Merge pull request #18 from PanQiWei/push_to_hub_integration

push_to_hub integration

auto_gptq/modeling/_base.py

@@ -10,6 +10,7 @@ import torch.nn as nn
 import transformers
 from safetensors.torch import load_file as safe_load, save_file as safe_save
 from transformers import AutoConfig, AutoModelForCausalLM, PreTrainedModel
+from transformers.utils.hub import PushToHubMixin
 
 from ._const import *
 from ._utils import *
@@ -19,7 +20,7 @@ logger = getLogger(__name__)
 
 
 @dataclass
-class BaseQuantizeConfig:
+class BaseQuantizeConfig(PushToHubMixin):
     bits: int = field(default=4, metadata={"choices": [2, 3, 4, 8]})
     damp_percent: float = field(default=0.01)
     desc_act: bool = field(default=True)
@@ -35,7 +36,7 @@ class BaseQuantizeConfig:
         if self.group_size != -1 and self.group_size <= 0:
             raise ValueError("unless equal to -1, group_size must greater then 0.")
 
-    def save_pretrained(self, save_dir: str):
+    def save_pretrained(self, save_dir: str, **kwargs):
         with open(join(save_dir, "quantize_config.json"), "w", encoding="utf-8") as f:
             json.dump(self.to_dict(), f, indent=2)
 
@@ -53,7 +54,7 @@ class BaseQuantizeConfig:
         }
 
 
-class BaseGPTQForCausalLM(nn.Module):
+class BaseGPTQForCausalLM(nn.Module, PushToHubMixin):
     layers_block_name: str = None
     outside_layer_modules: List[str] = None
     inside_layer_modules: List[List[str]] = None
@@ -293,6 +294,11 @@ class BaseGPTQForCausalLM(nn.Module):
         self.model.config.save_pretrained(save_dir)
         self.quantize_config.save_pretrained(save_dir)
 
+    def save_pretrained(self, save_dir: str, use_safetensors: bool = False, **kwargs):
+        """alias of save_quantized"""
+        logger.warning("you are using save_pretrained, which will re-direct to save_quantized.")
+        self.save_quantized(save_dir, use_safetensors)
+
     @classmethod
     def from_pretrained(
         cls,
@@ -322,7 +328,7 @@ class BaseGPTQForCausalLM(nn.Module):
 
         model = AutoModelForCausalLM.from_pretrained(pretrained_model_name_or_path, **model_init_kwargs)
         model_config = model.config.to_dict()
-        seq_len_keys = ["max_position_embeddings", "seq_length"]
+        seq_len_keys = ["max_position_embeddings", "seq_length", "n_positions"]
         if any([k in model_config for k in seq_len_keys]):
             for key in seq_len_keys:
                 if key in model_config:

auto_gptq/modeling/auto.py

@@ -1,4 +1,4 @@
-from ._base import BaseQuantizeConfig
+from ._base import BaseQuantizeConfig, BaseGPTQForCausalLM
 from ._utils import check_and_get_model_type
 from .bloom import BloomGPTQForCausalLM
 from .gpt_neox import GPTNeoXGPTQForCausalLM
@@ -33,7 +33,7 @@ class AutoGPTQForCausalLM:
         quantize_config: BaseQuantizeConfig,
         bf16: bool = False,
         **model_init_kwargs
-    ):
+    ) -> BaseGPTQForCausalLM:
         model_type = check_and_get_model_type(pretrained_model_name_or_path)
         return GPTQ_CAUSAL_LM_MODEL_MAP[model_type].from_pretrained(
             pretrained_model_name_or_path=pretrained_model_name_or_path,
@@ -49,7 +49,7 @@ class AutoGPTQForCausalLM:
         device: str = "cpu",
         use_safetensors: bool = False,
         use_triton: bool = False
-    ):
+    ) -> BaseGPTQForCausalLM:
         model_type = check_and_get_model_type(save_dir)
         return GPTQ_CAUSAL_LM_MODEL_MAP[model_type].from_quantized(
             save_dir=save_dir,

examples/README.md

@@ -57,3 +57,22 @@ CUDA_VISIBLE_DEVICES=0 python run_text_summarization_task.py --base_model_dir PA
 ```
 
 Use `--help` flag to see detailed descriptions for more command arguments.
+
+## Push To Hub
+> Commands in this chapter should be run under `push_to_hub` folder.
+
+You can upload and share your quantized model to Hugging Face Hub by using `push_to_hub` function.
+
+`push_quantized_model_to_hf_hub.py` provide a simple example to upload quantized model, tokenizer and configs at once.
+
+First, you need to login, run the following command in the virtual environment where Hugging Face Transformers is installed:
+```shell
+huggingface-cli login
+```
+
+Then run the script like this:
+```shell
+python push_quantized_model_to_hf_hub.py --quantized_model_dir PATH/TO/QUANTIZED/MODEL/DIR --tokenizer_dir PATH/TO/TOKENIZER/DIR --repo_id REPO/ID
+```
+
+Use `--help` flag to see detailed descriptions for more command arguments.

examples/evaluation/run_language_modeling_task.py

@@ -1,9 +1,10 @@
 import datasets
 from argparse import ArgumentParser
-from transformers import AutoTokenizer
 
+import torch
 from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
 from auto_gptq.eval_tasks import LanguageModelingTask
+from transformers import AutoTokenizer
 
 
 DATASET = "tatsu-lab/alpaca"
@@ -63,6 +64,7 @@ def main():
     task.model = None
     model.cpu()
     del model
+    torch.cuda.empty_cache()
 
     model = AutoGPTQForCausalLM.from_quantized(args.quantized_model_dir, device="cuda:0", use_triton=args.use_triton)
     task.model = model

examples/evaluation/run_sequence_classification_task.py

@@ -2,10 +2,10 @@ from argparse import ArgumentParser
 from functools import partial
 
 import datasets
-from transformers import AutoTokenizer
+import torch
-
 from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
 from auto_gptq.eval_tasks import SequenceClassificationTask
+from transformers import AutoTokenizer
 
 
 DATASET = "cardiffnlp/tweet_sentiment_multilingual"
@@ -67,6 +67,7 @@ def main():
     task.model = None
     model.cpu()
     del model
+    torch.cuda.empty_cache()
 
     model = AutoGPTQForCausalLM.from_quantized(args.quantized_model_dir, device="cuda:0", use_triton=args.use_triton)
     task.model = model

examples/evaluation/run_text_summarization_task.py

@@ -2,10 +2,10 @@ import os
 from argparse import ArgumentParser
 
 import datasets
-from transformers import AutoTokenizer, GenerationConfig
+import torch
-
 from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
 from auto_gptq.eval_tasks import TextSummarizationTask
+from transformers import AutoTokenizer, GenerationConfig
 
 
 os.system("pip install py7zr")
@@ -61,6 +61,7 @@ def main():
     task.model = None
     model.cpu()
     del model
+    torch.cuda.empty_cache()
 
     model = AutoGPTQForCausalLM.from_quantized(args.quantized_model_dir, device="cuda:0", use_triton=args.use_triton)
     task.model = model

examples/push_to_hub/push_quantized_model_to_hf_hub.py

@@ -0,0 +1,55 @@
+from argparse import ArgumentParser
+
+from auto_gptq import AutoGPTQForCausalLM
+from transformers import AutoTokenizer
+
+
+def main():
+    parser = ArgumentParser()
+    parser.add_argument("--quantized_model_dir", type=str, help="Directory that saves quantized model.")
+    parser.add_argument("--repo_id", type=str, help="The name of the repository you want to push to.")
+    parser.add_argument(
+        "--tokenizer_dir",
+        type=str,
+        default=None,
+        help="Directory that saves tokenizer, defaults to None, will not upload tokenizer if not specified."
+    )
+    parser.add_argument("--commit_message", type=str, default=None, help="Message to commit while pushing.")
+    parser.add_argument(
+        "--device",
+        type=str,
+        default="cpu",
+        choices=["cpu", "cuda"],
+        help="Which device to load the model."
+    )
+    parser.add_argument(
+        "--private",
+        action="store_true",
+        help="Whether or not the repository created should be private."
+    )
+    parser.add_argument(
+        "--use_temp_dir",
+        action="store_true",
+        help="Whether or not to use a temporary directory to store the files saved before they are pushed to the Hub."
+    )
+    args = parser.parse_args()
+
+    push_to_hub_kwargs = {
+        "repo_id": args.repo_id,
+        "commit_message": args.commit_message,
+        "private": args.private,
+        "use_temp_dir": args.use_temp_dir
+    }
+
+    model = AutoGPTQForCausalLM.from_quantized(args.quantized_model_dir, device=args.device)
+    model.push_to_hub(**push_to_hub_kwargs)
+    model.config.push_to_hub(**push_to_hub_kwargs)
+    model.quantize_config.push_to_hub(**push_to_hub_kwargs)
+
+    if args.tokenizer_dir:
+        tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_dir)
+        tokenizer.push_to_hub(**push_to_hub_kwargs)
+
+
+if __name__ == "__main__":
+    main()