[LLM]Windows unittest (#8356)

* win-unittest

* update

* update

* try llama 7b

* delete llama

* update

* add red-3b

* only test red-3b

* revert

* add langchain

* add dependency

* delete langchain

.github/workflows/llm_unit_tests_windows.yml

@@ -60,3 +60,15 @@ jobs:
           bash python/llm/test/run-llm-install-tests.sh
         env:
           ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+
+      - name: Run LLM unittests
+        shell: bash
+        run: |
+          bash python/llm/test/run-llm-windows-tests.sh
+        env:
+          ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+
+      - name: Clean up
+        shell: bash
+        run: |
+          rm -rf models

python/llm/test/run-llm-windows-tests.sh

@@ -0,0 +1,46 @@
+#!/bin/bash
+set -e
+
+export HF_TOKEN=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
+export ANALYTICS_ZOO_ROOT=${ANALYTICS_ZOO_ROOT}
+export LLM_HOME=${ANALYTICS_ZOO_ROOT}/python/llm/src
+export BLOOM_ORIGIN_PATH=${ANALYTICS_ZOO_ROOT}/models/bloom-560m
+# export LLAMA_ORIGIN_PATH=${ANALYTICS_ZOO_ROOT}/models/llama-7b-hf
+export GPTNEOX_ORIGIN_PATH=${ANALYTICS_ZOO_ROOT}/models/redpajama-3b
+export INT4_CKPT_DIR=${ANALYTICS_ZOO_ROOT}/models/converted_models
+# export LLAMA_INT4_CKPT_PATH=${INT4_CKPT_DIR}/bigdl_llm_llama_7b_q4_0.bin
+export GPTNEOX_INT4_CKPT_PATH=${INT4_CKPT_DIR}/bigdl_llm_redpajama_q4_0.bin
+export BLOOM_INT4_CKPT_PATH=${INT4_CKPT_DIR}/bigdl_llm_bloom_q4_0.bin
+
+echo "# Download the models"
+start=$(date "+%s")
+
+echo ${ANALYTICS_ZOO_ROOT}
+python ${ANALYTICS_ZOO_ROOT}/python/llm/test/win/download_from_huggingface.py
+
+now=$(date "+%s")
+time=$((now-start))
+echo "Models downloaded in:$time seconds"
+
+echo "# Start testing convert model"
+start=$(date "+%s")
+
+python -m pytest -s ${ANALYTICS_ZOO_ROOT}/python/llm/test/convert/test_convert_model.py -k 'test_convert_bloom'
+
+now=$(date "+%s")
+time=$((now-start))
+
+echo "Bigdl-llm convert model test finished"
+echo "Time used:$time seconds"
+
+
+echo "# Start testing inference"
+start=$(date "+%s")
+
+python -m pytest -s ${ANALYTICS_ZOO_ROOT}/python/llm/test/inference/test_call_models.py -k 'test_bloom_completion_success or test_bloom_completion_with_stream_success'
+
+now=$(date "+%s")
+time=$((now-start))
+
+echo "Bigdl-llm inference test finished"
+echo "Time used:$time seconds"

python/llm/test/win/download_from_huggingface.py

@@ -0,0 +1,37 @@
+#
+# Copyright 2016 The BigDL Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+
+import os
+from huggingface_hub import login
+from huggingface_hub import snapshot_download
+
+if __name__ == '__main__':
+    access_token_read = os.environ.get('HF_TOKEN')
+    login(token = access_token_read)
+
+
+    # Download bigscience/bloom-560m
+    snapshot_download(repo_id="bigscience/bloom-560m", local_dir="models/bloom-560m",
+                      local_dir_use_symlinks=False, ignore_patterns="*.safetensors")
+    
+    # # Download decapoda-research/llama-7b-hf
+    # snapshot_download(repo_id="decapoda-research/llama-7b-hf", local_dir="models/llama-7b-hf",
+    #                   local_dir_use_symlinks=False, ignore_patterns="*.safetensors")
+
+    # Download togethercomputer/RedPajama-INCITE-Chat-3B-v1
+    # snapshot_download(repo_id="togethercomputer/RedPajama-INCITE-Chat-3B-v1", local_dir="models/redpajama-3b",
+    #                   local_dir_use_symlinks=False, ignore_patterns="*.safetensors")