Merge branch 'main' into pinbump1111

Author: Jack-Khuu

.ci/scripts/run-docs

@@ -75,9 +75,6 @@ if [ "$1" == "advanced" ]; then
 fi
 
 if [ "$1" == "evaluation" ]; then
-
-    exit 0
-
         echo "::group::Create script to run evaluation"
         python3 torchchat/utils/scripts/updown.py --file torchchat/utils/docs/evaluation.md --replace 'llama3:stories15M,-l 3:-l 2' --suppress huggingface-cli,HF_TOKEN > ./run-evaluation.sh
         # for good measure, if something happened to updown processor,
@@ -95,7 +92,7 @@ fi
 if [ "$1" == "multimodal" ]; then
 
    # Expecting that this might fail this test as-is, because 
-   # it's the first on-pr test depending on githib secrets for access with HF token access
+   # it's the first on-pr test depending on github secrets for access with HF token access
 
         echo "::group::Create script to run multimodal"
         python3 torchchat/utils/scripts/updown.py --file docs/multimodal.md > ./run-multimodal.sh
@@ -111,3 +108,20 @@ if [ "$1" == "multimodal" ]; then
         bash -x ./run-multimodal.sh
         echo "::endgroup::"
 fi
+
+if [ "$1" == "native" ]; then
+
+        echo "::group::Create script to run native-execution"
+        python3 torchchat/utils/scripts/updown.py --file docs/native-execution.md > ./run-native.sh
+        # for good measure, if something happened to updown processor,
+        # and it did not error out, fail with an exit 1
+        echo "exit 1" >> ./run-native.sh
+        echo "::endgroup::"
+
+        echo "::group::Run native-execution"
+        echo "*******************************************"
+        cat ./run-native.sh
+        echo "*******************************************"
+        bash -x ./run-native.sh
+        echo "::endgroup::"
+fi

.github/workflows/run-readme-pr-mps.yml

@@ -10,6 +10,7 @@ jobs:
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     with:
       runner: macos-m1-14
+      timeout-minutes: 50
       script: |
           conda create -y -n test-readme-mps-macos python=3.10.11 llvm-openmp
           conda activate test-readme-mps-macos

.github/workflows/run-readme-pr.yml

@@ -227,3 +227,46 @@ jobs:
         echo "::endgroup::"
 
         TORCHCHAT_DEVICE=cpu .ci/scripts/run-docs multimodal
+
+  test-native-any:
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      runner: linux.g5.4xlarge.nvidia.gpu
+      gpu-arch-type: cuda
+      gpu-arch-version: "12.1"
+      timeout: 60
+      script: |
+        echo "::group::Print machine info"
+        uname -a
+        echo "::endgroup::"
+
+        echo "::group::Install newer objcopy that supports --set-section-alignment"
+        yum install -y  devtoolset-10-binutils
+        export PATH=/opt/rh/devtoolset-10/root/usr/bin/:$PATH
+        echo "::endgroup::"
+
+        .ci/scripts/run-docs native
+
+        echo "::group::Completion"
+        echo "tests complete"
+        echo "*******************************************"
+        echo "::endgroup::"
+
+  test-native-cpu:
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      runner: linux.g5.4xlarge.nvidia.gpu
+      gpu-arch-type: cuda
+      gpu-arch-version: "12.1"
+      timeout: 60
+      script: |
+        echo "::group::Print machine info"
+        uname -a
+        echo "::endgroup::"
+
+        echo "::group::Install newer objcopy that supports --set-section-alignment"
+        yum install -y  devtoolset-10-binutils
+        export PATH=/opt/rh/devtoolset-10/root/usr/bin/:$PATH
+        echo "::endgroup::"
+
+        TORCHCHAT_DEVICE=cpu .ci/scripts/run-docs native  

README.md

@@ -231,6 +231,8 @@ python3 torchchat.py server llama3.1
 ```
 [skip default]: end
 
+[shell default]: python3 torchchat.py server llama3.1 & server_pid=$!
+
 In another terminal, query the server using `curl`. Depending on the model configuration, this query might take a few minutes to respond.
 
 > [!NOTE]
@@ -244,8 +246,6 @@ Setting `stream` to "true" in the request emits a response in chunks. If `stream
 
 **Example Input + Output**
 
-[skip default]: begin
-
 ```
 curl http://127.0.0.1:5000/v1/chat/completions \
   -H "Content-Type: application/json" \
@@ -265,12 +265,14 @@ curl http://127.0.0.1:5000/v1/chat/completions \
     ]
   }'
 ```
+[skip default]: begin
 ```
 {"response":" I'm a software developer with a passion for building innovative and user-friendly applications. I have experience in developing web and mobile applications using various technologies such as Java, Python, and JavaScript. I'm always looking for new challenges and opportunities to learn and grow as a developer.\n\nIn my free time, I enjoy reading books on computer science and programming, as well as experimenting with new technologies and techniques. I'm also interested in machine learning and artificial intelligence, and I'm always looking for ways to apply these concepts to real-world problems.\n\nI'm excited to be a part of the developer community and to have the opportunity to share my knowledge and experience with others. I'm always happy to help with any questions or problems you may have, and I'm looking forward to learning from you as well.\n\nThank you for visiting my profile! I hope you find my information helpful and interesting. If you have any questions or would like to discuss any topics, please feel free to reach out to me. I"}
 ```
 
 [skip default]: end
 
+[shell default]: kill ${server_pid}
 
 </details>
 
@@ -664,6 +666,6 @@ awesome libraries and tools you've built around local LLM inference.
 
 torchchat is released under the [BSD 3 license](LICENSE). (Additional
 code in this distribution is covered by the MIT and Apache Open Source
-licenses.) However you may have other legal obligations that govern
+licenses.) However, you may have other legal obligations that govern
 your use of content, such as the terms of service for third-party
 models.

docs/ADVANCED-USERS.md

@@ -251,6 +251,8 @@ To improve performance, you can compile the model with `--compile`
 trading off the time to first token processed with time per token.  To
 improve performance further, you may also compile the prefill with
 `--compile-prefill`. This will increase further compilation times though. 
+For CPU, you can use `--max-autotune` to further improve the performance
+with `--compile` and `compile-prefill`. See [`max-autotune on CPU tutorial`](https://pytorch.org/tutorials/prototype/max_autotune_on_CPU_tutorial.html).
 
 Parallel prefill is not yet supported by exported models, and may be
 supported in a future release.

docs/model_customization.md

@@ -34,6 +34,9 @@ prefill with `--compile_prefill`.
 
 To learn more about compilation, check out: https://pytorch.org/get-started/pytorch-2.0/
 
+For CPU, you can use `--max-autotune` to further improve the performance with `--compile` and `compile-prefill`.
+
+See [`max-autotune on CPU tutorial`](https://pytorch.org/tutorials/prototype/max_autotune_on_CPU_tutorial.html).
 
 ## Model Precision
 

torchchat/utils/docs/evaluation.md

@@ -4,8 +4,13 @@
 
 # Evaluation Features
 
+<!--
+
 [shell default]: ./install/install_requirements.sh
 
+[shell default]: TORCHCHAT_ROOT=${PWD} ./torchchat/utils/scripts/install_et.sh
+
+-->
 
 Torchchat provides evaluation functionality for your language model on
 a variety of tasks using the
@@ -14,7 +19,7 @@ library.
 
 ## Usage
 
-The evaluation mode of `torchchat.py` script can be used to evaluate your language model on various tasks available in the `lm_eval` library such as "wikitext". You can specify the task(s) you want to evaluate using the `--tasks` option, and limit the evaluation using the `--limit` option. If no task is specified, it will default to evaluating on "wikitext".
+The evaluation mode of `torchchat.py` script can be used to evaluate your language model on various tasks available in the `lm_eval` library such as "wikitext". You can specify the task(s) you want to evaluate using the `--tasks` option, and limit the evaluation using the `--limit` option. If no task is specified, the task will default to evaluating on "wikitext".
 
 **Examples**