kind = "hf2q.config"
schema_version = 2
package = "hf2q"
[convert]
quant = "q4_k_m"
[serve]
host = "127.0.0.1"
port = 8081
scheduler = "inflight_batched"
max_slots = 1
repetition_penalty = 1.05
thinking_token_budget = 2048
tool_thinking_token_budget = 512