Merge branch 'main' into wangchang/adapt_quant_lm_head

intel · Jul 25, 2024 · c06bd11 · c06bd11
2 parents 8694152 + 2b4b188
commit c06bd11
Show file tree

Hide file tree

Showing 13 changed files with 25 additions and 13 deletions.
diff --git a/.github/workflows/Scaner_BDBA.yaml b/.github/workflows/Scaner_BDBA.yaml
@@ -2,6 +2,8 @@ name: Scanner BDBA
 
 on:
   workflow_dispatch:
+permissions:
+  contents: read
 
 jobs:
   bdba_job:

diff --git a/.github/workflows/Scaner_Coverity.yaml b/.github/workflows/Scaner_Coverity.yaml
@@ -1,8 +1,9 @@
 name: Scanner Coverity PYTHON
 
 on:
-    workflow_dispatch:
-
+  workflow_dispatch:
+permissions:
+  contents: read
 jobs:
   coverity_job:
     name: Coverity
@@ -14,8 +15,8 @@ jobs:
       - name: Scan
         uses: intel-innersource/frameworks.devops.github.actions.coverity@main
         with:
-          server: 'https://coverityent.devtools.intel.com/prod4'
+          server: "https://coverityent.devtools.intel.com/prod4"
           username: ${{secrets.COVERITY_NAME }}
           password: ${{secrets.COVERITY }}
           project: IntelExtensionForTransformers
-          stream: 'IntelExtensionForTransformers12cpp'
+          stream: "IntelExtensionForTransformers12cpp"
diff --git a/.github/workflows/Scaner_Trivy.yaml b/.github/workflows/Scaner_Trivy.yaml
@@ -1,7 +1,9 @@
 name: Trivy Scan for Containers
 
 on:
-    workflow_dispatch:
+  workflow_dispatch:
+permissions:
+  contents: read
 jobs:
   trivy_container_job:
     uses: "intel-innersource/frameworks.ai.infrastructure.code-scan-tools/.github/workflows/Scanner_Trivy.yml@one-ci-cd"

diff --git a/.github/workflows/chatbot-test.yml b/.github/workflows/chatbot-test.yml
@@ -27,6 +27,8 @@ on:
       - '!intel_extension_for_transformers/neural_chat/README.md'
 
   workflow_dispatch:
+permissions:
+  contents: read
 # If there is a new commit, the previous jobs will be canceled
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}

diff --git a/.github/workflows/chatbot_finetuning.yml b/.github/workflows/chatbot_finetuning.yml
@@ -2,7 +2,8 @@ name: Chat Bot Finetuning Test
 
 on:
   workflow_dispatch:
-
+permissions:
+  contents: read
 # If there is a new commit, the previous jobs will be canceled
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}

diff --git a/.github/workflows/probot.yml b/.github/workflows/probot.yml
@@ -9,7 +9,9 @@ on:
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}-${{ github.head_ref }}
   cancel-in-progress: true
-permissions: write-all
+permissions:
+  contents: read
+  pull-requests: write
 jobs:
   required-jobs:
     runs-on: ubuntu-latest

diff --git a/examples/huggingface/pytorch/code-generation/quantization/requirements.txt b/examples/huggingface/pytorch/code-generation/quantization/requirements.txt
@@ -11,5 +11,5 @@ tiktoken #code_gen
 neural-compressor
 intel_extension_for_pytorch==2.3.0
 git+https://github.com/huggingface/optimum-intel.git@50d867c13b22c22eda451ddb67bddb8159670f85
-auto-round==0.2
+git+https://github.com/intel/auto-round.git@61cf9eef4a3ccb5a2d83a557deb709091a548581
 git+https://github.com/bigcode-project/bigcode-evaluation-harness@094c7cc197d13a53c19303865e2056f1c7488ac1
diff --git a/examples/huggingface/pytorch/text-generation/quantization/requirements_GPU.txt b/examples/huggingface/pytorch/text-generation/quantization/requirements_GPU.txt
@@ -12,6 +12,6 @@ bitsandbytes  #baichuan
 transformers_stream_generator
 tiktoken  #qwen
 einops  #qwen
-auto-round
+git+https://github.com/intel/auto-round.git@e24b9074af6cdb099e31c92eb81b7f5e9a4a244e
 git+https://github.com/intel/neural-compressor.git
 lm-eval==0.4.3
diff --git a/examples/huggingface/pytorch/text-generation/quantization/requirements_cpu_woq.txt b/examples/huggingface/pytorch/text-generation/quantization/requirements_cpu_woq.txt
@@ -11,7 +11,7 @@ transformers_stream_generator
 tiktoken  #qwen
 einops  #qwen
 git+https://github.com/intel/neural-speed.git
-auto-round==0.2
+git+https://github.com/intel/auto-round.git@e24b9074af6cdb099e31c92eb81b7f5e9a4a244e
 git+https://github.com/intel/neural-compressor.git
 lm-eval==0.4.3
 huggingface_hub
diff --git a/intel_extension_for_transformers/neural_chat/pipeline/plugins/retrieval/requirements.txt b/intel_extension_for_transformers/neural_chat/pipeline/plugins/retrieval/requirements.txt
@@ -9,7 +9,7 @@ langchain_core==0.1.35
 langid
 markdown
 openpyxl
-protobuf==3.20
+protobuf==3.20.2
 PyMuPDF
 python-docx
 qdrant-client==1.9.0

diff --git a/tests/CI/test_quantization.py b/tests/CI/test_quantization.py
@@ -432,7 +432,7 @@ def test_quantization_for_llm(self):
         woq_model.eval()
         output = woq_model(dummy_input)
         if CpuInfo().bf16:
-            self.assertTrue(isclose(float(output[0][0][0][0]), 0.1513671875, rel_tol=1e-04))
+            self.assertTrue(isclose(float(output[0][0][0][0]), 0.150390625, rel_tol=1e-04))
 
     def test_export(self):
         # test model with model_id

diff --git a/tests/CI/test_weight_only.py b/tests/CI/test_weight_only.py
@@ -208,6 +208,7 @@ def test_auto_model_saving_loading(self):
                 module_list.append(name)
         self.assertTrue(len(module_list) > 0)
 
+    @unittest.skip("need bug fix.")
     def test_nf4_training(self):
         quantization_config = RtnConfig(bits=4, weight_dtype="nf4", scale_dtype="fp32")
         model = AutoModelForCausalLM.from_pretrained(
@@ -251,6 +252,7 @@ def test_nf4_training(self):
                 module.unmerge()
         model.merge_and_unload()
 
+    @unittest.skip("need bug fix.")
     def test_int8_training(self):
         model = AutoModelForCausalLM.from_pretrained(
             llama_model_path, load_in_8bit=True, use_neural_speed=False)

diff --git a/tests/requirements.txt b/tests/requirements.txt
@@ -1,13 +1,13 @@
 --extra-index-url https://download.pytorch.org/whl/cpu
 accelerate
 auto-gptq
-auto-round==0.2
 bitsandbytes
 datasets==2.16.1
 einops
 evaluate
 gguf
 git+https://github.com/huggingface/optimum-intel.git@50d867c13b22c22eda451ddb67bddb8159670f85
+git+https://github.com/intel/auto-round.git@61cf9eef4a3ccb5a2d83a557deb709091a548581
 git+https://github.com/intel/neural-compressor.git
 git+https://github.com/intel/neural-speed.git
 intel-extension-for-pytorch==2.3.0