phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-06-28 02:53:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

show sampling params in model describe

Browse source
This commit is contained in:
Ashwin Bharambe 2024-07-29 23:44:07 -07:00
parent 040c30ee54
commit 7306e6b167
1 changed files with 16 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

16
llama_toolchain/cli/model/describe.py
View file

@ -7,6 +7,8 @@
import argparse import argparse
import json import json
from enum import Enum
from llama_models.llama3_1.api.sku_list import llama3_1_model_list from llama_models.llama3_1.api.sku_list import llama3_1_model_list
from termcolor import colored from termcolor import colored
@ -15,6 +17,13 @@ from llama_toolchain.cli.subcommand import Subcommand
from llama_toolchain.cli.table import print_table from llama_toolchain.cli.table import print_table
class EnumEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj, Enum):
return obj.value
return super().default(obj)
class ModelDescribe(Subcommand): class ModelDescribe(Subcommand):
"""Show details about a model""" """Show details about a model"""
@ -48,6 +57,9 @@ class ModelDescribe(Subcommand):
model = by_id[args.model_id] model = by_id[args.model_id]
sampling_params = model.recommended_sampling_params.dict()
for k in ("max_tokens", "repetition_penalty"):
del sampling_params[k]
rows = [ rows = [
( (
colored("Model", "white", attrs=["bold"]), colored("Model", "white", attrs=["bold"]),
@ -57,6 +69,10 @@ class ModelDescribe(Subcommand):
("Description", model.description_markdown), ("Description", model.description_markdown),
("Context Length", f"{model.max_seq_length // 1024}K tokens"), ("Context Length", f"{model.max_seq_length // 1024}K tokens"),
("Weights format", model.quantization_format.value), ("Weights format", model.quantization_format.value),
(
"Recommended sampling params",
json.dumps(sampling_params, cls=EnumEncoder, indent=4),
),
("Model params.json", json.dumps(model.model_args, indent=4)), ("Model params.json", json.dumps(model.model_args, indent=4)),
] ]