Safetensors
GGUF
Turkish
llama
Llama-3
instruct
finetune
chatml
gpt4
synthetic data
distillation
function calling
json mode
axolotl
roleplaying
chat
Instructions to use tda45/TdAI with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use tda45/TdAI with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="tda45/TdAI", filename="llama.cpp/models/ggml-vocab-aquila.gguf", )
output = llm( "Once upon a time,", max_tokens=512, echo=True ) print(output)
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use tda45/TdAI with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf tda45/TdAI # Run inference directly in the terminal: llama cli -hf tda45/TdAI
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf tda45/TdAI # Run inference directly in the terminal: llama cli -hf tda45/TdAI
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf tda45/TdAI # Run inference directly in the terminal: ./llama-cli -hf tda45/TdAI
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf tda45/TdAI # Run inference directly in the terminal: ./build/bin/llama-cli -hf tda45/TdAI
Use Docker
docker model run hf.co/tda45/TdAI
- LM Studio
- Jan
- Ollama
How to use tda45/TdAI with Ollama:
ollama run hf.co/tda45/TdAI
- Unsloth Studio
How to use tda45/TdAI with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for tda45/TdAI to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for tda45/TdAI to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for tda45/TdAI to start chatting
- Atomic Chat new
- Docker Model Runner
How to use tda45/TdAI with Docker Model Runner:
docker model run hf.co/tda45/TdAI
- Lemonade
How to use tda45/TdAI with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull tda45/TdAI
Run and chat with the model
lemonade run user.TdAI-{{QUANT_TAG}}List all available models
lemonade list
| import { describe, it, expect } from 'vitest'; | |
| import { classifyContinueIntent } from '$lib/utils/agentic'; | |
| import { ContinueIntentKind, MessageRole, MessageType } from '$lib/enums'; | |
| import type { DatabaseMessage } from '$lib/types/database'; | |
| /** | |
| * Tests for the Continue button intent classifier. | |
| * | |
| * The classifier walks the persisted message history to decide which of three | |
| * resume paths a Continue click should take: | |
| * | |
| * A. append_text -> plain text assistant turn, resume with | |
| * continue_final_message. | |
| * B. rerun_turn -> assistant turn with tool_calls but no tool results yet, | |
| * the stream was cut mid turn and the tool_calls are | |
| * unrecoverable as a token level continuation. Drop the | |
| * target and rerun from the previous history. | |
| * C. next_turn -> assistant turn with tool_calls that were already | |
| * resolved by trailing tool results. Hand the history | |
| * back to the agentic loop so it starts the next turn. | |
| */ | |
| let nextId = 0; | |
| function makeMsg(role: MessageRole, opts: Partial<DatabaseMessage> = {}): DatabaseMessage { | |
| nextId++; | |
| return { | |
| id: `msg-${nextId}`, | |
| convId: 'conv-1', | |
| type: MessageType.TEXT, | |
| timestamp: nextId, | |
| role, | |
| content: '', | |
| parent: null, | |
| children: [], | |
| ...opts | |
| }; | |
| } | |
| function toolCall(id: string, name: string, args: string = '{}'): string { | |
| return JSON.stringify([{ id, type: 'function', function: { name, arguments: args } }]); | |
| } | |
| describe('classifyContinueIntent', () => { | |
| it('returns append_text for a plain text assistant turn at the tail', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'hello' }), | |
| makeMsg(MessageRole.ASSISTANT, { content: 'hi there' }) | |
| ]; | |
| const intent = classifyContinueIntent(messages, 1); | |
| expect(intent).toEqual({ kind: ContinueIntentKind.APPEND_TEXT }); | |
| }); | |
| it('returns append_text for a plain text assistant turn in the middle', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'q1' }), | |
| makeMsg(MessageRole.ASSISTANT, { content: 'a1' }), | |
| makeMsg(MessageRole.USER, { content: 'q2' }), | |
| makeMsg(MessageRole.ASSISTANT, { content: 'a2' }) | |
| ]; | |
| expect(classifyContinueIntent(messages, 1)).toEqual({ kind: ContinueIntentKind.APPEND_TEXT }); | |
| }); | |
| it('returns rerun_turn when the assistant has tool_calls without results', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'list files' }), | |
| makeMsg(MessageRole.ASSISTANT, { | |
| content: '', | |
| toolCalls: toolCall('call_1', 'bash_tool', '{"command":"ls"}') | |
| }) | |
| ]; | |
| const intent = classifyContinueIntent(messages, 1); | |
| expect(intent).toEqual({ kind: ContinueIntentKind.RERUN_TURN, truncateAfter: 0 }); | |
| }); | |
| it('returns next_turn when trailing tool results resolve the tool_calls', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'list files' }), | |
| makeMsg(MessageRole.ASSISTANT, { | |
| content: '', | |
| toolCalls: toolCall('call_1', 'bash_tool') | |
| }), | |
| makeMsg(MessageRole.TOOL, { content: 'file1\nfile2', toolCallId: 'call_1' }) | |
| ]; | |
| const intent = classifyContinueIntent(messages, 1); | |
| expect(intent).toEqual({ kind: ContinueIntentKind.NEXT_TURN, truncateAfter: 2 }); | |
| }); | |
| it('next_turn keeps all consecutive trailing tool results, not just one', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'do many things' }), | |
| makeMsg(MessageRole.ASSISTANT, { | |
| content: '', | |
| toolCalls: JSON.stringify([ | |
| { id: 'call_1', type: 'function', function: { name: 'a', arguments: '{}' } }, | |
| { id: 'call_2', type: 'function', function: { name: 'b', arguments: '{}' } } | |
| ]) | |
| }), | |
| makeMsg(MessageRole.TOOL, { content: 'r1', toolCallId: 'call_1' }), | |
| makeMsg(MessageRole.TOOL, { content: 'r2', toolCallId: 'call_2' }) | |
| ]; | |
| const intent = classifyContinueIntent(messages, 1); | |
| expect(intent).toEqual({ kind: ContinueIntentKind.NEXT_TURN, truncateAfter: 3 }); | |
| }); | |
| it('next_turn stops at the first non tool message after the target', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'go' }), | |
| makeMsg(MessageRole.ASSISTANT, { | |
| content: '', | |
| toolCalls: toolCall('call_1', 'a') | |
| }), | |
| makeMsg(MessageRole.TOOL, { content: 'r1', toolCallId: 'call_1' }), | |
| makeMsg(MessageRole.USER, { content: 'wait' }), | |
| makeMsg(MessageRole.TOOL, { content: 'late', toolCallId: 'call_1' }) | |
| ]; | |
| const intent = classifyContinueIntent(messages, 1); | |
| // truncateAfter must point at the contiguous tool block, not jump over | |
| // the user message to grab the dangling late tool. | |
| expect(intent).toEqual({ kind: ContinueIntentKind.NEXT_TURN, truncateAfter: 2 }); | |
| }); | |
| it('returns append_text when toolCalls is set but parses to empty array', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'q' }), | |
| makeMsg(MessageRole.ASSISTANT, { content: 'a', toolCalls: '[]' }) | |
| ]; | |
| expect(classifyContinueIntent(messages, 1)).toEqual({ kind: ContinueIntentKind.APPEND_TEXT }); | |
| }); | |
| it('returns append_text when toolCalls is malformed JSON', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'q' }), | |
| makeMsg(MessageRole.ASSISTANT, { content: 'a', toolCalls: '{not json' }) | |
| ]; | |
| expect(classifyContinueIntent(messages, 1)).toEqual({ kind: ContinueIntentKind.APPEND_TEXT }); | |
| }); | |
| it('returns append_text defensively when idx points at a non assistant message', () => { | |
| const messages = [ | |
| makeMsg(MessageRole.USER, { content: 'q' }), | |
| makeMsg(MessageRole.ASSISTANT, { content: 'a' }) | |
| ]; | |
| expect(classifyContinueIntent(messages, 0)).toEqual({ kind: ContinueIntentKind.APPEND_TEXT }); | |
| }); | |
| it('returns append_text defensively when idx is out of bounds', () => { | |
| const messages = [makeMsg(MessageRole.ASSISTANT, { content: 'a' })]; | |
| expect(classifyContinueIntent(messages, 5)).toEqual({ kind: ContinueIntentKind.APPEND_TEXT }); | |
| expect(classifyContinueIntent([], 0)).toEqual({ kind: ContinueIntentKind.APPEND_TEXT }); | |
| }); | |
| }); | |