hf_text-generation-inference/server/text_generation_server/models/flash_llama.py

79 lines
2.5 KiB
Python
Raw Normal View History

import torch
import torch.distributed
from opentelemetry import trace
from transformers import AutoConfig
feat(server): Add inference support for GPTQ (llama + falcon tested) + Quantization script (#438) Let's start discussing implementation. - Need to expose the quantization scripts (either included here or add doc on how to use https://github.com/qwopqwop200/GPTQ-for-LLaMa) - Make sure GPTQ works for multiple models (priority to Falcon). Currently it means that every place we use `get_{tensor|sharded}` to check for quantization. My idea is to reintegrate as much as possible into `utils/layer.py` by expanding `load_multi` to be a bit more generic. This might require some thinking, but ultimately the `qweight,qzeros,scales,g_idx` should be in a single place, and independant of bias presence. # 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 --> --------- Co-authored-by: Ubuntu <ubuntu@ip-172-31-41-161.ec2.internal> Co-authored-by: OlivierDehaene <olivier@huggingface.co>
2023-06-26 04:27:01 -06:00
from transformers.models.llama import LlamaTokenizer, LlamaTokenizerFast
from typing import Optional
from text_generation_server.models import FlashCausalLM
from text_generation_server.models.custom_modeling.flash_llama_modeling import (
FlashLlamaForCausalLM,
)
from text_generation_server.utils import (
initialize_torch_distributed,
weight_files,
Weights,
)
tracer = trace.get_tracer(__name__)
class FlashLlama(FlashCausalLM):
def __init__(
feat(server): GPTQ quantization (step1) (#277) Changes only the type from `bool` to `Option<Enum>` pretty much everywhere. - Use `Optional[str]` in Python (easier to manage than importing type everywhere). Except for the cli to get proper validation - Updated all models to handle gracefully new values. (Error out if unknown value, or gptq since not implemented). <!-- 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 -->
2023-05-12 06:46:41 -06:00
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("FlashLlama is only available on GPU")
feat(server): Add inference support for GPTQ (llama + falcon tested) + Quantization script (#438) Let's start discussing implementation. - Need to expose the quantization scripts (either included here or add doc on how to use https://github.com/qwopqwop200/GPTQ-for-LLaMa) - Make sure GPTQ works for multiple models (priority to Falcon). Currently it means that every place we use `get_{tensor|sharded}` to check for quantization. My idea is to reintegrate as much as possible into `utils/layer.py` by expanding `load_multi` to be a bit more generic. This might require some thinking, but ultimately the `qweight,qzeros,scales,g_idx` should be in a single place, and independant of bias presence. # 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 --> --------- Co-authored-by: Ubuntu <ubuntu@ip-172-31-41-161.ec2.internal> Co-authored-by: OlivierDehaene <olivier@huggingface.co>
2023-06-26 04:27:01 -06:00
try:
tokenizer = LlamaTokenizer.from_pretrained(
model_id,
revision=revision,
padding_side="left",
truncation_side="left",
trust_remote_code=trust_remote_code,
)
except Exception:
tokenizer = LlamaTokenizerFast.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=trust_remote_code
)
torch.distributed.barrier(group=self.process_group)
filenames = weight_files(model_id, revision=revision, extension=".safetensors")
weights = Weights(filenames, device, dtype, process_group=self.process_group)
config.quantize = quantize
model = FlashLlamaForCausalLM(config, weights)
torch.distributed.barrier(group=self.process_group)
super(FlashLlama, self).__init__(
model=model,
tokenizer=tokenizer,
num_layers=len(model.model.layers),
num_kv_heads=model.model.num_key_value_heads,
head_size=model.model.head_size,
dtype=dtype,
device=device,
rank=rank,
world_size=world_size,
)