Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

more ffn multipliers added #854

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

more ffn multipliers added #854

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b57acf8
more ffn multipliers added
Nov 22, 2024
6955785
using model name, non-existing architectures will crash
Nov 22, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 22 additions & 7 deletions src/fairseq2/recipes/llama/convert_checkpoint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@

from typing_extensions import override

from fairseq2.assets import get_asset_store
from fairseq2.console import get_error_console
from fairseq2.dependency import DependencyResolver
from fairseq2.logging import get_log_writer
Expand All @@ -35,9 +36,9 @@ class ConvertCheckpointCommandHandler(CliCommandHandler):
@override
def init_parser(self, parser: ArgumentParser, resolver: DependencyResolver) -> None:
parser.add_argument(
"--arch",
"--model",
metavar="ARCH_NAME",
help="architecture name to generate params.json",
help="model name to fetch architecture to generate params.json",
)

parser.add_argument(
Expand All @@ -64,8 +65,11 @@ def __call__(self, args: Namespace, resolver: DependencyResolver) -> None:

sys.exit(1)

if args.arch:
model_config = load_llama_config(args.arch)
asset_store = get_asset_store()
arch = asset_store.retrieve_card(args.model).field("model_arch").as_(str)

if arch:
model_config = load_llama_config(args.model)
else:
model_config = None

Expand Down Expand Up @@ -153,8 +157,19 @@ def __call__(self, args: Namespace, resolver: DependencyResolver) -> None:
if model_config.num_attn_heads != model_config.num_key_value_heads:
params["model"]["n_kv_heads"] = model_config.num_key_value_heads

if args.arch == "llama2_70b" or args.arch.startswith("llama3"):
params["model"]["ffn_dim_multiplier"] = 1.3
# we only specify archs where multiplier != 1.0
ffn_dim_multipliers = {
"llama2_70b": 1.3,
"llama3_8b": 1.3,
"llama3_70b": 1.3,
"llama3_1_8b": 1.3,
"llama3_1_70b": 1.3,
"llama3_1_405b": 1.2,
"llama3_2_1b": 1.5,
}

if arch in ffn_dim_multipliers:
params["model"]["ffn_dim_multiplier"] = ffn_dim_multipliers[arch]

try:
with args.output_dir.joinpath("params.json").open("w") as fp:
Expand All @@ -164,4 +179,4 @@ def __call__(self, args: Namespace, resolver: DependencyResolver) -> None:

sys.exit(1)

log.info("params.json generated for {}.", args.arch)
log.info("params.json generated for {}.", args.model)
Loading