Added support of HF and GenAI models into CLI (openvinotoolkit#887)

Author: AlexKoff88

.github/workflows/llm_bench-python.yml

@@ -40,7 +40,8 @@ jobs:
         python -m pip install --upgrade pip
         python -m pip install flake8 pytest black
         GIT_CLONE_PROTECTION_ACTIVE=false pip install -r ${{ env.LLM_BENCH_PYPATH }}/requirements.txt
-        pip install openvino-nightly
+        python -m pip install -U --pre openvino openvino-tokenizers openvino-genai --extra-index-url
+https://storage.openvinotoolkit.org/simple/wheels/nightly
         GIT_CLONE_PROTECTION_ACTIVE=false pip install -r ${{ env.WWB_PATH }}/requirements.txt
         GIT_CLONE_PROTECTION_ACTIVE=false pip install ${{ env.WWB_PATH }}
 
@@ -73,7 +74,7 @@ jobs:
         python ./llm_bench/python/benchmark.py -m ./ov_models/tiny-sd/pytorch/dldt/FP16/ -pf ./llm_bench/python/prompts/stable-diffusion.jsonl -d cpu -n 1
     - name: WWB Tests
       run: |
-        python -m pytest  ./llm_bench/python/who_what_benchmark/tests
+        python -m pytest llm_bench/python/who_what_benchmark/tests
   stateful:
     runs-on: ubuntu-20.04
     steps:
@@ -85,12 +86,13 @@ jobs:
         run: |
           GIT_CLONE_PROTECTION_ACTIVE=false python -m pip install -r llm_bench/python/requirements.txt
           python -m pip uninstall --yes openvino
-          python -m pip install openvino-nightly
+          python -m pip install -U --pre openvino openvino-tokenizers openvino-genai --extra-index-url
+https://storage.openvinotoolkit.org/simple/wheels/nightly
           python llm_bench/python/convert.py --model_id TinyLlama/TinyLlama-1.1B-Chat-v1.0 --output_dir . --stateful
           grep beam_idx pytorch/dldt/FP32/openvino_model.xml
       - name: WWB Tests
         run: |
           GIT_CLONE_PROTECTION_ACTIVE=false pip install -r llm_bench/python/who_what_benchmark/requirements.txt
           GIT_CLONE_PROTECTION_ACTIVE=false pip install llm_bench/python/who_what_benchmark/
           pip install pytest
-          python -m pytest  llm_bench/python/who_what_benchmark/tests
+          python -m pytest llm_bench/python/who_what_benchmark/tests

llm_bench/python/who_what_benchmark/README.md

@@ -55,27 +55,36 @@ metrics_per_prompt, metrics = evaluator.score(optimized_model, test_data=prompts
 ```sh
 wwb --help
 
-# run ground truth generation for uncompressed model on the first 32 samples from squad dataset
-# ground truth will be saved in llama_2_7b_squad_gt.csv file
+# Run ground truth generation for uncompressed model on the first 32 samples from squad dataset
+# Ground truth will be saved in llama_2_7b_squad_gt.csv file
 wwb --base-model meta-llama/Llama-2-7b-chat-hf --gt-data llama_2_7b_squad_gt.csv --dataset squad --split validation[:32] --dataset-field question
 
-# run comparison with compressed model on the first 32 samples from squad dataset
+# Run comparison with compressed model on the first 32 samples from squad dataset
 wwb --target-model /home/user/models/Llama_2_7b_chat_hf_int8 --gt-data llama_2_7b_squad_gt.csv --dataset squad --split validation[:32] --dataset-field question
 
-# output will be like this
+# Output will be like this
 #   similarity        FDT        SDT  FDT norm  SDT norm
 # 0    0.972823  67.296296  20.592593  0.735127  0.151505
 
-# run ground truth generation for uncompressed model on internal set of questions
-# ground truth will be saved in llama_2_7b_squad_gt.csv file
+# Run ground truth generation for uncompressed model on internal set of questions
+# Ground truth will be saved in llama_2_7b_squad_gt.csv file
 wwb --base-model meta-llama/Llama-2-7b-chat-hf --gt-data llama_2_7b_wwb_gt.csv
 
-# run comparison with compressed model on internal set of questions
+# Run comparison with compressed model on internal set of questions
 wwb --target-model /home/user/models/Llama_2_7b_chat_hf_int8 --gt-data llama_2_7b_wwb_gt.csv
 
-## Control the number of samples and use verbose mode to see the difference in the results
+# Use --num-samples to control the number of samples
 wwb --base-model meta-llama/Llama-2-7b-chat-hf --gt-data llama_2_7b_wwb_gt.csv --num-samples 10
+
+# Use -v for verbose mode to see the difference in the results
 wwb --target-model /home/user/models/Llama_2_7b_chat_hf_int8 --gt-data llama_2_7b_wwb_gt.csv  --num-samples 10 -v
+
+# Use --hf AutoModelForCausalLM to instantiate the model from model_id/folder
+wwb --base-model meta-llama/Llama-2-7b-chat-hf --gt-data llama_2_7b_wwb_gt.csv --hf
+
+# Use --language parameter to control the language of promts
+# Autodetection works for basic Chinese models 
+wwb --base-model meta-llama/Llama-2-7b-chat-hf --gt-data llama_2_7b_wwb_gt.csv --hf
 ```
 
 ### Supported metrics

llm_bench/python/who_what_benchmark/requirements.txt

@@ -1,8 +1,10 @@
 transformers>=4.35.2
 sentence-transformers>=2.2.2
-openvino>=2023.3.0
-openvino-telemetry>=2023.2.1
+openvino>=2024.3.0
+openvino-telemetry>=2024.3.0
 optimum-intel>=1.14
+openvino-tokenizers>=2024.3.0
+openvino-genai>=2024.3.0
 pandas>=2.0.3
 numpy>=1.23.5
 tqdm>=4.66.1

llm_bench/python/who_what_benchmark/tests/test_cli.py

@@ -32,17 +32,21 @@ def run_wwb(args):
 
 
 def setup_module():
+    from optimum.exporters.openvino.convert import export_tokenizer
+
     logger.info("Create models")
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     base_model = OVModelForCausalLM.from_pretrained(model_id)
     base_model.save_pretrained(base_model_path)
     tokenizer.save_pretrained(base_model_path)
+    export_tokenizer(tokenizer, base_model_path)
 
     target_model = OVModelForCausalLM.from_pretrained(
         model_id, quantization_config=OVWeightQuantizationConfig(bits=8)
     )
     target_model.save_pretrained(target_model_path)
     tokenizer.save_pretrained(target_model_path)
+    export_tokenizer(tokenizer, target_model_path)
 
 
 def teardown_module():
@@ -57,9 +61,10 @@ def test_target_model():
         "--num-samples", "2",
         "--device", "CPU"
     ])
+
     assert result.returncode == 0
-    assert "Metrics for model" in result.stdout
-    assert "## Reference text" not in result.stdout
+    assert "Metrics for model" in result.stderr
+    assert "## Reference text" not in result.stderr
 
 
 @pytest.fixture
@@ -76,8 +81,6 @@ def test_gt_data():
         "--num-samples", "2",
         "--device", "CPU"
     ])
-    import time
-    time.sleep(1)
     data = pd.read_csv(temp_file_name)
     os.remove(temp_file_name)
 
@@ -95,7 +98,7 @@ def test_output_directory():
             "--output", temp_dir
         ])
         assert result.returncode == 0
-        assert "Metrics for model" in result.stdout
+        assert "Metrics for model" in result.stderr
         assert os.path.exists(os.path.join(temp_dir, "metrics_per_qustion.csv"))
         assert os.path.exists(os.path.join(temp_dir, "metrics.csv"))
 
@@ -109,7 +112,7 @@ def test_verbose():
         "--verbose"
     ])
     assert result.returncode == 0
-    assert "## Reference text" in result.stdout
+    assert "## Diff " in result.stderr
 
 
 def test_language_autodetect():
@@ -127,3 +130,34 @@ def test_language_autodetect():
 
     assert result.returncode == 0
     assert "马克" in data["questions"].values[0]
+
+
+def test_hf_model():
+    with tempfile.NamedTemporaryFile(suffix=".csv") as tmpfile:
+        temp_file_name = tmpfile.name
+
+    result = run_wwb([
+        "--base-model", model_id,
+        "--gt-data", temp_file_name,
+        "--num-samples", "2",
+        "--device", "CPU",
+        "--hf"
+    ])
+    data = pd.read_csv(temp_file_name)
+    os.remove(temp_file_name)
+
+    assert result.returncode == 0
+    assert len(data["questions"].values) == 2
+
+
+def test_genai_model():
+    result = run_wwb([
+        "--base-model", base_model_path,
+        "--target-model", target_model_path,
+        "--num-samples", "2",
+        "--device", "CPU",
+        "--genai"
+    ])
+    assert result.returncode == 0
+    assert "Metrics for model" in result.stderr
+    assert "## Reference text" not in result.stderr

llm_bench/python/who_what_benchmark/whowhatbench/evaluator.py

@@ -96,7 +96,8 @@ def __init__(
         max_new_tokens=128,
         crop_question=True,
         num_samples=None,
-        language=None
+        language=None,
+        gen_answer_fn=None,
     ) -> None:
         assert (
             base_model is not None or gt_data is not None
@@ -116,7 +117,7 @@ def __init__(
                 self.language = autodetect_language(base_model)
 
         if base_model:
-            self.gt_data = self._generate_data(base_model)
+            self.gt_data = self._generate_data(base_model, gen_answer_fn)
         else:
             self.gt_data = pd.read_csv(gt_data, keep_default_na=False)
 

llm_bench/python/who_what_benchmark/whowhatbench/wwb.py

@@ -1,25 +1,61 @@
 import argparse
 import difflib
 import os
-
 import json
 import pandas as pd
+import logging
 from datasets import load_dataset
 from optimum.exporters import TasksManager
 from optimum.intel.openvino import OVModelForCausalLM
 from optimum.utils import NormalizedConfigManager, NormalizedTextConfig
-from transformers import AutoConfig, AutoTokenizer
+from transformers import AutoConfig, AutoTokenizer, AutoModelForCausalLM
 
 from . import Evaluator
 
+# Configure logging
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
 TasksManager._SUPPORTED_MODEL_TYPE["stablelm-epoch"] = TasksManager._SUPPORTED_MODEL_TYPE["llama"]
 NormalizedConfigManager._conf["stablelm-epoch"] = NormalizedTextConfig.with_args(
     num_layers="num_hidden_layers",
     num_attention_heads="num_attention_heads",
 )
 
 
-def load_model(model_id, device="CPU", ov_config=None):
+class GenAIModelWrapper():
+    """
+    A helper class to store additional attributes for GenAI models
+    """
+    def __init__(self, model, model_dir):
+        self.model = model
+        self.config = AutoConfig.from_pretrained(model_dir)
+
+    def __getattr__(self, attr):
+        if attr in self.__dict__:
+            return getattr(self, attr)
+        else:
+            return getattr(self.model, attr)
+
+
+def load_genai_pipeline(model_dir, device="CPU"):
+    try:
+        import openvino_genai
+    except ImportError:
+        logger.error("Failed to import openvino_genai package. Please install it.")
+        exit(-1)
+    logger.info("Using OpenVINO GenAI API")
+    return GenAIModelWrapper(openvino_genai.LLMPipeline(model_dir, device), model_dir)
+
+
+def load_model(model_id, device="CPU", ov_config=None, use_hf=False, use_genai=False):
+    if use_hf:
+        logger.info("Using HF Transformers API")
+        return AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True, device_map=device.lower())
+
+    if use_genai:
+        return load_genai_pipeline(model_id, device)
+
     if ov_config:
         with open(ov_config) as f:
             ov_options = json.load(f)
@@ -157,6 +193,16 @@ def parse_args():
         default=None,
         help="Used to select default prompts based on the primary model language, e.g. 'en', 'ch'.",
     )
+    parser.add_argument(
+        "--hf",
+        action="store_true",
+        help="Use AutoModelForCausalLM from transformers library to instantiate the model.",
+    )
+    parser.add_argument(
+        "--genai",
+        action="store_true",
+        help="Use LLMPipeline from transformers library to instantiate the model.",
+    )
 
     return parser.parse_args()
 
@@ -211,6 +257,11 @@ def diff_strings(a: str, b: str, *, use_loguru_colors: bool = False) -> str:
     return "".join(output)
 
 
+def genai_gen_answer(model, tokenizer, question, max_new_tokens, skip_question):
+    out = model.generate(question, max_new_tokens=max_new_tokens)
+    return out
+
+
 def main():
     args = parse_args()
     check_args(args)
@@ -228,24 +279,25 @@ def main():
             language=args.language,
         )
     else:
-        base_model = load_model(args.base_model, args.device, args.ov_config)
+        base_model = load_model(args.base_model, args.device, args.ov_config, args.hf, args.genai)
         evaluator = Evaluator(
             base_model=base_model,
             test_data=prompts,
             tokenizer=tokenizer,
             similarity_model_id=args.text_encoder,
             num_samples=args.num_samples,
             language=args.language,
+            gen_answer_fn=genai_gen_answer if args.genai else None
         )
         if args.gt_data:
             evaluator.dump_gt(args.gt_data)
         del base_model
 
     if args.target_model:
-        target_model = load_model(args.target_model, args.device, args.ov_config)
-        all_metrics_per_question, all_metrics = evaluator.score(target_model)
-        print("Metrics for model: ", args.target_model)
-        print(all_metrics)
+        target_model = load_model(args.target_model, args.device, args.ov_config, args.hf, args.genai)
+        all_metrics_per_question, all_metrics = evaluator.score(target_model, genai_gen_answer if args.genai else None)
+        logger.info("Metrics for model: %s", args.target_model)
+        logger.info(all_metrics)
 
         if args.output:
             if not os.path.exists(args.output):
@@ -269,11 +321,11 @@ def main():
                 actual_text += l2 + "\n"
                 diff += diff_strings(l1, l2) + "\n"
 
-            print("--------------------------------------------------------------------------------------")
-            print("## Reference text {}:\n".format(i + 1), ref_text)
-            print("## Actual text {}:\n".format(i + 1), actual_text)
-            print("## Diff {}: ".format(i + 1))
-            print(diff)
+            logger.info("--------------------------------------------------------------------------------------")
+            logger.info("## Reference text %d:\n%s", i + 1, ref_text)
+            logger.info("## Actual text %d:\n%s", i + 1, actual_text)
+            logger.info("## Diff %d: ", i + 1)
+            logger.info(diff)
 
 
 if __name__ == "__main__":