MCPcopy Create free account
hub / github.com/abetlen/llama-cpp-python / GptParams

Class GptParams

examples/low_level_api/common.py:12–74  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

10
11@dataclass
12class GptParams:
13 seed: int = -1
14 n_threads: int = min(4, os.cpu_count() or 1)
15 n_predict: int = 128
16 n_parts: int = -1
17 n_ctx: int = 512
18 n_batch: int = 8
19 n_keep: int = 0
20
21 ignore_eos: bool = False
22 logit_bias: dict[int, float] = field(default_factory=dict)
23 top_k: int = 40
24 top_p: float = 0.95
25 tfs_z: float = 1.00
26 typical_p: float = 1.00
27 temp: float = 0.80
28 repeat_penalty: float = 1.10
29 repeat_last_n: int = 64
30 frequency_penalty: float = 0.0
31 presence_penalty: float = 0.0
32 mirostat: int = 0
33 mirostat_tau: float = 5.0
34 mirostat_eta: float = 0.1
35
36 model: str = "./models/llama-7B/ggml-model.bin"
37 prompt: str = ""
38 path_session: str = ""
39 input_prefix: str = " "
40 input_suffix: str = ""
41 antiprompt: List[str] = field(default_factory=list)
42
43 lora_adapter: str = ""
44 lora_base: str = ""
45
46 memory_f16: bool = True
47 random_prompt: bool = False
48 use_color: bool = False
49 interactive: bool = False
50
51 embedding: bool = False
52 interactive_start: bool = False
53
54 instruct: bool = False
55 penalize_nl: bool = True
56 perplexity: bool = False
57 use_mmap: bool = True
58 use_mlock: bool = False
59 mem_test: bool = False
60 verbose_prompt: bool = False
61
62 file: str = None
63
64 # If chat ended prematurely, append this to the conversation to fix it.
65 # Set to "\nUser:" etc.
66 # This is an alternative to input_prefix which always adds it, so it potentially duplicates "User:""
67 fix_prefix: str = ""
68 input_echo: bool = (True,)
69

Callers 4

Miku.pyFile · 0.90
ReasonAct.pyFile · 0.90
Chat.pyFile · 0.90
gpt_params_parseFunction · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected

Used in the wild real call sites across dependent graphs

searching dependent graphs…