# Very Slow Fine Tuning Performance for Speech?

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fvery-slow-fine-tuning-performance-for-speech%2F25431%3E%5C
**Category:** 🤗Transformers\
**Created:** [November 2, 2022, 6:53pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fvery-slow-fine-tuning-performance-for-speech%2F25431 "2022-11-02T18:53:54Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![picheny](https://avatars.discourse-cdn.com/v4/letter/p/77aa72/32.png) [@picheny](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fpicheny)%5C
**Post date:** [November 2, 2022, 6:53pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fvery-slow-fine-tuning-performance-for-speech%2F25431%2F1 "2022-11-02T18:53:54Z")

</div>

I am trying to fine tune the Facebook 60K hour wav2vec2 model as described in Patrick von Platen’s article: [Fine-Tune Wav2Vec2 for English ASR in Hugging Face with 🤗 Transformers](https://huggingface.co/blog/fine-tune-wav2vec2-english)

I have about 200 hours of speech data. I am using 3 RTX8000 GPUs on a 48 Core Lenovo SR670 with 369 GB of memory (on our NYU compute cluster).

It seems to be using all 3 GPUs (utilization is high) but seems as slow as molasses. If I read the output right it looks like it will take over a week to fine tune the model with 30 epochs.

Does this sound correct?

Thanks  
Michael Picheny

---

<div class="post-metadata">

**Author:** ![seba3y](https://avatars.discourse-cdn.com/v4/letter/s/e68b1a/32.png) [@seba3y](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fseba3y)%5C
**Post date:** [August 14, 2023, 6:06pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fvery-slow-fine-tuning-performance-for-speech%2F25431%2F2 "2023-08-14T18:06:32Z")

</div>

are you found out the problem?

---

<div class="post-metadata">

**Author:** ![picheny](https://avatars.discourse-cdn.com/v4/letter/p/77aa72/32.png) [@picheny](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fpicheny)%5C
**Post date:** [August 14, 2023, 6:35pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fvery-slow-fine-tuning-performance-for-speech%2F25431%2F3 "2023-08-14T18:35:13Z")

</div>

I just ran it for a week.

---

<div class="post-metadata">

**Author:** ![seba3y](https://avatars.discourse-cdn.com/v4/letter/s/e68b1a/32.png) [@seba3y](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fseba3y)%5C
**Post date:** [August 14, 2023, 8:30pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fvery-slow-fine-tuning-performance-for-speech%2F25431%2F4 "2023-08-14T20:30:32Z")

</div>

dam, i’m working with 1200 hour of Bengal speech and use Lora and PEFT for finetune the model but time and resource usage is equivalent to full finetune of the model.