.8]  | 0.59 [0.39, 0.76] | 0.35 [0.31, 0.39] | 0.5 [0.23, 0.77]           | 0.04 [0.02, 0.07]       | 0.08 [0.03, 0.13]   | 0.33 [0.29, 0.37]       | 0.85 [0.79, 0.9]     | 0.47 [0.42, 0.52] | 0.41 [0.31, 0.52]             | 0.19 [0.13, 0.24]          | 0.26 [0.18, 0.32]      |
-| 1 | google/gemma-2-9b                      | 0.78 [0.77, 0.8]  | 0.2 [0.18, 0.23]      | 0.48 [0.45, 0.5]          | 0.49 [0.48, 0.51]             | 0.63 [0.54, 0.73] | 0.68 [0.59, 0.77] | 0.69 [0.63, 0.74] | 0.55 [0.42, 0.69] | 0.31 [0.22, 0.41] | 0.33 [0.29, 0.37] | 0.55 [0.25, 0.83]          | 0.03 [0.01, 0.06]       | 0.06 [0.02, 0.11]   | 0.32 [0.28, 0.36]       | 0.98 [0.95, 0.99]    | 0.48 [0.44, 0.52] | 0.57 [0.14, 1.0]              | 0.02 [0.01, 0.04]          | 0.04 [0.01, 0.08]      |
-| 2 | google/gemma-2-9b-it                   | 0.83 [0.82, 0.84] | 0.18 [0.16, 0.2]      | 0.6 [0.58, 0.62]          | 0.5 [0.49, 0.52]              | 0.62 [0.52, 0.71] | 0.86 [0.79, 0.93] | 0.8 [0.75, 0.85]  | 0.82 [0.72, 0.92] | 0.73 [0.58, 0.85] | 0.47 [0.43, 0.51] | 0.48 [0.28, 0.68]          | 0.06 [0.03, 0.1]        | 0.11 [0.05, 0.17]   | 0.4 [0.35, 0.46]        | 0.66 [0.59, 0.74]    | 0.5 [0.44, 0.56]  | 0.55 [0.49, 0.62]             | 0.72 [0.66, 0.78]          | 0.62 [0.57, 0.68]      |
-| 3 | zlatorog/Zlatorog_SFT_v2               | 0.82 [0.81, 0.84] | 0.35 [0.32, 0.38]     | 0.72 [0.7, 0.74]          | 0.71 [0.69, 0.72]             | 0.65 [0.56, 0.75] | 0.79 [0.71, 0.87] | 0.79 [0.74, 0.84] | 0.77 [0.65, 0.88] | 0.63 [0.46, 0.82] | 0.63 [0.59, 0.67] | 0.54 [0.48, 0.59]          | 0.93 [0.9, 0.96]        | 0.68 [0.63, 0.73]   | 0.8 [0.25, 1.0]         | 0.02 [0.01, 0.05]    | 0.04 [0.01, 0.1]  | 0.79 [0.73, 0.84]             | 0.9 [0.86, 0.94]           | 0.84 [0.8, 0.88]       |
-| 4 | cjvt/GaMS-1B                           | 0.52 [0.5, 0.54]  | 0.01 [0.0, 0.01]      | 0.05 [0.04, 0.05]         | 0.04 [0.04, 0.05]             | 0.61 [0.51, 0.7]  | 0.55 [0.45, 0.65] | 0.48 [0.42, 0.54] | 0.21 [0.1, 0.33]  | 0.19 [0.09, 0.27] | 0.33 [0.29, 0.36] | 0.34 [0.28, 0.4]           | 0.43 [0.36, 0.5]        | 0.38 [0.32, 0.44]   | 0.33 [0.25, 0.41]       | 0.26 [0.2, 0.33]     | 0.29 [0.22, 0.35] | 0.3 [0.23, 0.37]              | 0.28 [0.21, 0.34]          | 0.29 [0.22, 0.35]      |
-| 5 | cjvt/GaMS-1B-Chat                      | 0.62 [0.61, 0.64] | 0.03 [0.02, 0.04]     | 0.13 [0.12, 0.14]         | 0.07 [0.07, 0.08]             | 0.5 [0.4, 0.6]    | 0.55 [0.45, 0.65] | 0.47 [0.41, 0.53] | 0.5 [0.36, 0.64]  | 0.22 [0.18, 0.26] | 0.35 [0.31, 0.39] | 0.35 [0.31, 0.39]          | 0.99 [0.98, 1.0]        | 0.52 [0.47, 0.56]   | 0.0 [0.0, 0.0]          | 0.0 [0.0, 0.0]       | 0.0 [0.0, 0.0]    | 0.0 [0.0, 0.0]                | 0.0 [0.0, 0.0]             | 0.0 [0.0, 0.0]         |
-| 6 | utter-project/EuroLLM-9B               | 0.77 [0.76, 0.79] | 0.12 [0.1, 0.14]      | 0.53 [0.52, 0.55]         | 0.53 [0.52, 0.55]             | 0.55 [0.45, 0.65] | 0.78 [0.7, 0.86]  | 0.73 [0.68, 0.79] | 0.79 [0.67, 0.9]  | 0.55 [0.48, 0.62] | 0.39 [0.34, 0.43] | 0.64 [0.33, 1.0]           | 0.04 [0.01, 0.07]       | 0.07 [0.02, 0.12]   | 0.34 [0.3, 0.38]        | 0.95 [0.92, 0.98]    | 0.5 [0.45, 0.54]  | 0.87 [0.76, 0.96]             | 0.22 [0.15, 0.28]          | 0.35 [0.26, 0.42]      |
-| 7 | utter-project/EuroLLM-9B-Instruct      | 0.81 [0.79, 0.82] | 0.18 [0.15, 0.2]      | 0.64 [0.62, 0.65]         | 0.64 [0.62, 0.66]             | 0.6 [0.5, 0.69]   | 0.58 [0.48, 0.68] | 0.82 [0.77, 0.87] | 0.77 [0.65, 0.88] | 0.63 [0.46, 0.82] | 0.38 [0.34, 0.42] | 0.46 [0.41, 0.52]          | 0.76 [0.69, 0.82]       | 0.57 [0.52, 0.62]   | 0.24 [0.18, 0.3]        | 0.21 [0.15, 0.26]    | 0.22 [0.16, 0.28] | 0.31 [0.21, 0.42]             | 0.14 [0.09, 0.2]           | 0.2 [0.13, 0.26]       |
-| 8 | /models/hf_models/GaMS-9B-SecondRound  | 0.78 [0.76, 0.79] | 0.23 [0.21, 0.26]     | 0.63 [0.61, 0.65]         | 0.53 [0.52, 0.55]             | 0.59 [0.49, 0.68] | 0.65 [0.55, 0.75] | 0.77 [0.72, 0.82] | 0.68 [0.55, 0.8]  | 0.64 [0.52, 0.76] | 0.46 [0.42, 0.5]  | 0.49 [0.42, 0.57]          | 0.45 [0.38, 0.52]       | 0.47 [0.41, 0.53]   | 0.33 [0.25, 0.4]        | 0.29 [0.23, 0.36]    | 0.31 [0.24, 0.37] | 0.53 [0.46, 0.6]              | 0.62 [0.56, 0.7]           | 0.57 [0.51, 0.63]      |
-
-### 3-shot results
-
-|   | Model                                  | BoolQ_accuracy    | MultiRC_exact_match   | MultiRC_per_question_f1   | MultiRC_f1_over_all_answers   | WSC_accuracy      | COPA_accuracy     | RTE_accuracy      | CB_accuracy       | CB_f1             | NLI_accuracy      | NLI_precision_entailment   | NLI_recall_entailment   | NLI_f1_entailment   | NLI_precision_neutral   | NLI_recall_neutral   | NLI_f1_neutral    | NLI_precision_contradiction   | NLI_recall_contradiction   | NLI_f1_contradiction   |
-|--:|:---------------------------------------|:------------------|:----------------------|:--------------------------|:------------------------------|:------------------|:------------------|:------------------|:------------------|:------------------|:------------------|:---------------------------|:------------------------|:--------------------|:------------------------|:---------------------|:------------------|:------------------------------|:---------------------------|:-----------------------|
-| 0 | /models/hf_models/GaMS-9B-Parallel-2.0 | 0.83 [0.81, 0.84] | 0.36 [0.33, 0.39]     | 0.74 [0.72, 0.75]         | 0.74 [0.73, 0.76]             | 0.64 [0.55, 0.74] | 0.87 [0.8, 0.94]  | 0.78 [0.73, 0.83] | 0.84 [0.74, 0.94] | 0.59 [0.53, 0.64] | 0.48 [0.43, 0.52] | 0.38 [0.21, 0.56]          | 0.07 [0.03, 0.11]       | 0.11 [0.06, 0.18]   | 0.37 [0.32, 0.42]       | 0.67 [0.6, 0.74]     | 0.48 [0.42, 0.53] | 0.66 [0.59, 0.73]             | 0.72 [0.66, 0.78]          | 0.69 [0.63, 0.75]      |
-| 1 | google/gemma-2-9b                      | 0.82 [0.81, 0.83] | 0.37 [0.34, 0.4]      | 0.75 [0.73, 0.77]         | 0.75 [0.74, 0.77]             | 0.66 [0.57, 0.76] | 0.88 [0.82, 0.94] | 0.79 [0.74, 0.84] | 0.88 [0.79, 0.96] | 0.61 [0.56, 0.65] | 0.48 [0.44, 0.52] | 0.52 [0.37, 0.65]          | 0.15 [0.1, 0.19]        | 0.23 [0.15, 0.29]   | 0.37 [0.32, 0.42]       | 0.77 [0.71, 0.83]    | 0.5 [0.45, 0.55]  | 0.75 [0.68, 0.82]             | 0.56 [0.48, 0.63]          | 0.64 [0.57, 0.71]      |
-| 2 | google/gemma-2-9b-it                   | 0.84 [0.83, 0.85] | 0.15 [0.13, 0.18]     | 0.66 [0.64, 0.67]         | 0.65 [0.64, 0.67]             | 0.7 [0.61, 0.79]  | 0.89 [0.83, 0.95] | 0.82 [0.78, 0.87] | 0.84 [0.74, 0.94] | 0.75 [0.59, 0.88] | 0.65 [0.61, 0.69] | 0.66 [0.6, 0.72]           | 0.77 [0.71, 0.83]       | 0.71 [0.66, 0.76]   | 0.56 [0.46, 0.68]       | 0.28 [0.22, 0.35]    | 0.37 [0.3, 0.45]  | 0.68 [0.61, 0.74]             | 0.88 [0.83, 0.92]          | 0.76 [0.72, 0.81]      |
-| 3 | zlatorog/Zlatorog_SFT_v2               | 0.83 [0.82, 0.84] | 0.04 [0.02, 0.05]     | 0.54 [0.52, 0.55]         | 0.55 [0.53, 0.56]             | 0.51 [0.41, 0.61] | 0.66 [0.57, 0.75] | 0.73 [0.67, 0.78] | 0.7 [0.57, 0.82]  | 0.48 [0.4, 0.56]  | 0.56 [0.52, 0.6]  | 0.55 [0.49, 0.62]          | 0.62 [0.55, 0.69]       | 0.58 [0.53, 0.64]   | 0.39 [0.31, 0.49]       | 0.24 [0.19, 0.31]    | 0.3 [0.23, 0.37]  | 0.64 [0.58, 0.71]             | 0.79 [0.73, 0.85]          | 0.71 [0.66, 0.76]      |
-| 4 | cjvt/GaMS-1B                           | 0.49 [0.47, 0.5]  | 0.07 [0.06, 0.09]     | 0.41 [0.39, 0.43]         | 0.37 [0.35, 0.39]             | 0.58 [0.48, 0.67] | 0.49 [0.39, 0.59] | 0.47 [0.41, 0.53] | 0.43 [0.29, 0.56] | 0.21 [0.17, 0.25] | 0.32 [0.28, 0.36] | 0.36 [0.27, 0.44]          | 0.22 [0.16, 0.29]       | 0.27 [0.21, 0.34]   | 0.31 [0.27, 0.36]       | 0.77 [0.7, 0.83]     | 0.44 [0.39, 0.49] | 0.0 [0.0, 0.0]                | 0.0 [0.0, 0.0]             | 0.0 [0.0, 0.0]         |
-| 5 | cjvt/GaMS-1B-Chat                      | 0.59 [0.57, 0.61] | 0.04 [0.02, 0.05]     | 0.34 [0.32, 0.36]         | 0.16 [0.15, 0.16]             | 0.63 [0.54, 0.73] | 0.55 [0.45, 0.65] | 0.47 [0.41, 0.53] | 0.43 [0.29, 0.56] | 0.2 [0.16, 0.24]  | 0.36 [0.32, 0.4]  | 0.36 [0.32, 0.4]           | 0.99 [0.97, 1.0]        | 0.53 [0.48, 0.57]   | 0.47 [0.2, 0.73]        | 0.04 [0.01, 0.07]    | 0.07 [0.02, 0.13] | 0.0 [0.0, 0.0]                | 0.0 [0.0, 0.0]             | 0.0 [0.0, 0.0]         |
-| 6 | utter-project/EuroLLM-9B               | 0.81 [0.79, 0.82] | 0.17 [0.15, 0.2]      | 0.61 [0.59, 0.63]         | 0.58 [0.57, 0.6]              | 0.65 [0.56, 0.75] | 0.63 [0.53, 0.73] | 0.73 [0.67, 0.78] | 0.73 [0.61, 0.85] | 0.59 [0.47, 0.71] | 0.45 [0.41, 0.49] | 1.0 [0.0, 1.0]             | 0.02 [0.0, 0.04]        | 0.03 [0.0, 0.07]    | 0.38 [0.32, 0.45]       | 0.53 [0.46, 0.61]    | 0.44 [0.38, 0.51] | 0.49 [0.44, 0.55]             | 0.83 [0.78, 0.88]          | 0.62 [0.57, 0.67]      |
-| 7 | utter-project/EuroLLM-9B-Instruct      | 0.81 [0.79, 0.82] | 0.07 [0.05, 0.08]     | 0.52 [0.5, 0.53]          | 0.53 [0.51, 0.55]             | 0.65 [0.56, 0.75] | 0.69 [0.6, 0.78]  | 0.79 [0.74, 0.84] | 0.79 [0.67, 0.9]  | 0.71 [0.56, 0.84] | 0.42 [0.38, 0.47] | 0.61 [0.53, 0.69]          | 0.48 [0.41, 0.55]       | 0.53 [0.47, 0.6]    | 0.31 [0.26, 0.36]       | 0.57 [0.5, 0.64]     | 0.4 [0.35, 0.45]  | 0.54 [0.43, 0.65]             | 0.23 [0.17, 0.29]          | 0.32 [0.25, 0.39]      |
-| 8 | /models/hf_models/GaMS-9B-SecondRound  | 0.8 [0.78, 0.81]  | 0.15 [0.13, 0.17]     | 0.64 [0.62, 0.65]         | 0.58 [0.57, 0.6]              | 0.66 [0.57, 0.76] | 0.9 [0.84, 0.96]  | 0.82 [0.77, 0.87] | 0.84 [0.74, 0.94] | 0.73 [0.56, 0.87] | 0.56 [0.52, 0.6]  | 0.58 [0.5, 0.66]           | 0.38 [0.31, 0.46]       | 0.46 [0.39, 0.53]   | 0.42 [0.35, 0.5]        | 0.44 [0.37, 0.51]    | 0.43 [0.37, 0.49] | 0.65 [0.59, 0.71]             | 0.86 [0.8, 0.91]           | 0.74 [0.69, 0.79]      |
-
-
-## Model Details
-
-### Model Description
-
-<!-- Provide a longer summary of what this model is. -->
-
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+## Acknowledgment
 
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
+The model was developed within the [PoVeJMo](https://www.cjvt.si/povejmo/en/project/) research program (Adaptive Natural Language Processing with Large Language Models), particularly within the research project titled SloLLaMai -- Open-access computationally efficient models for Slovenian. The program is funded within the Recovery and Resilience Plan by the Slovenian Research and Innovation Agency (ARIS) and NextGenerationEU. The authors also acknowledge the financial support from the Slovenian Research and Innovation Agency (research core funding No. P6-0411 -- Language Resources and Technologies for Slovene).
 
-### Model Sources [optional]
+We thank everyone who worked on data collection and preparation, enabling us to train our model. Special thanks go to Nikola Ljubešić, Taja Kuzman, Tjaša Arčon, Jaka Čibej, Simon Krek, Tomaž Erjavec, Iztok Kosem and Tomaž Savodnik.
 
-<!-- Provide the basic links for the model. -->
+## Basic information
 
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+- **Developed by:** team of researchers at the University of Ljubljana, Faculty for Computer and Information Science. Team members: Domen Vreš, Iztok Lebar Bajec, Tjaša Arčon, Gašper Jelovčan and Marko Robnik-Šikonja.
+- **Languages:** Slovene, English (primary), Croatian, Bosnian and Serbian (secondary). The model might also work for other languages supported by Gemma 2, even though it was not continually pretrained on them.
+- **Base model:** [google/gemma2-9b](https://huggingface.co/google/gemma-2-9b)
+- **License:** [Gemma](https://ai.google.dev/gemma/terms)
 
-## Uses
+## Usage
 
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+The model can be run through `pipeline` API using the following code:
 
-### Direct Use
+```python
+from transformers import pipeline
 
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+model_id = "cjvt/GaMS-9B"
 
-[More Information Needed]
+pline = pipeline(
+    "text-generation",
+    model=model_id,
+    device_map="cuda" # replace with "mps" to run on a Mac device
+)
 
-### Downstream Use [optional]
+prompts = [
+    "The examples of antonyms are:\nhigh => low\nwide => narrow\nbig =>",
+    "Pristanek je bil prvi nadzorovani spust ameriškega vesoljskega plovila na površje Lune po Apollu 17 leta 1972, ko je na Luni pristala zadnja Nasina misija s posadko.\nDoslej so na Luni pristala vesoljska plovila le iz štirih drugih držav –",
+    "U četvrtak je bila prva polufinalna večer Dore, a komentari na društvenim mrežama ne prestaju. U nedjeljno finale prošli su:"
+]
 
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+sequences = pline(
+    prompts,
+    max_new_tokens=512,
+    num_return_sequences=1
+)
 
-[More Information Needed]
+for seq in sequences:
+    print("--------------------------")
+    print(f"Result: {seq[0]['generated_text']}")
+    print("--------------------------\n")
+```
 
-### Out-of-Scope Use
+For multi GPU inference set the `device_map` to `auto`:
 
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+```python
+from transformers import pipeline
 
-[More Information Needed]
+model_id = "cjvt/GaMS-9B"
 
-## Bias, Risks, and Limitations
+pline = pipeline(
+    "text-generation",
+    model=model_id,
+    device_map="auto"
+)
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+prompts = [
+    "The examples of antonyms are:\nhigh => low\nwide => narrow\nbig =>",
+    "Pristanek je bil prvi nadzorovani spust ameriškega vesoljskega plovila na površje Lune po Apollu 17 leta 1972, ko je na Luni pristala zadnja Nasina misija s posadko.\nDoslej so na Luni pristala vesoljska plovila le iz štirih drugih držav –",
+    "U četvrtak je bila prva polufinalna večer Dore, a komentari na društvenim mrežama ne prestaju. U nedjeljno finale prošli su:"
+]
 
-[More Information Needed]
+sequences = pline(
+    prompts,
+    max_new_tokens=512,
+    num_return_sequences=1
+)
 
-### Recommendations
+for seq in sequences:
+    print("--------------------------")
+    print(f"Result: {seq[0]['generated_text']}")
+    print("--------------------------\n")
+```
 
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-
-## How to Get Started with the Model
-
-Use the code below to get started with the model.
-
-[More Information Needed]
-
-## Training Details
-
-### Training Data
-
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-
-[More Information Needed]
-
-### Training Procedure
-
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+## Data
 
-#### Preprocessing [optional]
+### CPT Data
 
-[More Information Needed]
+Model was continually pre-trained in two stages. In the first stage, parallel English-Slovene (and Croatian in some cases) corpora was used to align the languages. In the second stage, the model was trained on separate English, Slovene, Croatian, Bosnian and Serbian corpora.
 
+#### Parallel alignment corpora
 
-#### Training Hyperparameters
+| Corpus | Alignment level | # Tokens | Percentage |
+| :----- | :------- | :------: | :--------: |
+| KAS Abstracts | Document level | 31 M | 1.65 % |
+| DGT | Separate documents | 697 M | 36.56 % |
+| MaCoCu Parallel | Separate documents | 430 M | 22.53 % |
+| CC-News | Paragraph level | 749 M | 39.25 % |
+| Total | | 1.91 B | |
 
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+Explanation of each alignment level:
+- Document level: Parallel documents were concatenated into a single document
+- Separate documents: Parallel documents were not explicitly aligned
+- Paragraph level: Paragraphs of parallel documents were concatenated (the first paragraph of Slovene/English document was followed by the first paragraph in the other language, which was then followed by the second paragraph in the first language and so on)
 
-#### Speeds, Sizes, Times [optional]
+#### Second stage corpora
 
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+| Corpus | Language | # Tokens | Percentage |
+| :----- | :------- | :------: | :--------: |
+| [KAS](https://www.clarin.si/repository/xmlui/handle/11356/1448)	 | Slovene  | 2.77 B   | 20.34 %    |
+| [MetaFida](https://www.clarin.si/repository/xmlui/handle/11356/1775)*	| Slovene | 4.66 B | 34.18 %    |
+| [Wikipedia-En](https://huggingface.co/datasets/zidsi/wikipedia_markdown) (Date: January 23rd 2025) | English | 5.45 B |	39.99 % |
+| [Wikipedia-Sl](https://huggingface.co/datasets/zidsi/wikipedia_markdown) (Date: January 1st 2025) | Slovene | 0.16 B |	1.19 %  |
+| [Wikipedia-Hr](https://huggingface.co/datasets/zidsi/wikipedia_markdown) (Date: January 1st 2025) | Croatian | 0.15 B | 1.13 % |
+| [Wikipedia-Bs](https://huggingface.co/datasets/zidsi/wikipedia_markdown) (Date: January 1st 2025) | Bosnian | 0.07 B | 0.50 %  |
+| [Wikipedia-Sr-Latin](https://huggingface.co/datasets/zidsi/wikipedia_markdown)*  | Serbian | 0.36 B | 2.68 % |
+| Total | | 13.62 B | |
 
-[More Information Needed]
+Remarks:
+- The following corpora was excluded from MetaFida: dgt15_sl, classlawiki_sl, tweet_sl, janes_tweet, janes_forum, janes_news
+- Serbian Wikipedia was converted from Cyrillic to Latin
 
 ## Evaluation
 
-<!-- This section describes the evaluation protocols and provides the results. -->
-
-### Testing Data, Factors & Metrics
-
-#### Testing Data
-
-<!-- This should link to a Dataset Card if possible. -->
-
-[More Information Needed]
+The models were evaluated using [Slovene SuperGLUE](https://slobench.cjvt.si/leaderboard/view/3) collection of classification tasks on [SloBench](https://slobench.cjvt.si). Instruct version of the model was also evaluated on translation [from English to Slovene](https://slobench.cjvt.si/leaderboard/view/8) and [from Slovene to English](https://slobench.cjvt.si/leaderboard/view/7) Additionally, we evaluated our models on [Slovenian-LLM-Eval](https://huggingface.co/datasets/cjvt/slovenian-llm-eval).
 
-#### Factors
+Code for evaluation:
+- [SloBench tasks](https://github.com/SloLama/slobench_evaluation)
+- [Slovenian-LLM-Eval](https://github.com/SloLama/slovenian-llm-eval)
 
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-
-[More Information Needed]
-
-#### Metrics
-
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-
-[More Information Needed]
-
-### Results
-
-[More Information Needed]
-
-#### Summary
-
-
-
-## Model Examination [optional]
-
-<!-- Relevant interpretability work for the model goes here -->
-
-[More Information Needed]
-
-## Environmental Impact
-
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-
-## Technical Specifications [optional]
-
-### Model Architecture and Objective
-
-[More Information Needed]
-
-### Compute Infrastructure
-
-[More Information Needed]
-
-#### Hardware
-
-[More Information Needed]
-
-#### Software
-
-[More Information Needed]
-
-## Citation [optional]
-
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-
-**BibTeX:**
-
-[More Information Needed]
-
-**APA:**
-
-[More Information Needed]
-
-## Glossary [optional]
-
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-
-[More Information Needed]
-
-## More Information [optional]
-
-[More Information Needed]
+## Slovenian-LLM-Eval results
 
-## Model Card Authors [optional]
+Comparison between GaMS models, base Gemma 2 models and SlovenianGPT (open source model for Slovene based on Mistral 7B) is shown in the figure below. All models were evaluated in 0-shot scenario.
 
-[More Information Needed]
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/652d40a78fa1fbb0aae165bb/tDyAjB2dgYXv1dLpFHikd.png)
 
-## Model Card Contact
+## Slobench Results
 
-[More Information Needed]
+GaMS 2B, 9B and 27B models were evaluated in 3-shot scenario, except for MultiRC and translation tasks, where 0-shot was used. GaMS-9B-Instruct was evaluated in 0-shot scenarion on all tasks. We used guided decoding to ensure the correct format of the responses.
+
+### Slovene SuperGLUE
+
+| Rank | Title                 | Average | BoolQ Accuracy | CB Accuracy | CB F1 Score | CB Average | COPA Accuracy | MultiRC EM | MultiRC F1a Score | MultiRC Average | RTE Accuracy | WSC Accuracy |
+|------|------------------------|---------|---------------|-------------|-------------|------------|--------------|------------|----------------|----------------|-------------|-------------|
+| 1    | GaMS-27B              | 0.7601  | 0.8333        | 0.6440      | 0.5864      | 0.6152     | 0.9540       | 0.3904     | 0.7504         | 0.5704         | 0.7931      | 0.7945      |
+| 2    | PrešernGPT 0.1        | 0.7568  | 0.8333        | 0.8520      | 0.5868      | 0.7194     | 0.9740       | 0.4985     | 0.8061         | 0.6523         | 0.8276      | 0.5342      |
+| 3    | Gemma 2 27B           | 0.7546  | 0.8333        | 0.6680      | 0.5972      | 0.6326     | 0.9140       | 0.4174     | 0.7295         | 0.5735         | 0.8276      | 0.7466      |
+| 4    | GaMS-9B               | 0.7309  | 0.7000        | 0.8400      | 0.7955      | 0.8178     | 0.9000       | 0.3243     | 0.6551         | 0.4897         | 0.7931      | 0.6849      |
+| 5    | GaMS-9B-Instruct      | 0.6997  | 0.8000        | 0.7960      | 0.7128      | 0.7544     | 0.8140       | 0.0721     | 0.6174         | 0.3447         | 0.7931      | 0.6918      |
+| 6    | Gemma 2 9B            | 0.6980  | 0.8333        | 0.8240      | 0.5683      | 0.6962     | 0.8700       | 0.2282     | 0.5310         | 0.3796         | 0.7241      | 0.6849      |
+| 8    | CroSloEngual BERT     | 0.6078  | 0.7333        | 0.7920      | 0.7437      | 0.7679     | 0.5720       | 0.0931     | 0.5241         | 0.3086         | 0.6552      | 0.6096      |
+| 11   | SlovenianGPT-Chat     | 0.5078  | 0.7333        | 0.3920      | 0.3829      | 0.3874     | 0.6840       | 0.2432     | 0.4944         | 0.3688         | 0.5172      | 0.3562      |
+| 12   | Gemma 2 2B            | 0.4876  | 0.6333        | 0.4520      | 0.2123      | 0.3321     | 0.5180       | 0.1471     | 0.4419         | 0.2945         | 0.5862      | 0.5616      |
+| 13   | GaMS-2B               | 0.4790  | 0.5667        | 0.6080      | 0.4880      | 0.5480     | 0.5240       | 0.0631     | 0.5234         | 0.2932         | 0.5517      | 0.3904      |
+| 14   | GaMS-1B               | 0.4604  | 0.5000        | 0.6200      | 0.4565      | 0.5382     | 0.4920       | 0.1351     | 0.2675         | 0.2013         | 0.4828      | 0.5479      |
+| 15   | GaMS-1B-Chat          | 0.4570  | 0.8000        | 0.4880      | 0.3023      | 0.3951     | 0.4840       | 0.1081     | 0.2428         | 0.1755         | 0.5172      | 0.3692      |
+
+
+### English to Slovene translation (first 11 models on the benchmark)
+
+| Rank | Title                           | BERT score | BLEU (avg) | METEOR (avg) | CHRF (avg) | BLEU (corpus) | CHRF (corpus) |
+|------|---------------------------------|------------|------------|--------------|------------|---------------|---------------|
+| 1    | DeepL Translator                | 0.8812     | 0.3153     | 0.5902       | 0.6205     | 0.3599        | 0.6205        |
+| 2    | gemini-1.5-pro                  | 0.8791     | 0.3124     | 0.5895       | 0.6176     | 0.3569        | 0.6176        |
+| 3    | Sonnet 3.5                      | 0.8789     | 0.3059     | 0.5783       | 0.6204     | 0.3442        | 0.6204        |
+| 4    | gpt-4o                          | 0.8784     | 0.2958     | 0.5811       | 0.6138     | 0.3379        | 0.6138        |
+| 5    | EuroLLM-9B-Instruct             | 0.8741     | 0.2927     | 0.5792       | 0.6055     | 0.3273        | 0.6055        |
+| 6    | seamless-m4t-v2-large           | 0.8731     | 0.2780     | 0.5599       | 0.5947     | 0.3085        | 0.5947        |
+| 7    | **GaMS-9B-Instruct**            | 0.8713     | 0.2773     | 0.5616       | 0.5928     | 0.3209        | 0.5928        |
+| 8    | Zlatorog                        | 0.8706     | 0.2834     | 0.5633       | 0.6014     | 0.2903        | 0.6014        |
+| 9    | RSDO-DS4-NMT 1.2.2              | 0.8705     | 0.2794     | 0.5634       | 0.5956     | 0.3226        | 0.5956        |
+| 9    | META LLAMA 3.1 405B             | 0.8705     | 0.2637     | 0.5497       | 0.5930     | 0.3063        | 0.5930        |
+| 11   | RSDO-DS4-NMT 1.2                | 0.8698     | 0.2781     | 0.5602       | 0.5970     | 0.3177        | 0.5970        |
+
+### Slovene to English translation (first 10 models on the benchmark)
+
+| Rank | Title               | BERT score | BLEU (avg) | METEOR (avg) | CHRF (avg) | BLEU (corpus) | CHRF (corpus) |
+|------|---------------------|------------|------------|--------------|------------|---------------|---------------|
+| 1    | gpt-4o             | 0.9496     | 0.3161     | 0.6655       | 0.6297     | 0.3496        | 0.6297        |
+| 2    | gemini-1.5-pro     | 0.9489     | 0.3117     | 0.6560       | 0.6237     | 0.3502        | 0.6237        |
+| 3    | gpt-4o-mini        | 0.9466     | 0.2976     | 0.6493       | 0.6197     | 0.3328        | 0.6197        |
+| 4    | **GaMS-9B-Instruct**   | 0.9454     | 0.2821     | 0.6275       | 0.6018     | 0.3141        | 0.6018        |
+| 5    | ChatGPTv1         | 0.9449     | 0.2852     | 0.6415       | 0.6096     | 0.3171        | 0.6096        |
+| 6    | RSDO-DS4-NMT 1.2.4 | 0.9434     | 0.2839     | 0.6227       | 0.5967     | 0.3290        | 0.5967        |
+| 7    | RSDO-DS4-NMT 1.2.6 | 0.9433     | 0.2832     | 0.6207       | 0.5944     | 0.3295        | 0.5944        |
+| 8    | RSDO-DS4-NMT 1.2.2 | 0.9431     | 0.2785     | 0.6184       | 0.5933     | 0.3240        | 0.5933        |
+| 8    | RSDO-DS4-NMT 1.2   | 0.9431     | 0.2805     | 0.6201       | 0.5941     | 0.3231        | 0.5941        |
+| 10   | eTranslation SLEN  | 0.9414     | 0.2729     | 0.6175       | 0.5913     | 0.3119        | 0.5913        |
+
+
+## Usage and Limitations (taken from Gemma 2)
+
+These models have certain limitations that users should be aware of.
+
+### Intended Usage
+
+Open Large Language Models (LLMs) have a wide range of applications across
+various industries and domains. The following list of potential uses is not
+comprehensive. The purpose of this list is to provide contextual information
+about the possible use-cases that the model creators considered as part of model
+training and development.
+
+* Content Creation and Communication
+  * Text Generation: These models can be used to generate creative text formats
+    such as poems, scripts, code, marketing copy, and email drafts.
+  * Chatbots and Conversational AI: Power conversational interfaces for customer
+    service, virtual assistants, or interactive applications.
+  * Text Summarization: Generate concise summaries of a text corpus, research
+    papers, or reports.
+* Research and Education
+  * Natural Language Processing (NLP) Research: These models can serve as a
+    foundation for researchers to experiment with NLP techniques, develop
+    algorithms, and contribute to the advancement of the field.
+  * Language Learning Tools: Support interactive language learning experiences,
+    aiding in grammar correction or providing writing practice.
+  * Knowledge Exploration: Assist researchers in exploring large bodies of text
+    by generating summaries or answering questions about specific topics.
+
+### Limitations
+
+* Training Data
+  * The quality and diversity of the training data significantly influence the
+    model's capabilities. Biases or gaps in the training data can lead to
+    limitations in the model's responses.
+  * The scope of the training dataset determines the subject areas the model can
+    handle effectively.
+* Context and Task Complexity
+  * LLMs are better at tasks that can be framed with clear prompts and
+    instructions. Open-ended or highly complex tasks might be challenging.
+  * A model's performance can be influenced by the amount of context provided
+    (longer context generally leads to better outputs, up to a certain point).
+* Language Ambiguity and Nuance
+  * Natural language is inherently complex. LLMs might struggle to grasp subtle
+    nuances, sarcasm, or figurative language.
+* Factual Accuracy
+  * LLMs generate responses based on information they learned from their
+    training datasets, but they are not knowledge bases. They may generate
+    incorrect or outdated factual statements.
+* Common Sense
+  * LLMs rely on statistical patterns in language. They might lack the ability
+    to apply common sense reasoning in certain situations.
+
+### Ethical Considerations and Risks
+
+The development of large language models (LLMs) raises several ethical concerns.
+In creating an open model, we have carefully considered the following:
+
+* Bias and Fairness
+  * LLMs trained on large-scale, real-world text data can reflect socio-cultural
+    biases embedded in the training material. These models underwent careful
+    scrutiny, input data pre-processing described and posterior evaluations
+    reported in this card.
+* Misinformation and Misuse
+  * LLMs can be misused to generate text that is false, misleading, or harmful.
+  * Guidelines are provided for responsible use with the model, see the
+    [Responsible Generative AI Toolkit][rai-toolkit].
+* Transparency and Accountability:
+  * This model card summarizes details on the models' architecture,
+    capabilities, limitations, and evaluation processes.
+  * A responsibly developed open model offers the opportunity to share
+    innovation by making LLM technology accessible to developers and researchers
+    across the AI ecosystem.
+
+Risks identified and mitigations:
+
+* Perpetuation of biases: It's encouraged to perform continuous monitoring
+  (using evaluation metrics, human review) and the exploration of de-biasing
+  techniques during model training, fine-tuning, and other use cases.
+* Generation of harmful content: Mechanisms and guidelines for content safety
+  are essential. Developers are encouraged to exercise caution and implement
+  appropriate content safety safeguards based on their specific product policies
+  and application use cases.
+* Misuse for malicious purposes: Technical limitations and developer and
+  end-user education can help mitigate against malicious applications of LLMs.
+  Educational resources and reporting mechanisms for users to flag misuse are
+  provided. Prohibited uses of Gemma models are outlined in the
+  [Gemma Prohibited Use Policy][prohibited-use].
+* Privacy violations: Models were trained on data filtered for removal of PII
+  (Personally Identifiable Information). Developers are encouraged to adhere to
+  privacy regulations with privacy-preserving techniques.