Skip to content

feat: add QA and MaskedLM task for FP8 encoder instantiation #148

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

Merged
chichun-charlie-liu merged 1 commit into from
Jul 7, 2025
Merged

feat: add QA and MaskedLM task for FP8 encoder instantiation #148

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
26 changes: 22 additions & 4 deletions fms_mo/run_quant.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,11 @@
from datasets import load_from_disk
from huggingface_hub.errors import HFValidationError
from torch.cuda import OutOfMemoryError
from transformers import AutoTokenizer
from transformers import (
AutoModelForMaskedLM,
AutoModelForQuestionAnswering,
AutoTokenizer,
)
import torch
import transformers

Expand Down Expand Up @@ -204,9 +208,23 @@ def run_fp8(model_args, data_args, opt_args, fp8_args):

logger = set_log_level(opt_args.log_level, "fms_mo.run_fp8")

model = SparseAutoModelForCausalLM.from_pretrained(
model_args.model_name_or_path, torch_dtype=model_args.torch_dtype
)
if model_args.task_type == "lm":
model = SparseAutoModelForCausalLM.from_pretrained(
model_args.model_name_or_path,
torch_dtype=model_args.torch_dtype,
)
elif model_args.task_type == "qa":
model = AutoModelForQuestionAnswering.from_pretrained(
model_args.model_name_or_path,
torch_dtype=model_args.torch_dtype,
)
elif model_args.task_type == "mlm":
model = AutoModelForMaskedLM.from_pretrained(
model_args.model_name_or_path,
torch_dtype=model_args.torch_dtype,
)
else:
raise ValueError(f"Unsupported task: {model_args.task_type}")
tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path)

recipe = QuantizationModifier(
Expand Down
11 changes: 11 additions & 0 deletions fms_mo/training_args.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,17 @@ class ModelArguments(TypeChecker):
"""Dataclass for model related arguments."""

model_name_or_path: str = field(default="facebook/opt-125m")
task_type: str = field(
default="lm",
metadata={
"choices": ["lm", "qa", "mlm"],
"help": (
"Instantiate model for selected task: 'lm' (language modeling), 'qa' "
"(question answering, for encoders), 'mlm' (masked language modeling, "
"for encoders)."
),
},
)
torch_dtype: str = field(default="bfloat16")
device_map: Optional[str] = field(
default=None,
Expand Down
Loading