dnrops
/
transformers
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

QDQBert example update (#16395) 

* update Dockerfile and utils_qa

* Update README.md
This commit is contained in:
Shang Zhang 2022-03-28 02:47:52 -07:00 committed by GitHub
parent f6f6866e9e
commit 7ecbb9c5e4
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 18 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
examples/research_projects/quantization-qdqbert/Dockerfile
View File

@ -12,7 +12,7 @@
# 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.
FROM nvcr.io/nvidia/pytorch:21.07-py3
FROM nvcr.io/nvidia/pytorch:22.02-py3
LABEL maintainer="Hugging Face"
LABEL repository="transformers"
@ -20,11 +20,7 @@ RUN apt-get update
RUN apt-get install sudo
RUN python3 -m pip install --no-cache-dir --upgrade pip
RUN python3 -m pip install --no-cache-dir --ignore-installed ruamel.yaml \
mkl \
absl-py \
yamlpy \
tensorboardX
RUN python3 -m pip install --no-cache-dir --ignore-installed pycuda
RUN python3 -m pip install --no-cache-dir \
pytorch-quantization --extra-index-url https://pypi.ngc.nvidia.com
@ -34,4 +30,4 @@ RUN cd transformers/ && \
python3 -m pip install --no-cache-dir .
RUN python3 -m pip install --no-cache-dir datasets \
accelerate
accelerate

3
examples/research_projects/quantization-qdqbert/README.md
View File

@ -39,9 +39,6 @@ Run the docker:
docker run --gpus all --privileged --rm -it --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864 bert_quantization:latest
```
*Note that the current NGC pytorch container (pytorch:21.07-py3) has TensorRT 8.0 which doesn't meet the requiremnt of TensorRT >= 8.2. One can either update the Dockerfile with the latest [NGC pytorch container](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/pytorch) once it supports TensorRT 8.2, or manually download and install [TensorRT >= 8.2](https://developer.nvidia.com/nvidia-tensorrt-download) in the container.*
In the container:
```
cd transformers/examples/research_projects/quantization-qdqbert/

23
examples/research_projects/quantization-qdqbert/utils_qa.py
View File

@ -73,10 +73,12 @@ def postprocess_qa_predictions(
log_level (:obj:`int`, `optional`, defaults to ``logging.WARNING``):
``logging`` log level (e.g., ``logging.WARNING``)
"""
assert len(predictions) == 2, "`predictions` should be a tuple with two elements (start_logits, end_logits)."
if len(predictions) != 2:
raise ValueError("`predictions` should be a tuple with two elements (start_logits, end_logits).")
all_start_logits, all_end_logits = predictions
assert len(predictions[0]) == len(features), f"Got {len(predictions[0])} predictions and {len(features)} features."
if len(predictions[0]) != len(features):
raise ValueError(f"Got {len(predictions[0])} predictions and {len(features)} features.")
# Build a map example to its corresponding features.
example_id_to_index = {k: i for i, k in enumerate(examples["id"])}
@ -135,7 +137,9 @@ def postprocess_qa_predictions(
start_index >= len(offset_mapping)
or end_index >= len(offset_mapping)
or offset_mapping[start_index] is None
or len(offset_mapping[start_index]) < 2
or offset_mapping[end_index] is None
or len(offset_mapping[end_index]) < 2
):
continue
# Don't consider answers with a length that is either < 0 or > max_answer_length.
@ -145,6 +149,7 @@ def postprocess_qa_predictions(
# provided).
if token_is_max_context is not None and not token_is_max_context.get(str(start_index), False):
continue
prelim_predictions.append(
{
"offsets": (offset_mapping[start_index][0], offset_mapping[end_index][1]),
@ -212,7 +217,8 @@ def postprocess_qa_predictions(
# If we have an output_dir, let's save all those dicts.
if output_dir is not None:
assert os.path.isdir(output_dir), f"{output_dir} is not a directory."
if not os.path.isdir(output_dir):
raise EnvironmentError(f"{output_dir} is not a directory.")
prediction_file = os.path.join(
output_dir, "predictions.json" if prefix is None else f"{prefix}_predictions.json"
@ -283,12 +289,12 @@ def postprocess_qa_predictions_with_beam_search(
log_level (:obj:`int`, `optional`, defaults to ``logging.WARNING``):
``logging`` log level (e.g., ``logging.WARNING``)
"""
assert len(predictions) == 5, "`predictions` should be a tuple with five elements."
if len(predictions) != 5:
raise ValueError("`predictions` should be a tuple with five elements.")
start_top_log_probs, start_top_index, end_top_log_probs, end_top_index, cls_logits = predictions
assert len(predictions[0]) == len(
features
), f"Got {len(predictions[0])} predicitions and {len(features)} features."
if len(predictions[0]) != len(features):
raise ValueError(f"Got {len(predictions[0])} predictions and {len(features)} features.")
# Build a map example to its corresponding features.
example_id_to_index = {k: i for i, k in enumerate(examples["id"])}
@ -400,7 +406,8 @@ def postprocess_qa_predictions_with_beam_search(
# If we have an output_dir, let's save all those dicts.
if output_dir is not None:
assert os.path.isdir(output_dir), f"{output_dir} is not a directory."
if not os.path.isdir(output_dir):
raise EnvironmentError(f"{output_dir} is not a directory.")
prediction_file = os.path.join(
output_dir, "predictions.json" if prefix is None else f"{prefix}_predictions.json"