Merge pull request huggingface#63 from jamesthesnake/ra

Ra

.github/workflows/build-docker-images.yml

@@ -71,6 +71,16 @@ jobs:
     name: "Latest PyTorch + DeepSpeed"
     runs-on: ubuntu-latest
     steps:
+      - name: Cleanup disk
+        run: |
+          sudo ls -l /usr/local/lib/
+          sudo ls -l /usr/share/
+          sudo du -sh /usr/local/lib/
+          sudo du -sh /usr/share/
+          sudo rm -rf /usr/local/lib/android
+          sudo rm -rf /usr/share/dotnet
+          sudo du -sh /usr/local/lib/
+          sudo du -sh /usr/share/
       -
         name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v2
@@ -98,6 +108,16 @@ jobs:
     name: "Latest PyTorch + DeepSpeed (Push CI - Daily Build)"
     runs-on: ubuntu-latest
     steps:
+      - name: Cleanup disk
+        run: |
+          sudo ls -l /usr/local/lib/
+          sudo ls -l /usr/share/
+          sudo du -sh /usr/local/lib/
+          sudo du -sh /usr/share/
+          sudo rm -rf /usr/local/lib/android
+          sudo rm -rf /usr/share/dotnet
+          sudo du -sh /usr/local/lib/
+          sudo du -sh /usr/share/
       -
         name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v2

README.md

@@ -115,6 +115,19 @@ In Multimodal tasks:
 
 **[Write With Transformer](https://transformer.huggingface.co)**, built by the Hugging Face team, is the official demo of this repo’s text generation capabilities.
 
+
+## 100 projects using Transformers
+
+Transformers is more than a toolkit to use pretrained models: it's a community of projects built around it and the 
+Hugging Face Hub. We want Transformers to enable developers, researchers, students, professors, engineers, and anyone 
+else to build their dream projects.
+
+In order to celebrate the 100,000 stars of transformers, we have decided to put the spotlight on the
+community, and we have created the [awesome-transformers](./awesome-transformers.md) page which lists 100
+incredible projects built in the vicinity of transformers.
+
+If you own or use a project that you believe should be part of the list, please open a PR to add it!
+
 ## If you are looking for custom support from the Hugging Face team
 
 <a target="_blank" href="https://huggingface.co/support">

docker/transformers-all-latest-gpu/Dockerfile

@@ -35,7 +35,7 @@ RUN [ "$PYTORCH" != "pre" ] && python3 -m pip install --no-cache-dir -U $VERSION
 RUN python3 -m pip install --no-cache-dir -U tensorflow==2.12 protobuf==3.20.3 tensorflow_text tensorflow_probability
 RUN python3 -m pip uninstall -y flax jax
 
-RUN python3 -m pip install --no-cache-dir intel_extension_for_pytorch==$INTEL_TORCH_EXT+cpu -f https://software.intel.com/ipex-whl-stable
+RUN python3 -m pip install --no-cache-dir intel_extension_for_pytorch==$INTEL_TORCH_EXT+cpu -f https://developer.intel.com/ipex-whl-stable-cpu
 
 RUN python3 -m pip install --no-cache-dir git+https://github.com/facebookresearch/detectron2.git pytesseract
 RUN python3 -m pip install -U "itsdangerous<2.1.0"

docker/transformers-pytorch-deepspeed-latest-gpu/Dockerfile

@@ -1,12 +1,12 @@
 # https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel_22-08.html#rel_22-08
-FROM nvcr.io/nvidia/pytorch:22.08-py3
+FROM nvcr.io/nvidia/pytorch:22.12-py3
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
 
 ARG PYTORCH='2.0.1'
 # Example: `cu102`, `cu113`, etc.
-ARG CUDA='cu117'
+ARG CUDA='cu118'
 
 RUN apt -y update
 RUN apt install -y libaio-dev
@@ -15,6 +15,8 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip
 ARG REF=main
 RUN git clone https://github.com/huggingface/transformers && cd transformers && git checkout $REF
 
+RUN python3 -m pip uninstall -y torch torchvision torchaudio
+
 # Install latest release PyTorch
 # (PyTorch must be installed before pre-compiling any DeepSpeed c++/cuda ops.)
 # (https://www.deepspeed.ai/tutorials/advanced-install/#pre-install-deepspeed-ops)

docker/transformers-pytorch-deepspeed-nightly-gpu/Dockerfile

@@ -1,11 +1,11 @@
 # https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel_22-08.html#rel_22-08
-FROM nvcr.io/nvidia/pytorch:22.08-py3
+FROM nvcr.io/nvidia/pytorch:22.12-py3
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
 
 # Example: `cu102`, `cu113`, etc.
-ARG CUDA='cu117'
+ARG CUDA='cu118'
 
 RUN apt -y update
 RUN apt install -y libaio-dev
@@ -14,6 +14,8 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip
 ARG REF=main
 RUN git clone https://github.com/huggingface/transformers && cd transformers && git checkout $REF
 
+RUN python3 -m pip uninstall -y torch torchvision torchaudio
+
 # Install **nightly** release PyTorch (flag `--pre`)
 # (PyTorch must be installed before pre-compiling any DeepSpeed c++/cuda ops.)
 # (https://www.deepspeed.ai/tutorials/advanced-install/#pre-install-deepspeed-ops)

docker/transformers-pytorch-gpu/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:11.7.1-cudnn8-devel-ubuntu20.04
+FROM nvidia/cuda:11.8.0-cudnn8-devel-ubuntu20.04
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
@@ -16,7 +16,7 @@ ARG PYTORCH='2.0.1'
 ARG TORCH_VISION=''
 ARG TORCH_AUDIO=''
 # Example: `cu102`, `cu113`, etc.
-ARG CUDA='cu117'
+ARG CUDA='cu118'
 
 RUN [ ${#PYTORCH} -gt 0 ] && VERSION='torch=='$PYTORCH'.*' ||  VERSION='torch'; python3 -m pip install --no-cache-dir -U $VERSION --extra-index-url https://download.pytorch.org/whl/$CUDA
 RUN [ ${#TORCH_VISION} -gt 0 ] && VERSION='torchvision=='TORCH_VISION'.*' ||  VERSION='torchvision'; python3 -m pip install --no-cache-dir -U $VERSION --extra-index-url https://download.pytorch.org/whl/$CUDA

src/transformers/modeling_utils.py

@@ -207,22 +207,29 @@ def get_parameter_dtype(parameter: Union[nn.Module, GenerationMixin, "ModuleUtil
         # if no floating dtype was found return whatever the first dtype is
         return last_dtype
 
-    else:
-        # For nn.DataParallel compatibility in PyTorch > 1.5
-        def find_tensor_attributes(module: nn.Module) -> List[Tuple[str, Tensor]]:
-            tuples = [(k, v) for k, v in module.__dict__.items() if torch.is_tensor(v)]
-            return tuples
-
-        gen = parameter._named_members(get_members_fn=find_tensor_attributes)
-        last_tuple = None
-        for tuple in gen:
-            last_tuple = tuple
-            if tuple[1].is_floating_point():
-                return tuple[1].dtype
-
+    # For nn.DataParallel compatibility in PyTorch > 1.5
+    def find_tensor_attributes(module: nn.Module) -> List[Tuple[str, Tensor]]:
+        tuples = [(k, v) for k, v in module.__dict__.items() if torch.is_tensor(v)]
+        return tuples
+
+    gen = parameter._named_members(get_members_fn=find_tensor_attributes)
+    last_tuple = None
+    for tuple in gen:
+        last_tuple = tuple
+        if tuple[1].is_floating_point():
+            return tuple[1].dtype
+
+    if last_tuple is not None:
         # fallback to the last dtype
         return last_tuple[1].dtype
 
+    # fallback to buffer dtype
+    for t in parameter.buffers():
+        last_dtype = t.dtype
+        if t.is_floating_point():
+            return t.dtype
+    return last_dtype
+
 
 def get_state_dict_float_dtype(state_dict):
     """

src/transformers/models/albert/modeling_tf_albert.py

@@ -44,7 +44,7 @@
     keras_serializable,
     unpack_inputs,
 )
-from ...tf_utils import shape_list, stable_softmax
+from ...tf_utils import check_embeddings_within_bounds, shape_list, stable_softmax
 from ...utils import (
     MULTIPLE_CHOICE_DUMMY_INPUTS,
     ModelOutput,
@@ -188,16 +188,7 @@ def call(
             raise ValueError("Need to provide either `input_ids` or `input_embeds`.")
 
         if input_ids is not None:
-            # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-            # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-            tf.debugging.assert_less(
-                input_ids,
-                tf.cast(self.config.vocab_size, dtype=input_ids.dtype),
-                message=(
-                    "input_ids must be smaller than the embedding layer's input dimension (got"
-                    f" {tf.math.reduce_max(input_ids)} >= {self.config.vocab_size})"
-                ),
-            )
+            check_embeddings_within_bounds(input_ids, self.config.vocab_size)
             inputs_embeds = tf.gather(params=self.weight, indices=input_ids)
 
         input_shape = shape_list(inputs_embeds)[:-1]

src/transformers/models/bart/modeling_tf_bart.py

@@ -40,7 +40,7 @@
     keras_serializable,
     unpack_inputs,
 )
-from ...tf_utils import shape_list, stable_softmax
+from ...tf_utils import check_embeddings_within_bounds, shape_list, stable_softmax
 from ...utils import (
     ContextManagers,
     add_code_sample_docstrings,
@@ -763,16 +763,7 @@ def call(
             if hasattr(self.embed_tokens, "load_weight_prefix"):
                 context.append(tf.name_scope(self.embed_tokens.load_weight_prefix + "/"))
             with ContextManagers(context):
-                # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-                # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-                tf.debugging.assert_less(
-                    input_ids,
-                    tf.cast(self.embed_tokens.input_dim, dtype=input_ids.dtype),
-                    message=(
-                        "input_ids must be smaller than the embedding layer's input dimension (got"
-                        f" {tf.math.reduce_max(input_ids)} >= {self.embed_tokens.input_dim})"
-                    ),
-                )
+                check_embeddings_within_bounds(input_ids, self.embed_tokens.input_dim)
                 inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale
 
         embed_pos = self.embed_positions(input_shape)
@@ -965,16 +956,7 @@ def call(
             if hasattr(self.embed_tokens, "load_weight_prefix"):
                 context.append(tf.name_scope(self.embed_tokens.load_weight_prefix + "/"))
             with ContextManagers(context):
-                # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-                # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-                tf.debugging.assert_less(
-                    input_ids,
-                    tf.cast(self.embed_tokens.input_dim, dtype=input_ids.dtype),
-                    message=(
-                        "input_ids must be smaller than the embedding layer's input dimension (got"
-                        f" {tf.math.reduce_max(input_ids)} >= {self.embed_tokens.input_dim})"
-                    ),
-                )
+                check_embeddings_within_bounds(input_ids, self.embed_tokens.input_dim)
                 inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale
 
         hidden_states = inputs_embeds

src/transformers/models/bert/modeling_tf_bert.py

@@ -49,7 +49,7 @@
     keras_serializable,
     unpack_inputs,
 )
-from ...tf_utils import shape_list, stable_softmax
+from ...tf_utils import check_embeddings_within_bounds, shape_list, stable_softmax
 from ...utils import (
     DUMMY_INPUTS,
     MULTIPLE_CHOICE_DUMMY_INPUTS,
@@ -198,16 +198,7 @@ def call(
             raise ValueError("Need to provide either `input_ids` or `input_embeds`.")
 
         if input_ids is not None:
-            # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-            # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-            tf.debugging.assert_less(
-                input_ids,
-                tf.cast(self.config.vocab_size, dtype=input_ids.dtype),
-                message=(
-                    "input_ids must be smaller than the embedding layer's input dimension (got"
-                    f" {tf.math.reduce_max(input_ids)} >= {self.config.vocab_size})"
-                ),
-            )
+            check_embeddings_within_bounds(input_ids, self.config.vocab_size)
             inputs_embeds = tf.gather(params=self.weight, indices=input_ids)
 
         input_shape = shape_list(inputs_embeds)[:-1]

src/transformers/models/blenderbot/modeling_tf_blenderbot.py

@@ -38,7 +38,7 @@
     keras_serializable,
     unpack_inputs,
 )
-from ...tf_utils import shape_list, stable_softmax
+from ...tf_utils import check_embeddings_within_bounds, shape_list, stable_softmax
 from ...utils import (
     ContextManagers,
     add_code_sample_docstrings,
@@ -746,16 +746,7 @@ def call(
             if hasattr(self.embed_tokens, "load_weight_prefix"):
                 context.append(tf.name_scope(self.embed_tokens.load_weight_prefix + "/"))
             with ContextManagers(context):
-                # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-                # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-                tf.debugging.assert_less(
-                    input_ids,
-                    tf.cast(self.embed_tokens.input_dim, dtype=input_ids.dtype),
-                    message=(
-                        "input_ids must be smaller than the embedding layer's input dimension (got"
-                        f" {tf.math.reduce_max(input_ids)} >= {self.embed_tokens.input_dim})"
-                    ),
-                )
+                check_embeddings_within_bounds(input_ids, self.embed_tokens.input_dim)
                 inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale
 
         embed_pos = self.embed_positions(input_shape)
@@ -956,16 +947,7 @@ def call(
             if hasattr(self.embed_tokens, "load_weight_prefix"):
                 context.append(tf.name_scope(self.embed_tokens.load_weight_prefix + "/"))
             with ContextManagers(context):
-                # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-                # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-                tf.debugging.assert_less(
-                    input_ids,
-                    tf.cast(self.embed_tokens.input_dim, dtype=input_ids.dtype),
-                    message=(
-                        "input_ids must be smaller than the embedding layer's input dimension (got"
-                        f" {tf.math.reduce_max(input_ids)} >= {self.embed_tokens.input_dim})"
-                    ),
-                )
+                check_embeddings_within_bounds(input_ids, self.embed_tokens.input_dim)
                 inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale
 
         hidden_states = inputs_embeds

src/transformers/models/blenderbot_small/modeling_tf_blenderbot_small.py

@@ -37,7 +37,7 @@
     keras_serializable,
     unpack_inputs,
 )
-from ...tf_utils import shape_list, stable_softmax
+from ...tf_utils import check_embeddings_within_bounds, shape_list, stable_softmax
 from ...utils import (
     ContextManagers,
     add_code_sample_docstrings,
@@ -752,16 +752,7 @@ def call(
             if hasattr(self.embed_tokens, "load_weight_prefix"):
                 context.append(tf.name_scope(self.embed_tokens.load_weight_prefix + "/"))
             with ContextManagers(context):
-                # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-                # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-                tf.debugging.assert_less(
-                    input_ids,
-                    tf.cast(self.embed_tokens.input_dim, dtype=input_ids.dtype),
-                    message=(
-                        "input_ids must be smaller than the embedding layer's input dimension (got"
-                        f" {tf.math.reduce_max(input_ids)} >= {self.embed_tokens.input_dim})"
-                    ),
-                )
+                check_embeddings_within_bounds(input_ids, self.embed_tokens.input_dim)
                 inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale
 
         embed_pos = self.embed_positions(input_shape)
@@ -961,16 +952,7 @@ def call(
             if hasattr(self.embed_tokens, "load_weight_prefix"):
                 context.append(tf.name_scope(self.embed_tokens.load_weight_prefix + "/"))
             with ContextManagers(context):
-                # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-                # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-                tf.debugging.assert_less(
-                    input_ids,
-                    tf.cast(self.embed_tokens.input_dim, dtype=input_ids.dtype),
-                    message=(
-                        "input_ids must be smaller than the embedding layer's input dimension (got"
-                        f" {tf.math.reduce_max(input_ids)} >= {self.embed_tokens.input_dim})"
-                    ),
-                )
+                check_embeddings_within_bounds(input_ids, self.embed_tokens.input_dim)
                 inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale
 
         # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]

src/transformers/models/blip/modeling_tf_blip.py

@@ -29,7 +29,7 @@
     shape_list,
     unpack_inputs,
 )
-from ...tf_utils import stable_softmax
+from ...tf_utils import check_embeddings_within_bounds, stable_softmax
 from ...utils import (
     ModelOutput,
     add_start_docstrings,
@@ -316,16 +316,7 @@ def call(
             raise ValueError("You have to specify either input_ids or inputs_embeds")
 
         if inputs_embeds is None:
-            # Note: tf.gather, on which the embedding layer is based, won't check positive out of bound
-            # indices on GPU, returning zeros instead. This is a dangerous silent behavior.
-            tf.debugging.assert_less(
-                input_ids,
-                tf.cast(self.config.vocab_size, dtype=input_ids.dtype),
-                message=(
-                    "input_ids must be smaller than the embedding layer's input dimension (got"
-                    f" {tf.math.reduce_max(input_ids)} >= {self.config.vocab_size})"
-                ),
-            )
+            check_embeddings_within_bounds(input_ids, self.config.vocab_size)
             inputs_embeds = tf.gather(params=self.weight, indices=input_ids)
 
         input_shape = shape_list(inputs_embeds)[:-1]