SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

GPT-J specific half precision on CPU note (#22086)

Browse Source 
* re: #21989

* update re: #21989

* removed cpu option

* make style
This commit is contained in:
Maria Khalusova
2023-03-10 14:03:43 -05:00
committed by GitHub
parent 2f4cdd97f5
commit bdec2768bd
1 changed files with 13 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
docs/source/en/model_doc/gptj.mdx
View File

@@ -21,21 +21,22 @@ This model was contributed by [Stella Biderman](https://huggingface.co/stellaath
Tips: Tips:
- To load [GPT-J](https://huggingface.co/EleutherAI/gpt-j-6B) in float32 one would need at least 2x model size CPU - To load [GPT-J](https://huggingface.co/EleutherAI/gpt-j-6B) in float32 one would need at least 2x model size
RAM: 1x for initial weights and another 1x to load the checkpoint. So for GPT-J it would take at least 48GB of CPU RAM: 1x for initial weights and another 1x to load the checkpoint. So for GPT-J it would take at least 48GB
RAM to just load the model. To reduce the CPU RAM usage there are a few options. The `torch_dtype` argument can be RAM to just load the model. To reduce the RAM usage there are a few options. The `torch_dtype` argument can be
used to initialize the model in half-precision. And the `low_cpu_mem_usage` argument can be used to keep the RAM used to initialize the model in half-precision on a CUDA device only. There is also a fp16 branch which stores the fp16 weights,
usage to 1x. There is also a [fp16 branch](https://huggingface.co/EleutherAI/gpt-j-6B/tree/float16) which stores which could be used to further minimize the RAM usage:
the fp16 weights, which could be used to further minimize the RAM usage. Combining all this it should take roughly
12.1GB of CPU RAM to load the model.
```python ```python
>>> from transformers import GPTJForCausalLM >>> from transformers import GPTJForCausalLM
>>> import torch >>> import torch
>>> device = "cuda"
>>> model = GPTJForCausalLM.from_pretrained( >>> model = GPTJForCausalLM.from_pretrained(
... "EleutherAI/gpt-j-6B", revision="float16", torch_dtype=torch.float16, low_cpu_mem_usage=True ... "EleutherAI/gpt-j-6B",
... ) ... revision="float16",
... torch_dtype=torch.float16,
... ).to(device)
``` ```
- The model should fit on 16GB GPU for inference. For training/fine-tuning it would take much more GPU RAM. Adam - The model should fit on 16GB GPU for inference. For training/fine-tuning it would take much more GPU RAM. Adam
@@ -85,7 +86,8 @@ model.
>>> from transformers import GPTJForCausalLM, AutoTokenizer >>> from transformers import GPTJForCausalLM, AutoTokenizer
>>> import torch >>> import torch
>>> model = GPTJForCausalLM.from_pretrained("EleutherAI/gpt-j-6B", torch_dtype=torch.float16) >>> device = "cuda"
>>> model = GPTJForCausalLM.from_pretrained("EleutherAI/gpt-j-6B", torch_dtype=torch.float16).to(device)
>>> tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B") >>> tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B")
>>> prompt = ( >>> prompt = (
@@ -94,7 +96,7 @@ model.
... "researchers was the fact that the unicorns spoke perfect English." ... "researchers was the fact that the unicorns spoke perfect English."
... ) ... )
>>> input_ids = tokenizer(prompt, return_tensors="pt").input_ids >>> input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(device)
>>> gen_tokens = model.generate( >>> gen_tokens = model.generate(
... input_ids, ... input_ids,