Spaces:
Running
Running
File size: 1,808 Bytes
2852136 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 |
# Copyright 2024 the LlamaFactory team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
from transformers.utils import is_flash_attn_2_available, is_torch_sdpa_available
from llamafactory.hparams import get_infer_args
from llamafactory.model import load_model, load_tokenizer
TINY_LLAMA = os.environ.get("TINY_LLAMA", "llamafactory/tiny-random-Llama-3")
INFER_ARGS = {
"model_name_or_path": TINY_LLAMA,
"template": "llama3",
}
def test_attention():
attention_available = ["off"]
if is_torch_sdpa_available():
attention_available.append("sdpa")
if is_flash_attn_2_available():
attention_available.append("fa2")
llama_attention_classes = {
"off": "LlamaAttention",
"sdpa": "LlamaSdpaAttention",
"fa2": "LlamaFlashAttention2",
}
for requested_attention in attention_available:
model_args, _, finetuning_args, _ = get_infer_args({"flash_attn": requested_attention, **INFER_ARGS})
tokenizer_module = load_tokenizer(model_args)
model = load_model(tokenizer_module["tokenizer"], model_args, finetuning_args)
for module in model.modules():
if "Attention" in module.__class__.__name__:
assert module.__class__.__name__ == llama_attention_classes[requested_attention]
|