MindsDB 25.7.4.0__py3-none-any.whl → 25.8.3.0__py3-none-any.whl

mindsdb/integrations/handlers/huggingface_handler/huggingface_handler.py

@@ -1,383 +1,360 @@
-# from typing import Dict, Optional
-
-# import pandas as pd
-# import transformers
-# from huggingface_hub import HfApi
-
-# from mindsdb.integrations.handlers.huggingface_handler.settings import FINETUNE_MAP
-# from mindsdb.integrations.libs.base import BaseMLEngine
-# from mindsdb.utilities import log
-
-# logger = log.getLogger(__name__)
-
-
-# class HuggingFaceHandler(BaseMLEngine):
-#     name = "huggingface"
-
-#     @staticmethod
-#     def create_validation(target, args=None, **kwargs):
-
-#         if "using" in args:
-#             args = args["using"]
-
-#         hf_api = HfApi()
-
-#         # check model is pytorch based
-#         metadata = hf_api.model_info(args["model_name"])
-#         if "pytorch" not in metadata.tags:
-#             raise Exception(
-#                 "Currently only PyTorch models are supported (https://huggingface.co/models?library=pytorch&sort=downloads). To request another library, please contact us on our community slack (https://mindsdb.com/joincommunity)."
-#             )
-
-#         # check model task
-#         supported_tasks = [
-#             "text-classification",
-#             "text-generation",
-#             "zero-shot-classification",
-#             "translation",
-#             "summarization",
-#             "text2text-generation",
-#             "fill-mask",
-#         ]
-
-#         if metadata.pipeline_tag not in supported_tasks:
-#             raise Exception(
-#                 f'Not supported task for model: {metadata.pipeline_tag}.\
-#              Should be one of {", ".join(supported_tasks)}'
-#             )
-
-#         if "task" not in args:
-#             args["task"] = metadata.pipeline_tag
-#         elif args["task"] != metadata.pipeline_tag:
-#             raise Exception(
-#                 f'Task mismatch for model: {args["task"]}!={metadata.pipeline_tag}'
-#             )
-
-#         input_keys = list(args.keys())
-
-#         # task, model_name, input_column is essential
-#         for key in ["task", "model_name", "input_column"]:
-#             if key not in args:
-#                 raise Exception(f'Parameter "{key}" is required')
-#             input_keys.remove(key)
-
-#         # check tasks input
-
-#         if args["task"] == "zero-shot-classification":
-#             key = "candidate_labels"
-#             if key not in args:
-#                 raise Exception(
-#                     '"candidate_labels" is required for zero-shot-classification'
-#                 )
-#             input_keys.remove(key)
-
-#         if args["task"] == "translation":
-#             keys = ["lang_input", "lang_output"]
-#             for key in keys:
-#                 if key not in args:
-#                     raise Exception(f"{key} is required for translation")
-#                 input_keys.remove(key)
-
-#         if args["task"] == "summarization":
-#             keys = ["min_output_length", "max_output_length"]
-#             for key in keys:
-#                 if key not in args:
-#                     raise Exception(f"{key} is required for summarization")
-#                 input_keys.remove(key)
-
-#         # optional keys
-#         for key in ["labels", "max_length", "truncation_policy"]:
-#             if key in input_keys:
-#                 input_keys.remove(key)
-
-#         if len(input_keys) > 0:
-#             raise Exception(f'Not expected parameters: {", ".join(input_keys)}')
-
-#     def create(self, target, args=None, **kwargs):
-#         # TODO change BaseMLEngine api?
-#         if "using" in args:
-#             args = args["using"]
-
-#         args["target"] = target
-
-#         model_name = args["model_name"]
-#         hf_model_storage_path = self.engine_storage.folder_get(model_name)  # real
-
-#         if args["task"] == "translation":
-#             args[
-#                 "task_proper"
-#             ] = f"translation_{args['lang_input']}_to_{args['lang_output']}"
-#         else:
-#             args["task_proper"] = args["task"]
-
-#         logger.debug(f"Checking file system for {model_name}...")
-
-#         ####
-#         # Check if pipeline has already been downloaded
-#         try:
-#             pipeline = transformers.pipeline(task=args['task_proper'], model=hf_model_storage_path,
-#                                              tokenizer=hf_model_storage_path)
-#             logger.debug('Model already downloaded!')
-#         ####
-#         # Otherwise download it
-#         except (ValueError, OSError):
-#             try:
-#                 logger.debug(f"Downloading {model_name}...")
-#                 pipeline = transformers.pipeline(task=args['task_proper'], model=model_name)
-
-#                 pipeline.save_pretrained(hf_model_storage_path)
-
-#                 logger.debug(f"Saved to {hf_model_storage_path}")
-#             except Exception:
-#                 raise Exception(
-#                     "Error while downloading and setting up the model. Please try a different model. We're working on expanding the list of supported models, so we would appreciate it if you let us know about this in our community slack (https://mindsdb.com/joincommunity)."
-#                 )  # noqa
-#         ####
-
-#         if "max_length" in args:
-#             pass
-#         elif "max_position_embeddings" in pipeline.model.config.to_dict().keys():
-#             args["max_length"] = pipeline.model.config.max_position_embeddings
-#         elif "max_length" in pipeline.model.config.to_dict().keys():
-#             args["max_length"] = pipeline.model.config.max_length
-#         else:
-#             logger.debug('No max_length found!')
-
-#         labels_default = pipeline.model.config.id2label
-#         labels_map = {}
-#         if "labels" in args:
-#             for num in labels_default.keys():
-#                 labels_map[labels_default[num]] = args["labels"][num]
-#             args["labels_map"] = labels_map
-#         else:
-#             for num in labels_default.keys():
-#                 labels_map[labels_default[num]] = labels_default[num]
-#             args["labels_map"] = labels_map
-
-#         # store and persist in model folder
-#         self.model_storage.json_set("args", args)
-
-#         # persist changes to handler folder
-#         self.engine_storage.folder_sync(model_name)
-
-#     # todo move infer tasks to a seperate file
-#     def predict_text_classification(self, pipeline, item, args):
-#         top_k = args.get("top_k", 1000)
-
-#         result = pipeline(
-#             [item], top_k=top_k, truncation=True, max_length=args["max_length"]
-#         )[0]
-
-#         final = {}
-#         explain = {}
-#         if type(result) == dict:
-#             result = [result]
-#         final[args["target"]] = args["labels_map"][result[0]["label"]]
-#         for elem in result:
-#             if args["labels_map"]:
-#                 explain[args["labels_map"][elem["label"]]] = elem["score"]
-#             else:
-#                 explain[elem["label"]] = elem["score"]
-#         final[f"{args['target']}_explain"] = explain
-#         return final
-
-#     def predict_text_generation(self, pipeline, item, args):
-#         result = pipeline([item], max_length=args["max_length"])[0]
-
-#         final = {}
-#         final[args["target"]] = result["generated_text"]
-
-#         return final
-
-#     def predict_zero_shot(self, pipeline, item, args):
-#         top_k = args.get("top_k", 1000)
-
-#         result = pipeline(
-#             [item],
-#             candidate_labels=args["candidate_labels"],
-#             truncation=True,
-#             top_k=top_k,
-#             max_length=args["max_length"],
-#         )[0]
-
-#         final = {}
-#         final[args["target"]] = result["labels"][0]
-
-#         explain = dict(zip(result["labels"], result["scores"]))
-#         final[f"{args['target']}_explain"] = explain
-
-#         return final
-
-#     def predict_translation(self, pipeline, item, args):
-#         result = pipeline([item], max_length=args["max_length"])[0]
-
-#         final = {}
-#         final[args["target"]] = result["translation_text"]
-
-#         return final
-
-#     def predict_summarization(self, pipeline, item, args):
-#         result = pipeline(
-#             [item],
-#             min_length=args["min_output_length"],
-#             max_length=args["max_output_length"],
-#         )[0]
-
-#         final = {}
-#         final[args["target"]] = result["summary_text"]
-
-#         return final
-
-#     def predict_text2text(self, pipeline, item, args):
-#         result = pipeline([item], max_length=args["max_length"])[0]
-
-#         final = {}
-#         final[args["target"]] = result["generated_text"]
-
-#         return final
-
-#     def predict_fill_mask(self, pipeline, item, args):
-#         result = pipeline([item])[0]
-
-#         final = {}
-#         final[args["target"]] = result[0]["sequence"]
-#         explain = {elem["sequence"]: elem["score"] for elem in result}
-#         final[f"{args['target']}_explain"] = explain
-
-#         return final
-
-#     def predict(self, df, args=None):
-
-#         fnc_list = {
-#             "text-classification": self.predict_text_classification,
-#             "text-generation": self.predict_text_generation,
-#             "zero-shot-classification": self.predict_zero_shot,
-#             "translation": self.predict_translation,
-#             "summarization": self.predict_summarization,
-#             "fill-mask": self.predict_fill_mask,
-#         }
-
-#         # get stuff from model folder
-#         args = self.model_storage.json_get("args")
-
-#         task = args["task"]
-
-#         if task not in fnc_list:
-#             raise RuntimeError(f"Unknown task: {task}")
-
-#         fnc = fnc_list[task]
-
-#         try:
-#             # load from model storage (finetuned models will use this)
-#             hf_model_storage_path = self.model_storage.folder_get(
-#                 args["model_name"]
-#             )
-#             pipeline = transformers.pipeline(
-#                 task=args["task_proper"],
-#                 model=hf_model_storage_path,
-#                 tokenizer=hf_model_storage_path,
-#             )
-#         except (ValueError, OSError):
-#             # load from engine storage (i.e. 'common' models)
-#             hf_model_storage_path = self.engine_storage.folder_get(
-#                 args["model_name"]
-#             )
-#             pipeline = transformers.pipeline(
-#                 task=args["task_proper"],
-#                 model=hf_model_storage_path,
-#                 tokenizer=hf_model_storage_path,
-#             )
-
-#         input_column = args["input_column"]
-#         if input_column not in df.columns:
-#             raise RuntimeError(f'Column "{input_column}" not found in input data')
-#         input_list = df[input_column]
-
-#         max_tokens = pipeline.tokenizer.model_max_length
-
-#         results = []
-#         for item in input_list:
-#             if max_tokens is not None:
-#                 tokens = pipeline.tokenizer.encode(item)
-#                 if len(tokens) > max_tokens:
-#                     truncation_policy = args.get("truncation_policy", "strict")
-#                     if truncation_policy == "strict":
-#                         results.append(
-#                             {
-#                                 "error": f"Tokens count exceed model limit: {len(tokens)} > {max_tokens}"
-#                             }
-#                         )
-#                         continue
-#                     elif truncation_policy == "left":
-#                         tokens = tokens[
-#                             -max_tokens + 1: -1
-#                         ]  # cut 2 empty tokens from left and right
-#                     else:
-#                         tokens = tokens[
-#                             1: max_tokens - 1
-#                         ]  # cut 2 empty tokens from left and right
-
-#                     item = pipeline.tokenizer.decode(tokens)
-
-#             item = str(item)
-#             try:
-#                 result = fnc(pipeline, item, args)
-#             except Exception as e:
-#                 msg = str(e).strip()
-#                 if msg == "":
-#                     msg = e.__class__.__name__
-#                 result = {"error": msg}
-#             results.append(result)
-
-#         pred_df = pd.DataFrame(results)
-
-#         return pred_df
-
-#     def describe(self, attribute: Optional[str] = None) -> pd.DataFrame:
-#         args = self.model_storage.json_get("args")
-#         if attribute == "args":
-#             return pd.DataFrame(args.items(), columns=["key", "value"])
-#         elif attribute == "metadata":
-#             hf_api = HfApi()
-#             metadata = hf_api.model_info(args["model_name"])
-#             data = metadata.__dict__
-#             return pd.DataFrame(list(data.items()), columns=["key", "value"])
-#         else:
-#             tables = ["args", "metadata"]
-#             return pd.DataFrame(tables, columns=["tables"])
-
-#     def finetune(
-#         self, df: Optional[pd.DataFrame] = None, args: Optional[Dict] = None
-#     ) -> None:
-#         finetune_args = args if args else {}
-#         args = self.base_model_storage.json_get("args")
-#         args.update(finetune_args)
-
-#         model_name = args["model_name"]
-#         model_folder = self.model_storage.folder_get(model_name)
-#         args["model_folder"] = model_folder
-#         model_folder_name = model_folder.split("/")[-1]
-#         task = args["task"]
-
-#         if task not in FINETUNE_MAP:
-#             raise KeyError(
-#                 f"{task} is not currently supported, please choose a supported task - {', '.join(FINETUNE_MAP)}"
-#             )
-
-#         tokenizer, trainer = FINETUNE_MAP[task](df, args)
-
-#         try:
-#             trainer.train()
-#             trainer.save_model(
-#                 model_folder
-#             )  # TODO: save entire pipeline instead https://huggingface.co/docs/transformers/main_classes/pipelines#transformers.Pipeline.save_pretrained
-#             tokenizer.save_pretrained(model_folder)
-
-#             # persist changes
-#             self.model_storage.json_set("args", args)
-#             self.model_storage.folder_sync(model_folder_name)
-
-#         except Exception as e:
-#             err_str = f"Finetune failed with error: {str(e)}"
-#             logger.debug(err_str)
-#             raise Exception(err_str)
+from typing import Dict, Optional
+
+import pandas as pd
+import transformers
+from huggingface_hub import HfApi
+
+from mindsdb.integrations.handlers.huggingface_handler.settings import FINETUNE_MAP
+from mindsdb.integrations.libs.base import BaseMLEngine
+from mindsdb.utilities import log
+
+logger = log.getLogger(__name__)
+
+
+class HuggingFaceHandler(BaseMLEngine):
+    name = "huggingface"
+
+    @staticmethod
+    def create_validation(target, args=None, **kwargs):
+        if "using" in args:
+            args = args["using"]
+
+        hf_api = HfApi()
+
+        # check model is pytorch based
+        metadata = hf_api.model_info(args["model_name"])
+        if "pytorch" not in metadata.tags:
+            raise Exception(
+                "Currently only PyTorch models are supported (https://huggingface.co/models?library=pytorch&sort=downloads). To request another library, please contact us on our community slack (https://mindsdb.com/joincommunity)."
+            )
+
+        # check model task
+        supported_tasks = [
+            "text-classification",
+            "text-generation",
+            "zero-shot-classification",
+            "translation",
+            "summarization",
+            "text2text-generation",
+            "fill-mask",
+        ]
+
+        if metadata.pipeline_tag not in supported_tasks:
+            raise Exception(
+                f"Not supported task for model: {metadata.pipeline_tag}.\
+             Should be one of {', '.join(supported_tasks)}"
+            )
+
+        if "task" not in args:
+            args["task"] = metadata.pipeline_tag
+        elif args["task"] != metadata.pipeline_tag:
+            raise Exception(f"Task mismatch for model: {args['task']}!={metadata.pipeline_tag}")
+
+        input_keys = list(args.keys())
+
+        # task, model_name, input_column is essential
+        for key in ["task", "model_name", "input_column"]:
+            if key not in args:
+                raise Exception(f'Parameter "{key}" is required')
+            input_keys.remove(key)
+
+        # check tasks input
+
+        if args["task"] == "zero-shot-classification":
+            key = "candidate_labels"
+            if key not in args:
+                raise Exception('"candidate_labels" is required for zero-shot-classification')
+            input_keys.remove(key)
+
+        if args["task"] == "translation":
+            keys = ["lang_input", "lang_output"]
+            for key in keys:
+                if key not in args:
+                    raise Exception(f"{key} is required for translation")
+                input_keys.remove(key)
+
+        if args["task"] == "summarization":
+            keys = ["min_output_length", "max_output_length"]
+            for key in keys:
+                if key not in args:
+                    raise Exception(f"{key} is required for summarization")
+                input_keys.remove(key)
+
+        # optional keys
+        for key in ["labels", "max_length", "truncation_policy"]:
+            if key in input_keys:
+                input_keys.remove(key)
+
+        if len(input_keys) > 0:
+            raise Exception(f"Not expected parameters: {', '.join(input_keys)}")
+
+    def create(self, target, args=None, **kwargs):
+        # TODO change BaseMLEngine api?
+        if "using" in args:
+            args = args["using"]
+
+        args["target"] = target
+
+        model_name = args["model_name"]
+        hf_model_storage_path = self.engine_storage.folder_get(model_name)  # real
+
+        if args["task"] == "translation":
+            args["task_proper"] = f"translation_{args['lang_input']}_to_{args['lang_output']}"
+        else:
+            args["task_proper"] = args["task"]
+
+        logger.debug(f"Checking file system for {model_name}...")
+
+        ####
+        # Check if pipeline has already been downloaded
+        try:
+            pipeline = transformers.pipeline(
+                task=args["task_proper"], model=hf_model_storage_path, tokenizer=hf_model_storage_path
+            )
+            logger.debug("Model already downloaded!")
+        ####
+        # Otherwise download it
+        except (ValueError, OSError):
+            try:
+                logger.debug(f"Downloading {model_name}...")
+                pipeline = transformers.pipeline(task=args["task_proper"], model=model_name)
+
+                pipeline.save_pretrained(hf_model_storage_path)
+
+                logger.debug(f"Saved to {hf_model_storage_path}")
+            except Exception:
+                raise Exception(
+                    "Error while downloading and setting up the model. Please try a different model. We're working on expanding the list of supported models, so we would appreciate it if you let us know about this in our community slack (https://mindsdb.com/joincommunity)."
+                )  # noqa
+        ####
+
+        if "max_length" in args:
+            pass
+        elif "max_position_embeddings" in pipeline.model.config.to_dict().keys():
+            args["max_length"] = pipeline.model.config.max_position_embeddings
+        elif "max_length" in pipeline.model.config.to_dict().keys():
+            args["max_length"] = pipeline.model.config.max_length
+        else:
+            logger.debug("No max_length found!")
+
+        labels_default = pipeline.model.config.id2label
+        labels_map = {}
+        if "labels" in args:
+            for num in labels_default.keys():
+                labels_map[labels_default[num]] = args["labels"][num]
+            args["labels_map"] = labels_map
+        else:
+            for num in labels_default.keys():
+                labels_map[labels_default[num]] = labels_default[num]
+            args["labels_map"] = labels_map
+
+        # store and persist in model folder
+        self.model_storage.json_set("args", args)
+
+        # persist changes to handler folder
+        self.engine_storage.folder_sync(model_name)
+
+    # todo move infer tasks to a seperate file
+    def predict_text_classification(self, pipeline, item, args):
+        top_k = args.get("top_k", 1000)
+
+        result = pipeline([item], top_k=top_k, truncation=True, max_length=args["max_length"])[0]
+
+        final = {}
+        explain = {}
+        if type(result) == dict:
+            result = [result]
+        final[args["target"]] = args["labels_map"][result[0]["label"]]
+        for elem in result:
+            if args["labels_map"]:
+                explain[args["labels_map"][elem["label"]]] = elem["score"]
+            else:
+                explain[elem["label"]] = elem["score"]
+        final[f"{args['target']}_explain"] = explain
+        return final
+
+    def predict_text_generation(self, pipeline, item, args):
+        result = pipeline([item], max_length=args["max_length"])[0]
+
+        final = {}
+        final[args["target"]] = result["generated_text"]
+
+        return final
+
+    def predict_zero_shot(self, pipeline, item, args):
+        top_k = args.get("top_k", 1000)
+
+        result = pipeline(
+            [item],
+            candidate_labels=args["candidate_labels"],
+            truncation=True,
+            top_k=top_k,
+            max_length=args["max_length"],
+        )[0]
+
+        final = {}
+        final[args["target"]] = result["labels"][0]
+
+        explain = dict(zip(result["labels"], result["scores"]))
+        final[f"{args['target']}_explain"] = explain
+
+        return final
+
+    def predict_translation(self, pipeline, item, args):
+        result = pipeline([item], max_length=args["max_length"])[0]
+
+        final = {}
+        final[args["target"]] = result["translation_text"]
+
+        return final
+
+    def predict_summarization(self, pipeline, item, args):
+        result = pipeline(
+            [item],
+            min_length=args["min_output_length"],
+            max_length=args["max_output_length"],
+        )[0]
+
+        final = {}
+        final[args["target"]] = result["summary_text"]
+
+        return final
+
+    def predict_text2text(self, pipeline, item, args):
+        result = pipeline([item], max_length=args["max_length"])[0]
+
+        final = {}
+        final[args["target"]] = result["generated_text"]
+
+        return final
+
+    def predict_fill_mask(self, pipeline, item, args):
+        result = pipeline([item])[0]
+
+        final = {}
+        final[args["target"]] = result[0]["sequence"]
+        explain = {elem["sequence"]: elem["score"] for elem in result}
+        final[f"{args['target']}_explain"] = explain
+
+        return final
+
+    def predict(self, df, args=None):
+        fnc_list = {
+            "text-classification": self.predict_text_classification,
+            "text-generation": self.predict_text_generation,
+            "zero-shot-classification": self.predict_zero_shot,
+            "translation": self.predict_translation,
+            "summarization": self.predict_summarization,
+            "fill-mask": self.predict_fill_mask,
+        }
+
+        # get stuff from model folder
+        args = self.model_storage.json_get("args")
+
+        task = args["task"]
+
+        if task not in fnc_list:
+            raise RuntimeError(f"Unknown task: {task}")
+
+        fnc = fnc_list[task]
+
+        try:
+            # load from model storage (finetuned models will use this)
+            hf_model_storage_path = self.model_storage.folder_get(args["model_name"])
+            pipeline = transformers.pipeline(
+                task=args["task_proper"],
+                model=hf_model_storage_path,
+                tokenizer=hf_model_storage_path,
+            )
+        except (ValueError, OSError):
+            # load from engine storage (i.e. 'common' models)
+            hf_model_storage_path = self.engine_storage.folder_get(args["model_name"])
+            pipeline = transformers.pipeline(
+                task=args["task_proper"],
+                model=hf_model_storage_path,
+                tokenizer=hf_model_storage_path,
+            )
+
+        input_column = args["input_column"]
+        if input_column not in df.columns:
+            raise RuntimeError(f'Column "{input_column}" not found in input data')
+        input_list = df[input_column]
+
+        max_tokens = pipeline.tokenizer.model_max_length
+
+        results = []
+        for item in input_list:
+            if max_tokens is not None:
+                tokens = pipeline.tokenizer.encode(item)
+                if len(tokens) > max_tokens:
+                    truncation_policy = args.get("truncation_policy", "strict")
+                    if truncation_policy == "strict":
+                        results.append({"error": f"Tokens count exceed model limit: {len(tokens)} > {max_tokens}"})
+                        continue
+                    elif truncation_policy == "left":
+                        tokens = tokens[-max_tokens + 1 : -1]  # cut 2 empty tokens from left and right
+                    else:
+                        tokens = tokens[1 : max_tokens - 1]  # cut 2 empty tokens from left and right
+
+                    item = pipeline.tokenizer.decode(tokens)
+
+            item = str(item)
+            try:
+                result = fnc(pipeline, item, args)
+            except Exception as e:
+                msg = str(e).strip()
+                if msg == "":
+                    msg = e.__class__.__name__
+                result = {"error": msg}
+            results.append(result)
+
+        pred_df = pd.DataFrame(results)
+
+        return pred_df
+
+    def describe(self, attribute: Optional[str] = None) -> pd.DataFrame:
+        args = self.model_storage.json_get("args")
+        if attribute == "args":
+            return pd.DataFrame(args.items(), columns=["key", "value"])
+        elif attribute == "metadata":
+            hf_api = HfApi()
+            metadata = hf_api.model_info(args["model_name"])
+            data = metadata.__dict__
+            return pd.DataFrame(list(data.items()), columns=["key", "value"])
+        else:
+            tables = ["args", "metadata"]
+            return pd.DataFrame(tables, columns=["tables"])
+
+    def finetune(self, df: Optional[pd.DataFrame] = None, args: Optional[Dict] = None) -> None:
+        finetune_args = args if args else {}
+        args = self.base_model_storage.json_get("args")
+        args.update(finetune_args)
+
+        model_name = args["model_name"]
+        model_folder = self.model_storage.folder_get(model_name)
+        args["model_folder"] = model_folder
+        model_folder_name = model_folder.split("/")[-1]
+        task = args["task"]
+
+        if task not in FINETUNE_MAP:
+            raise KeyError(
+                f"{task} is not currently supported, please choose a supported task - {', '.join(FINETUNE_MAP)}"
+            )
+
+        tokenizer, trainer = FINETUNE_MAP[task](df, args)
+
+        try:
+            trainer.train()
+            trainer.save_model(
+                model_folder
+            )  # TODO: save entire pipeline instead https://huggingface.co/docs/transformers/main_classes/pipelines#transformers.Pipeline.save_pretrained
+            tokenizer.save_pretrained(model_folder)
+
+            # persist changes
+            self.model_storage.json_set("args", args)
+            self.model_storage.folder_sync(model_folder_name)
+
+        except Exception as e:
+            err_str = f"Finetune failed with error: {str(e)}"
+            logger.debug(err_str)
+            raise Exception(err_str)