trollek / thoughtstream

A small language model that simulates a QuietStar like behaviour in a simplified agentic way.

3B

4 Pulls Updated 8 weeks ago

... /
params
9833446801f9 · 135B
{ "num_ctx": 8192, "num_predict": 2048, "stop": [ "<|im_end|>", "<|im_start|>" ], "temperature": 0.5, "top_k": 40, "top_p": 0.85 }