Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The initialization of
data_type
in DeepSpeedMLP appears to be incorrect whenint8
inference is requested, leading to seemingly wrong text generation.For instance, running the following command (using inference-test.py from the DeepSpeedExamples repo)
deepspeed --num_gpus 1 inference/huggingface/text-generation/inference-test.py --model bigscience/bloom-7b1 --batch_size 1 --use_kernel --use_meta_tensor --dtype int8
produces the following output:
in=DeepSpeed is a machine learning framework
out=DeepSpeed is a machine learning framework 6 10 tràm kú Tá aixướm de de截útil恩opció 15 Beraz Tá P L Tá 7 80pata terbakar aixjai尼 Tá l'Oficina de al 3'hi aix-Els L Tá T Haut-Commissariat Tá 80juntament de de a 20 10 kú
The proposed fix modifies the change to the
data_type
initialization introduced in #3425