Tool Calling for firefunction-v2 GGUF

[odrobnik]

Which version of LM Studio?
LM Studio 0.3.6

Which operating system?
macOS

What is the bug?
Sending a chat completion request to LM Studio for firefunction-v2, including Tools results in

Error: received prediction-error t.fromSerializedError (/Applications/LM Studio.app/Contents/Resources/app/.webpack/main/index.js:382:62222) at m.o (/Applications/LM Studio.app/Contents/Resources/app/.webpack/main/index.js:25:155460) at m.emit (node:events:519:28) at m.onChildMessage (/Applications/LM Studio.app/Contents/Resources/app/.webpack/main/index.js:25:140859) at m.onChildMessage (/Applications/LM Studio.app/Contents/Resources/app/.webpack/main/index.js:25:164118) at ForkUtilityProcess.<anonymous> (/Applications/LM Studio.app/Contents/Resources/app/.webpack/main/index.js:25:140194) at ForkUtilityProcess.emit (node:events:519:28) at ForkUtilityProcess.a.emit (node:electron/js2c/browser_init:2:71823) - Caused By: SyntaxError: Unknown statement type: Identifier at /Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:33:115985 at i (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:33:116050) at R (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:33:120440) at new se (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:33:137373) at /Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:10:36458 at async t.formatPrompt (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:10:34152) at async t.getContextOverflowPolicy (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:10:49555) at async t.temp_createLlamaPredictionArgs (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:8:366412) at async t.LLMEngineWrapper.predictTokens (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:10:26226) at async Object.predictTokens (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:15:7181)

Screenshots

Additional Info

For reference: tool calling works ok using the same model and ollama.

firefunction-v2 is a fine tune of Llama 3 to support multiple tool calls in parallel by making one JSON with multiple function calls.

[yagil]

Thanks @odrobnik ! We're investigating