use autotokenizer

intel-analytics · Nov 10, 2023 · 7afb188 · 7afb188
1 parent efc651d
commit 7afb188
Show file tree

Hide file tree

Showing 2 changed files with 7 additions and 1 deletion.
diff --git a/.github/workflows/llm-harness-evaluation.yml b/.github/workflows/llm-harness-evaluation.yml
@@ -25,10 +25,16 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
+        # include:
+        #   python-version: ["3.9"]
+        #   model_name: "stablelm-3b-4e1t"
+        #   task: "arc"
+        #   precision: ["sym_int4"] #options: sym_int4, fp4, nf4, mixed_4bit, fp8
         python-version: ["3.9"]
         model_name: ["stablelm-3b-4e1t"]
         task: ["truthfulqa"]
         precision: ["mixed_4bit"] #options: sym_int4, fp4, nf4, mixed_4bit, fp8
+
     runs-on: [self-hosted, llm, accuracy, temp-arc01]
     env:
       ANALYTICS_ZOO_ROOT: ${{ github.workspace }}

diff --git a/python/llm/dev/benchmark/harness/bigdl_llm.py b/python/llm/dev/benchmark/harness/bigdl_llm.py
@@ -66,7 +66,7 @@ def __init__(
         self._device = device
         self.model = model.to(device)
 
-        self.tokenizer = LlamaTokenizer.from_pretrained(pretrained, trust_remote_code=True)
+        self.tokenizer = AutoTokenizer.from_pretrained(pretrained, trust_remote_code=True)
 
         # setup for automatic batch size detection
         if batch_size == 'auto':