Update README.md

README.md

@@ -11,6 +11,86 @@ language:
 - en
 ---
 
+# adrieljleo/indonesia-function-call-lora-v2
+This model is a fine-tuned version of unsloth/Meta-Llama-3.1-8B-Instruct, trained using the Unsloth framework to support Function Calling capabilities in a Bahasa Indonesia conversational setting. It is designed for use cases where a virtual assistant interacts with users in Indonesian, while still utilizing English-based function definitions, arguments, and schemas — suitable for applications such as virtual agents, chat-based interfaces, and backend integration with APIs or tool usage.
+
+# Training Overview
+- **Base Model:** unsloth/Meta-Llama-3.1-8B-Instruct
+- **Framework:** Unsloth
+- **Languages:**
+  - Input/Output Conversations: Bahasa Indonesia
+  - Function Schemas/Descriptions: English
+
+# Datasets Trained
+This model was trained using translated and aligned function-calling datasets based on English instruction-following data:
+- adrieljleo/skripzi_func_call_revision
+  Translated version of glaiveai/glaive-function-calling-v2
+
+- adrieljleo/skripzi_parallel_revision
+  Translated version of unaidedelf87777/parallel-function_calling-10k
+
+These datasets were created and curated as part of an undergraduate thesis focusing on bridging language gaps in LLM function-calling systems.
+
+#  Features
+- Bahasa Indonesia instructions, questions, and conversations
+- Function definitions, arguments, and descriptions remain in English for compatibility with tool-based APIs and automation pipelines that exists
+- Ready for integration into applications involving Tool Use, Code Execution, or Conversational Interfaces
+
+# Chat Template
+```
+<|begin_of_text|><|start_header_id|>system<|end_header_id|>
+
+Environment: ipython
+
+Cutting Knowledge Date: December 2023
+
+Today Date: {datetime now}
+
+You have access to the following functions:
+
+Use the function '{function name}' to: {function description}
+{function tool}
+
+Think very carefully before calling functions.
+If you choose to call a function, reply in the following format:
+<{start_tag}={function_name}>{parameters}{end_tag}
+where
+
+start_tag => `<function`
+parameters => a JSON dict with the function argument name as key and function argument value as value.
+end_tag => `</function>`
+
+Here is an example,
+<function=example_function_name>{"example_name": "example_value"}</function>
+
+Reminder:
+- If looking for real time information use relevant functions before falling back to brave_search
+- Function calls MUST follow the specified format, start with <function= and end with </function>
+- Required parameters MUST be specified, if it missing then ask it
+- Function calls can be multiple parallel by having a new line each
+- Put the entire function call reply on one line
+- The arguments value MUST be the same as original query and cannot be summarized
+```
+
+The example of function tool json schema
+```
+{
+  "name": "transfer_funds",
+  "description": "Transfer money to a specified recipient.",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "recipient_name": {"type": "string", "description": "Name of the recipient"},
+      "amount": {"type": "number", "description": "Amount to transfer"}
+    },
+    "required": ["recipient_name", "amount"]
+  }
+}
+```
+
+# Statements
+This model was developed as part of an undergraduate thesis to explore the effectiveness of LLMs in Indonesian tool-use scenarios. Benchmarking has been conducted using the Berkeley Function Calling Leaderboard, where the model achieved a high score, outperforming several other open-source LLMs with Function Calling capabilities. However, the results should be interpreted with caution, as there were some misalignments in the custom benchmark classes used during evaluation. Further evaluation and refinement are planned, and updated results will be shared soon.
+
 # Uploaded  model
 
 - **Developed by:** adrieljleo