# Incrementally finetuning a HF model in SageMaker
**URL:**
**Author:**  [@MaximusDecimusMeridi](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FMaximusDecimusMeridi)%5C
**Post date:** [May 2, 2022, 7:46pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fincrementally-finetuning-a-hf-model-in-sagemaker%2F17443%2F1 "2022-05-02T19:46:31Z")
If I have an already pretrained model, and want to train it on some additional data, how do I go about that in SageMaker?
In this case I’ve trained a model loaded with AutoModelForSequenceClassification using “distilbert-base-uncased”.
---
**Author:**  [@philschmid](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fphilschmid)%5C
**Post date:** [May 3, 2022, 6:19am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fincrementally-finetuning-a-hf-model-in-sagemaker%2F17443%2F2 "2022-05-03T06:19:15Z")
You can provide your trained model as model\_id instead of the base model this would further train it. But you need to make sure that you have the same labels etc.
---
**Author:**  [@petarulev](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fpetarulev)%5C
**Post date:** [May 3, 2022, 8:24am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fincrementally-finetuning-a-hf-model-in-sagemaker%2F17443%2F3 "2022-05-03T08:24:00Z")
> [@MaximusDecimusMeridi](#):
>
> In this case I’ve trained a model loaded with AutoModelForSequenceClassification using “distilbert-base-uncased”.
Hello!
Consider using huggingface’s Trainer and TrainingArguments. Please check this page: [Trainer](https://huggingface.co/docs/transformers/main_classes/trainer) .
If you have any questions on how to do it, let me know and I will forward some other links or help you do it. 🙂
---
**Author:**  [@MaximusDecimusMeridi](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FMaximusDecimusMeridi)%5C
**Post date:** [May 3, 2022, 3:39pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fincrementally-finetuning-a-hf-model-in-sagemaker%2F17443%2F4 "2022-05-03T15:39:35Z")
> [@petarulev](#):
>
> If you have any questions on how to do it, let me know and I will forward some other links or help you do it.
Thank you. Yes if you can please share any other links with additional context/backgrouynd that would be great 🙏
---
**Author:**  [@MaximusDecimusMeridi](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FMaximusDecimusMeridi)%5C
**Post date:** [May 3, 2022, 8:45pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fincrementally-finetuning-a-hf-model-in-sagemaker%2F17443%2F5 "2022-05-03T20:45:10Z")
> [@philschmid](#):
>
> You can provide your trained model as model\_id
Can I do this without using the hub? Would I just unzip the model into a local directory and load this and train with data and use same labels?
Thanks
---
**Author:**  [@marshmellow77](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fmarshmellow77)%5C
**Post date:** [May 4, 2022, 2:30am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fincrementally-finetuning-a-hf-model-in-sagemaker%2F17443%2F6 "2022-05-04T02:30:14Z")
@MaximusDecimusMeridi - if you want to do this without using the model hub then, yes, downloading the model from S3, unzipping it into a local folder, and using `model = AutoModelForSequenceClassification.from_pretrained('path/to/model/folder')` should work!
Cheers
Heiko
---
**Author:**  [@philschmid](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fphilschmid)%5C
**Post date:** [May 4, 2022, 6:05am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fincrementally-finetuning-a-hf-model-in-sagemaker%2F17443%2F7 "2022-05-04T06:05:45Z")
You can also provide the model in a similar fashion you do with your training in the estimator, e.g. `huggingface_estimator.fit({"train":s3_train_uri,"model":s3_model_uri})` which should download the s3\_model\_uri into /opt/ml/input/model or rather speaking stores the path in the env var `SM_CHANNEL_MODEL`