Fixing exl2 (by disabling cuda graphs)

Dockerfile

@@ -250,6 +250,7 @@ RUN cd server && \
     pip install nvidia-nccl-cu12==2.22.3
 
 ENV LD_PRELOAD=/opt/conda/lib/python3.10/site-packages/nvidia/nccl/lib/libnccl.so.2
+ENV EXLLAMA_NO_FLASH_ATTN=1
 
 # Deps before the binaries
 # The binaries change on every build given we burn the SHA into them

integration-tests/models/test_flash_llama_exl2.py

@@ -6,6 +6,9 @@ def flash_llama_exl2_handle(launcher):
     with launcher(
         "turboderp/Llama-3-8B-Instruct-exl2",
         revision="2.5bpw",
+        # TODO
+        # Exl2 is currently broken with cuda graphs.
+        cuda_graphs=[0],
         # Set max input length to avoid OOM due to extremely large
         # scratch buffer.
         max_input_length=1024,

server/poetry.lock

@@ -1070,6 +1070,30 @@ win32-setctime = {version = ">=1.0.0", markers = "sys_platform == \"win32\""}
 [package.extras]
 dev = ["Sphinx (>=4.1.1)", "black (>=19.10b0)", "colorama (>=0.3.4)", "docutils (==0.16)", "flake8 (>=3.7.7)", "isort (>=5.1.1)", "pytest (>=4.6.2)", "pytest-cov (>=2.7.1)", "sphinx-autobuild (>=0.7.1)", "sphinx-rtd-theme (>=0.4.3)", "tox (>=3.9.0)"]
 
+[[package]]
+name = "markdown-it-py"
+version = "3.0.0"
+description = "Python port of markdown-it. Markdown parsing, done right!"
+optional = false
+python-versions = ">=3.8"
+files = [
+    {file = "markdown-it-py-3.0.0.tar.gz", hash = "sha256:e3f60a94fa066dc52ec76661e37c851cb232d92f9886b15cb560aaada2df8feb"},
+    {file = "markdown_it_py-3.0.0-py3-none-any.whl", hash = "sha256:355216845c60bd96232cd8d8c40e8f9765cc86f46880e43a8fd22dc1a1a8cab1"},
+]
+
+[package.dependencies]
+mdurl = ">=0.1,<1.0"
+
+[package.extras]
+benchmarking = ["psutil", "pytest", "pytest-benchmark"]
+code-style = ["pre-commit (>=3.0,<4.0)"]
+compare = ["commonmark (>=0.9,<1.0)", "markdown (>=3.4,<4.0)", "mistletoe (>=1.0,<2.0)", "mistune (>=2.0,<3.0)", "panflute (>=2.3,<3.0)"]
+linkify = ["linkify-it-py (>=1,<3)"]
+plugins = ["mdit-py-plugins"]
+profiling = ["gprof2dot"]
+rtd = ["jupyter_sphinx", "mdit-py-plugins", "myst-parser", "pyyaml", "sphinx", "sphinx-copybutton", "sphinx-design", "sphinx_book_theme"]
+testing = ["coverage", "pytest", "pytest-cov", "pytest-regressions"]
+
 [[package]]
 name = "markupsafe"
 version = "2.1.5"
@@ -1207,6 +1231,17 @@ torch = "*"
 type = "url"
 url = "https://github.com/danieldk/marlin-kernels/releases/download/v0.2.0/marlin_kernels-0.2.0+cu123torch2.4-cp39-cp39-linux_x86_64.whl"
 
+[[package]]
+name = "mdurl"
+version = "0.1.2"
+description = "Markdown URL utilities"
+optional = false
+python-versions = ">=3.7"
+files = [
+    {file = "mdurl-0.1.2-py3-none-any.whl", hash = "sha256:84008a41e51615a49fc9966191ff91509e3c40b939176e643fd50a5c2196b8f8"},
+    {file = "mdurl-0.1.2.tar.gz", hash = "sha256:bb413d29f5eea38f31dd4754dd7377d4465116fb207585f97bf925588687c1ba"},
+]
+
 [[package]]
 name = "mpmath"
 version = "1.3.0"
@@ -2277,6 +2312,20 @@ files = [
 [package.dependencies]
 typing-extensions = ">=4.6.0,<4.7.0 || >4.7.0"
 
+[[package]]
+name = "pygments"
+version = "2.18.0"
+description = "Pygments is a syntax highlighting package written in Python."
+optional = false
+python-versions = ">=3.8"
+files = [
+    {file = "pygments-2.18.0-py3-none-any.whl", hash = "sha256:b8e6aca0523f3ab76fee51799c488e38782ac06eafcf95e7ba832985c8e7b13a"},
+    {file = "pygments-2.18.0.tar.gz", hash = "sha256:786ff802f32e91311bff3889f6e9a86e81505fe99f2735bb6d60ae0c5004f199"},
+]
+
+[package.extras]
+windows-terminal = ["colorama (>=0.4.6)"]
+
 [[package]]
 name = "pytest"
 version = "7.4.4"
@@ -2508,6 +2557,24 @@ urllib3 = ">=1.21.1,<3"
 socks = ["PySocks (>=1.5.6,!=1.5.7)"]
 use-chardet-on-py3 = ["chardet (>=3.0.2,<6)"]
 
+[[package]]
+name = "rich"
+version = "13.7.1"
+description = "Render rich text, tables, progress bars, syntax highlighting, markdown and more to the terminal"
+optional = false
+python-versions = ">=3.7.0"
+files = [
+    {file = "rich-13.7.1-py3-none-any.whl", hash = "sha256:4edbae314f59eb482f54e9e30bf00d33350aaa94f4bfcd4e9e3110e64d0d7222"},
+    {file = "rich-13.7.1.tar.gz", hash = "sha256:9be308cb1fe2f1f57d67ce99e95af38a1e2bc71ad9813b0e247cf7ffbcc3a432"},
+]
+
+[package.dependencies]
+markdown-it-py = ">=2.2.0"
+pygments = ">=2.13.0,<3.0.0"
+
+[package.extras]
+jupyter = ["ipywidgets (>=7.5.1,<9)"]
+
 [[package]]
 name = "rpds-py"
 version = "0.19.0"
@@ -3584,4 +3651,4 @@ torch = ["torch"]
 [metadata]
 lock-version = "2.0"
 python-versions = ">=3.9,<3.13"
-content-hash = "a89867b23017d2efa8a7aa14d4764bcbd3b4dea9bfbf06a7a68464cb184ac6a1"
+content-hash = "0ff7a244a409b616490cb238995bbe28dedf67ccb8855edafa2b71ee2e777dbd"

server/pyproject.toml

@@ -46,6 +46,7 @@ marlin-kernels = [
   { url = "https://github.com/danieldk/marlin-kernels/releases/download/v0.2.0/marlin_kernels-0.2.0+cu123torch2.4-cp311-cp311-linux_x86_64.whl", python = "~3.11", optional = true },
   { url = "https://github.com/danieldk/marlin-kernels/releases/download/v0.2.0/marlin_kernels-0.2.0+cu123torch2.4-cp312-cp312-linux_x86_64.whl", python = "~3.12", optional = true },
 ]
+rich = "^13.7.1"
 
 [tool.poetry.extras]
 torch = ["torch"]

server/requirements_cuda.txt

@@ -16,6 +16,8 @@ huggingface-hub==0.23.5 ; python_version >= "3.9" and python_version < "3.13"
 idna==3.7 ; python_version >= "3.9" and python_version < "3.13"
 importlib-metadata==7.1.0 ; python_version >= "3.9" and python_version < "3.13"
 loguru==0.6.0 ; python_version >= "3.9" and python_version < "3.13"
+markdown-it-py==3.0.0 ; python_version >= "3.9" and python_version < "3.13"
+mdurl==0.1.2 ; python_version >= "3.9" and python_version < "3.13"
 numpy==1.26.4 ; python_version >= "3.9" and python_version < "3.13"
 opentelemetry-api==1.25.0 ; python_version >= "3.9" and python_version < "3.13"
 opentelemetry-exporter-otlp-proto-common==1.25.0 ; python_version >= "3.9" and python_version < "3.13"
@@ -32,9 +34,11 @@ pillow==10.4.0 ; python_version >= "3.9" and python_version < "3.13"
 prometheus-client==0.20.0 ; python_version >= "3.9" and python_version < "3.13"
 protobuf==4.25.3 ; python_version >= "3.9" and python_version < "3.13"
 py-cpuinfo==9.0.0 ; python_version >= "3.9" and python_version < "3.13"
+pygments==2.18.0 ; python_version >= "3.9" and python_version < "3.13"
 pyyaml==6.0.1 ; python_version >= "3.9" and python_version < "3.13"
 regex==2024.5.15 ; python_version >= "3.9" and python_version < "3.13"
 requests==2.32.3 ; python_version >= "3.9" and python_version < "3.13"
+rich==13.7.1 ; python_version >= "3.9" and python_version < "3.13"
 safetensors==0.4.3 ; python_version >= "3.9" and python_version < "3.13"
 scipy==1.13.1 ; python_version >= "3.9" and python_version < "3.13"
 sentencepiece==0.1.99 ; python_version >= "3.9" and python_version < "3.13"

server/requirements_intel.txt

@@ -16,6 +16,8 @@ huggingface-hub==0.23.5 ; python_version >= "3.9" and python_version < "3.13"
 idna==3.7 ; python_version >= "3.9" and python_version < "3.13"
 importlib-metadata==7.1.0 ; python_version >= "3.9" and python_version < "3.13"
 loguru==0.6.0 ; python_version >= "3.9" and python_version < "3.13"
+markdown-it-py==3.0.0 ; python_version >= "3.9" and python_version < "3.13"
+mdurl==0.1.2 ; python_version >= "3.9" and python_version < "3.13"
 numpy==1.26.4 ; python_version >= "3.9" and python_version < "3.13"
 opentelemetry-api==1.25.0 ; python_version >= "3.9" and python_version < "3.13"
 opentelemetry-exporter-otlp-proto-common==1.25.0 ; python_version >= "3.9" and python_version < "3.13"
@@ -32,9 +34,11 @@ pillow==10.4.0 ; python_version >= "3.9" and python_version < "3.13"
 prometheus-client==0.20.0 ; python_version >= "3.9" and python_version < "3.13"
 protobuf==4.25.3 ; python_version >= "3.9" and python_version < "3.13"
 py-cpuinfo==9.0.0 ; python_version >= "3.9" and python_version < "3.13"
+pygments==2.18.0 ; python_version >= "3.9" and python_version < "3.13"
 pyyaml==6.0.1 ; python_version >= "3.9" and python_version < "3.13"
 regex==2024.5.15 ; python_version >= "3.9" and python_version < "3.13"
 requests==2.32.3 ; python_version >= "3.9" and python_version < "3.13"
+rich==13.7.1 ; python_version >= "3.9" and python_version < "3.13"
 safetensors==0.4.3 ; python_version >= "3.9" and python_version < "3.13"
 scipy==1.13.1 ; python_version >= "3.9" and python_version < "3.13"
 sentencepiece==0.1.99 ; python_version >= "3.9" and python_version < "3.13"

server/requirements_rocm.txt

@@ -16,6 +16,8 @@ huggingface-hub==0.23.5 ; python_version >= "3.9" and python_version < "3.13"
 idna==3.7 ; python_version >= "3.9" and python_version < "3.13"
 importlib-metadata==7.1.0 ; python_version >= "3.9" and python_version < "3.13"
 loguru==0.6.0 ; python_version >= "3.9" and python_version < "3.13"
+markdown-it-py==3.0.0 ; python_version >= "3.9" and python_version < "3.13"
+mdurl==0.1.2 ; python_version >= "3.9" and python_version < "3.13"
 numpy==1.26.4 ; python_version >= "3.9" and python_version < "3.13"
 opentelemetry-api==1.25.0 ; python_version >= "3.9" and python_version < "3.13"
 opentelemetry-exporter-otlp-proto-common==1.25.0 ; python_version >= "3.9" and python_version < "3.13"
@@ -32,9 +34,11 @@ pillow==10.4.0 ; python_version >= "3.9" and python_version < "3.13"
 prometheus-client==0.20.0 ; python_version >= "3.9" and python_version < "3.13"
 protobuf==4.25.3 ; python_version >= "3.9" and python_version < "3.13"
 py-cpuinfo==9.0.0 ; python_version >= "3.9" and python_version < "3.13"
+pygments==2.18.0 ; python_version >= "3.9" and python_version < "3.13"
 pyyaml==6.0.1 ; python_version >= "3.9" and python_version < "3.13"
 regex==2024.5.15 ; python_version >= "3.9" and python_version < "3.13"
 requests==2.32.3 ; python_version >= "3.9" and python_version < "3.13"
+rich==13.7.1 ; python_version >= "3.9" and python_version < "3.13"
 safetensors==0.4.3 ; python_version >= "3.9" and python_version < "3.13"
 scipy==1.13.1 ; python_version >= "3.9" and python_version < "3.13"
 sentencepiece==0.1.99 ; python_version >= "3.9" and python_version < "3.13"