mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-20 06:12:07 +00:00
a0d55358d2
- Current PR is not great because we're side stepping the
`Weights.__init__` but Weights shouldn't requires anything related
to the config or the model_id as it aims to be a simple Wrapper
over multi file loading.
- Ideal solution would be to use something like Rust enum
```
enum Quantize{
Bitandbytes(Bitsandbytes),
GPTQ(bits: usize, groupsize: usize)
```
And passing that around during load. Unfortunately we don't
have access to this, so for now, side-stepping seems easier.
- Re-enabling groupsize<0 with exllama (confirmed it works.)
Helps #601
In next steps we should make sure our quantization script uses that
format and make it standard.
# What does this PR do?
<!--
Congratulations! You've made it this far! You're not quite done yet
though.
Once merged, your PR is going to appear in the release notes with the
title you set, so make sure it's a great title that fully reflects the
extent of your awesome contribution.
Then, please replace this with a description of the change and which
issue is fixed (if applicable). Please also include relevant motivation
and context. List any dependencies (if any) that are required for this
change.
Once you're done, someone will review your PR shortly (see the section
"Who can review?" below to tag some potential reviewers). They may
suggest changes to make the code even better. If no one reviewed your PR
after a week has passed, don't hesitate to post a new comment
@-mentioning the same persons---sometimes notifications get lost.
-->
<!-- Remove if not applicable -->
Fixes # (issue)
## Before submitting
- [ ] This PR fixes a typo or improves the docs (you can dismiss the
other checks if that's the case).
- [ ] Did you read the [contributor
guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#start-contributing-pull-requests),
Pull Request section?
- [ ] Was this discussed/approved via a Github issue or the
[forum](https://discuss.huggingface.co/)? Please add a link
to it if that's the case.
- [ ] Did you make sure to update the documentation with your changes?
Here are the
[documentation
guidelines](https://github.com/huggingface/transformers/tree/main/docs),
and
[here are tips on formatting
docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation).
- [ ] Did you write any new necessary tests?
## Who can review?
Anyone in the community is free to review the PR once the tests have
passed. Feel free to tag
members/contributors who may be interested in your PR.
<!-- Your PR will be replied to more quickly if you can figure out the
right person to tag with @
@OlivierDehaene OR @Narsil
-->
88 lines
2.8 KiB
Python
88 lines
2.8 KiB
Python
import torch
|
|
import torch.distributed
|
|
|
|
from opentelemetry import trace
|
|
from transformers import AutoTokenizer, AutoConfig
|
|
from typing import Optional, List
|
|
import json
|
|
import os
|
|
|
|
from huggingface_hub import hf_hub_download
|
|
from text_generation_server.models import FlashCausalLM
|
|
from text_generation_server.models.custom_modeling.flash_santacoder_modeling import (
|
|
FlashSantacoderForCausalLM,
|
|
)
|
|
from text_generation_server.utils import (
|
|
initialize_torch_distributed,
|
|
weight_files,
|
|
Weights,
|
|
)
|
|
|
|
tracer = trace.get_tracer(__name__)
|
|
|
|
|
|
class FlashSantacoderSharded(FlashCausalLM):
|
|
def __init__(
|
|
self,
|
|
model_id: str,
|
|
revision: Optional[str] = None,
|
|
quantize: Optional[str] = None,
|
|
dtype: Optional[torch.dtype] = None,
|
|
trust_remote_code: bool = False,
|
|
):
|
|
self.process_group, rank, world_size = initialize_torch_distributed()
|
|
if torch.cuda.is_available():
|
|
device = torch.device(f"cuda:{rank}")
|
|
dtype = torch.float16 if dtype is None else dtype
|
|
else:
|
|
raise NotImplementedError("FlashSantacoderSharded is only available on GPU")
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(
|
|
model_id,
|
|
revision=revision,
|
|
padding_side="left",
|
|
truncation_side="left",
|
|
trust_remote_code=trust_remote_code,
|
|
)
|
|
|
|
config = AutoConfig.from_pretrained(
|
|
model_id,
|
|
revision=revision,
|
|
trust_remote_code=True,
|
|
)
|
|
config.quantize = quantize
|
|
config.transpose = config.architectures[0].startswith("GPT2")
|
|
|
|
torch.distributed.barrier(group=self.process_group)
|
|
filenames = weight_files(model_id, revision=revision, extension=".safetensors")
|
|
weights = Weights(
|
|
filenames,
|
|
device=device,
|
|
dtype=dtype,
|
|
process_group=self.process_group,
|
|
aliases={"transformer.wte.weight": ["lm_head.weight"]},
|
|
)
|
|
if config.quantize == "gptq":
|
|
weights._set_gptq_params(model_id)
|
|
|
|
model = FlashSantacoderForCausalLM(config, weights)
|
|
|
|
torch.distributed.barrier(group=self.process_group)
|
|
super(FlashSantacoderSharded, self).__init__(
|
|
model=model.to(device),
|
|
tokenizer=tokenizer,
|
|
num_layers=len(model.transformer.h),
|
|
num_kv_heads=1,
|
|
head_size=model.transformer.head_size,
|
|
dtype=dtype,
|
|
device=device,
|
|
rank=rank,
|
|
world_size=world_size,
|
|
)
|
|
|
|
def decode(self, generated_ids: List[int]) -> str:
|
|
# Do not skip special tokens as they are used for custom parsing rules of the generated text
|
|
return self.tokenizer.decode(
|
|
generated_ids, skip_special_tokens=False, clean_up_tokenization_spaces=False
|
|
)
|