anocha-mcomp commited on
Commit
4d01f76
·
verified ·
1 Parent(s): b3d731b

Update model card from Confluence: Model Card: Gemma-SEA-LION-v4-4B-VL

Browse files
model_card_gemma-sea-lion-v4-4b-vl.md CHANGED
@@ -4,9 +4,9 @@
4
 
5
  SEA-LION is a collection of Large Language Models (LLMs) which have been pretrained and instruct-tuned for the Southeast Asia (SEA) region.
6
 
7
- **Gemma-SEA-LION-v4-4B-VL** is a 4-billion parameter Vision-Language Model (VLM) built upon the gemma-3-4b-it architecture. To ensure **domain adaptation** for the region, the model underwent rigorous post-training on a curated dataset of approximately 8.54 **million** instruction-text pairs.
8
 
9
- This extensive post-training instills **multilingual** and **multicultural** fluency, covering English and 4 key SEA languages: Burmese, Malay, Tagalog and Tamil.
10
 
11
  Gemma-SEA-LION-v4-4B-VL inherits the following features from gemma-3-4b-it:
12
 
@@ -93,21 +93,26 @@ print(output[0]["generated_text"][-1]["content"])
93
 
94
  The instruction fine-tuning text dataset comprises of a collection of OSS & synthetic data.
95
 
96
- **Datasets for linguistic training**:
97
 
98
  - <https://huggingface.co/datasets/aisingapore/filtered_joined_prompt_v1_resp_qwen_32b>
99
  - <https://huggingface.co/datasets/aisingapore/translated_english_500k_formatted_resp_qwen32b>
100
 
101
- **Datasets for tool-calling**:
102
 
103
  - <https://huggingface.co/datasets/allenai/Dolci-Instruct-SFT-Tool-Use>
104
  - <https://huggingface.co/datasets/Agent-Ark/Toucan-1.5M>
105
 
 
 
 
 
 
106
  ### Training Procedure
107
 
108
  #### Training Hyperparameters
109
 
110
- - **Training regime:** Our post-training workflow consists of instruction fine-tuning and model merging.
111
 
112
  ## Evaluation
113
 
 
4
 
5
  SEA-LION is a collection of Large Language Models (LLMs) which have been pretrained and instruct-tuned for the Southeast Asia (SEA) region.
6
 
7
+ **Gemma-SEA-LION-v4-4B-VL** is a 4-billion parameter Vision-Language Model (VLM) built upon the gemma-3-4b-it architecture. To ensure **domain adaptation** for the region, the model underwent rigorous post-training on a curated dataset of approximately **8.54 million** instruction-text pairs.
8
 
9
+ This extensive post-training instills **multilingual** and **multicultural** fluency, covering English and key SEA languages with a heavy emphasis on Burmese, Malay, Tagalog and Tamil.
10
 
11
  Gemma-SEA-LION-v4-4B-VL inherits the following features from gemma-3-4b-it:
12
 
 
93
 
94
  The instruction fine-tuning text dataset comprises of a collection of OSS & synthetic data.
95
 
96
+ **Datasets for Instruction Fine Tuning**:
97
 
98
  - <https://huggingface.co/datasets/aisingapore/filtered_joined_prompt_v1_resp_qwen_32b>
99
  - <https://huggingface.co/datasets/aisingapore/translated_english_500k_formatted_resp_qwen32b>
100
 
101
+ **Datasets for Tool-calling**:
102
 
103
  - <https://huggingface.co/datasets/allenai/Dolci-Instruct-SFT-Tool-Use>
104
  - <https://huggingface.co/datasets/Agent-Ark/Toucan-1.5M>
105
 
106
+ **Datasets for Reinforcement Learning:**
107
+
108
+ - Instruction following: <https://huggingface.co/datasets/nvidia/Nemotron-Cascade-RL-Instruction-Following>
109
+ - Tool: <https://huggingface.co/datasets/openai/gsm8k>
110
+
111
  ### Training Procedure
112
 
113
  #### Training Hyperparameters
114
 
115
+ - **Training regime:** Our post-training workflow consists of Instruction Fine-tuning, Model Merging and Reinforcement Learning.
116
 
117
  ## Evaluation
118