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.
- 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
- 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
- Atomic Chat
File size: 3,335 Bytes
15c3607 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 | /**
* Detects whether a model's chat template supports thinking/reasoning control.
*
* The server "/props" endpoint does NOT expose a supports_thinking flag.
* It is computed internally by common_chat_templates_support_enable_thinking
* in common/chat.cpp. A proper server flag would make this unnecessary.
*
* Detection order (most reliable first):
* 1. Thinking-control Jinja2 variables === pass-through via chat_template_kwargs
* 2. Thinking-control Jinja2 conditionals === template-native on/off logic
* 3. Paired thinking-content tag pairs === models that output special tags
*/
const THINKING_KWARG_VARS = ['enable_thinking', 'reasoning_effort', 'thinking_budget'];
/**
* Paired thinking-content tag patterns.
*
* Inspected: llama-cpp-deepseek-r1/v3, nim-nemotron-{3,4}-nano, qwen-qwq-32b,
* qwen-3-32b, google-gemma-4-31b-it, kimikimi-k2-thinking, apertus-8b-instruct,
* mistralai-Mistral-Small-3.2-24B, ByteDance-Seed-OSS.
*
* The self-closing entry is Kimi-K2, Gemma4 fixed-length pair,
* where both tags always appear adjacent with no content between.
*/
const THINKING_TAG_PATTERNS: Array<[string, string | null]> = [
['<think>', '</think>'],
['<|channel>thought', '<|channel|>'],
['<|think|>', '</|think|>'],
['<seed:think|>', '</seed:think|>'],
['<think></think>', null]
];
const JINJA_THINKING_CONDITIONALS: RegExp[] = [
// Matches: {% if enable thinking %}, {% if enable_thinking %}, {% if (enable_thinking is defined) %}
// Handles: underscore-separated (enable_thinking), space-separated (enable thinking),
// and optional parens/brackets before enable (if (enable_thinking )
/\{%-?\s*if\s+\(?\s*\w*enable[\s_]+\w*(thinking|think|reasoning)/i,
/\{%-?\s*if\s+\w*(thinking|reasoning)\s*(is not|==|!=)/i,
/\{%-?\s*if\s+not\s+\w*enable/i,
/\{%-?\s*if\s+ns\.enable_thinking/i
];
/** Guards against false positives:
* - Generic thought keyword (tool descriptions say "chain of thought")
* - Qwen vertical-bar token (used for ALL tool calls, not thinking)
*/
export function detectThinkingSupport(t: string): boolean {
if (!t) return false;
for (const kwarg of THINKING_KWARG_VARS) {
const regex = new RegExp(
`(\\{\\{[^{}]*\\b${kwarg}\\b[^{}]*\\}\\}|\\{%[^{}]*\\b${kwarg}\\b[^{}]*%\\})`,
'i'
);
if (regex.test(t)) return true;
}
for (const p of JINJA_THINKING_CONDITIONALS) {
if (p.test(t)) return true;
}
for (const [s, e] of THINKING_TAG_PATTERNS) {
if (t.includes(s) && (!e || t.includes(e))) return true;
}
return false;
}
export function detectThinkingSupportWithReason(t: string): { supported: boolean; reason: string } {
if (!t) return { supported: false, reason: 'No chat template available' };
for (const kwarg of THINKING_KWARG_VARS) {
const regex = new RegExp(
`(\\{\\{[^{}]*\\b${kwarg}\\b[^{}]*\\}\\}|\\{%[^{}]*\\b${kwarg}\\b[^{}]*%\\})`,
'i'
);
if (regex.test(t)) {
return { supported: true, reason: 'Found: ' + kwarg };
}
}
for (const p of JINJA_THINKING_CONDITIONALS) {
if (p.test(t)) return { supported: true, reason: 'Found: thinking conditional' };
}
for (const [s, e] of THINKING_TAG_PATTERNS) {
if (t.includes(s) && (!e || t.includes(e))) {
return { supported: true, reason: 'Found: ' + s + (e ? ' .. ' + e : ' (self)') };
}
}
return { supported: false, reason: 'No thinking patterns found' };
}
|