Align fork with HPU upstream code (#465)

vllm-project#6143 got merged, but it's
based on an older revision of HPU components. This PR aligns the two.

docs/source/getting_started/gaudi-installation.rst

@@ -1,5 +1,5 @@
-vLLM with Intel® Gaudi® AI Accelerators
-=========================================
+Installation with Intel® Gaudi® AI Accelerators
+===============================================
 
 This README provides instructions on running vLLM with Intel Gaudi devices.
 
@@ -22,22 +22,22 @@ Requirements
 
 
 Quick start using Dockerfile
-============================
+----------------------------
 .. code:: console
 
    $ docker build -f Dockerfile.hpu -t vllm-hpu-env  .
    $ docker run -it --runtime=habana -e HABANA_VISIBLE_DEVICES=all -e OMPI_MCA_btl_vader_single_copy_mechanism=none --cap-add=sys_nice --net=host --rm vllm-hpu-env
 
 
 .. tip::
-   If you're observing the following error: ``docker: Error response from daemon: Unknown runtime specified habana.``, please refer to "Install Using Containers" section of `Intel Gaudi Software Stack and Driver Installation <https://docs.habana.ai/en/v1.18.0/Installation_Guide/Bare_Metal_Fresh_OS.html`__. Make sure you have ``habana-container-runtime`` package installed and that ```habana`` container runtime is registered.
+   If you're observing the following error: ``docker: Error response from daemon: Unknown runtime specified habana.``, please refer to "Install Using Containers" section of `Intel Gaudi Software Stack and Driver Installation <https://docs.habana.ai/en/v1.18.0/Installation_Guide/Bare_Metal_Fresh_OS.html>`__. Make sure you have ``habana-container-runtime`` package installed and that ``habana`` container runtime is registered.
 
 
 Build from source
-=================
+-----------------
 
 Environment verification
-------------------------
+~~~~~~~~~~~~~~~~~~~~~~~~
 
 To verify that the Intel Gaudi software was correctly installed, run:
 
@@ -49,11 +49,11 @@ To verify that the Intel Gaudi software was correctly installed, run:
    $ pip list | grep neural # verify that neural_compressor is installed
 
 Refer to `Intel Gaudi Software Stack
-Verification <https://docs.habana.ai/en/latest/Installation_Guide/Platform_Upgrade_and_Unboxing.html#system-verifications-and-final-tests>`__
+Verification <https://docs.habana.ai/en/latest/Installation_Guide/SW_Verification.html#platform-upgrade>`__
 for more details.
 
 Run Docker Image
-----------------
+~~~~~~~~~~~~~~~~
 
 It is highly recommended to use the latest Docker image from Intel Gaudi
 vault. Refer to the `Intel Gaudi
@@ -68,7 +68,16 @@ Use the following commands to run a Docker image:
    $ docker run -it --runtime=habana -e HABANA_VISIBLE_DEVICES=all -e OMPI_MCA_btl_vader_single_copy_mechanism=none --cap-add=sys_nice --net=host --ipc=host vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
 
 Build and Install vLLM
----------------------------
+~~~~~~~~~~~~~~~~~~~~~~
+
+To build and install vLLM from source, run:
+
+.. code:: console
+
+   $ git clone https://github.com/vllm-project/vllm.git
+   $ cd vllm
+   $ python setup.py develop
+
 
 Currently, the latest features and performance optimizations are developed in Gaudi's `vLLM-fork <https://github.com/HabanaAI/vllm-fork>`__ and we periodically upstream them to vLLM main repo. To install latest `HabanaAI/vLLM-fork <https://github.com/HabanaAI/vllm-fork>`__, run the following:
 
@@ -77,16 +86,16 @@ Currently, the latest features and performance optimizations are developed in Ga
    $ git clone https://github.com/HabanaAI/vllm-fork.git
    $ cd vllm-fork
    $ git checkout habana_main
-   $ pip install -e .
+   $ python setup.py develop
 
 
 Supported Features
 ==================
 
 -  `Offline batched
-   inference <https://github.com/HabanaAI/vllm-fork/blob/habana_main/docs/source/getting_started/quickstart.rst#offline-batched-inference>`__
+   inference <https://docs.vllm.ai/en/latest/getting_started/quickstart.html#offline-batched-inference>`__
 -  Online inference via `OpenAI-Compatible
-   Server <https://github.com/HabanaAI/vllm-fork/blob/habana_main/docs/source/getting_started/quickstart.rst#openai-compatible-server>`__
+   Server <https://docs.vllm.ai/en/latest/getting_started/quickstart.html#openai-compatible-server>`__
 -  HPU autodetection - no need to manually select device within vLLM
 -  Paged KV cache with algorithms enabled for Intel Gaudi accelerators
 -  Custom Intel Gaudi implementations of Paged Attention, KV cache ops,

docs/source/index.rst

@@ -43,8 +43,7 @@ vLLM is flexible and easy to use with:
 * Tensor parallelism and pipeline parallelism support for distributed inference
 * Streaming outputs
 * OpenAI-compatible API server
-* Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs and GPUs, PowerPC CPUs, TPU, and AWS Trainium and Inferentia Accelerators.
-* (Experimental) Support for Intel® Gaudi® 2 accelerators
+* Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs, Gaudi® accelerators and GPUs, PowerPC CPUs, TPU, and AWS Trainium and Inferentia Accelerators.
 * Prefix caching support
 * Multi-lora support
 

requirements-hpu.txt

@@ -2,7 +2,7 @@
 -r requirements-common.txt
 
 # Dependencies for HPU code
-ray == 2.32.0
+ray
 triton
 pandas
 tabulate

setup.py

@@ -418,7 +418,7 @@ def get_vllm_version() -> str:
         neuron_version = str(get_neuronxcc_version())
         if neuron_version != MAIN_CUDA_VERSION:
             neuron_version_str = neuron_version.replace(".", "")[:3]
-            version += f"+neuron{neuron_version_str}"
+            version += f"{sep}neuron{neuron_version_str}"
     elif _is_hpu():
         # Get the Intel Gaudi Software Suite version
         gaudi_sw_version = str(get_gaudi_sw_version())

vllm/model_executor/layers/logits_processor.py

@@ -111,8 +111,9 @@ def _prune_hidden_states(
     hidden_states: torch.Tensor,
     sampling_metadata: SamplingMetadata,
 ) -> torch.Tensor:
-    # NOTE(kzawora): This is needed for Gaudi - in some scenarios (warmup,
-    # profile_run) we might not have selected_token_indices, so we skip pruning.
+    # NOTE(kzawora): The if guard is needed for Gaudi - in some scenarios
+    # (warmup, profile_run) we might not have selected_token_indices,
+    # so we skip pruning.
     if sampling_metadata.selected_token_indices is not None:
         return hidden_states.index_select(
             0, sampling_metadata.selected_token_indices)