2024-05-31 11:51:42 +00:00
|
|
|
from io import BytesIO
|
|
|
|
from PIL import Image
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
import torch
|
2023-12-14 14:59:38 +00:00
|
|
|
import time
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
from dataclasses import dataclass
|
|
|
|
from opentelemetry import trace
|
2023-09-27 10:22:09 +00:00
|
|
|
from transformers import (
|
2024-10-02 09:22:13 +00:00
|
|
|
AutoConfig,
|
2023-09-27 10:22:09 +00:00
|
|
|
AutoProcessor,
|
|
|
|
AutoTokenizer,
|
|
|
|
PreTrainedTokenizerBase,
|
|
|
|
ProcessorMixin,
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
from typing import Optional, Tuple, List, Type, Dict
|
|
|
|
|
|
|
|
from text_generation_server.models import Model
|
|
|
|
from text_generation_server.models.types import (
|
|
|
|
Batch,
|
2023-12-11 11:46:30 +00:00
|
|
|
Tokens,
|
2023-08-17 12:38:49 +00:00
|
|
|
Generation,
|
|
|
|
GeneratedText,
|
|
|
|
)
|
|
|
|
from text_generation_server.pb import generate_pb2
|
|
|
|
from text_generation_server.utils import NextTokenChooser, StoppingCriteria, Sampling
|
2024-10-02 09:22:13 +00:00
|
|
|
import torch.distributed
|
|
|
|
from text_generation_server.models.custom_modeling.idefics_modeling import (
|
|
|
|
IdeficsForVisionText2Text,
|
|
|
|
)
|
|
|
|
from text_generation_server.utils import (
|
|
|
|
initialize_torch_distributed,
|
|
|
|
weight_files,
|
|
|
|
Weights,
|
|
|
|
)
|
|
|
|
from text_generation_server.utils.quantization import get_loader
|
|
|
|
|
|
|
|
from text_generation_server.utils.import_utils import SYSTEM
|
2023-09-27 10:22:09 +00:00
|
|
|
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
tracer = trace.get_tracer(__name__)
|
|
|
|
|
|
|
|
|
|
|
|
@dataclass
|
|
|
|
class IdeficsCausalLMBatch(Batch):
|
|
|
|
batch_id: int
|
|
|
|
requests: List[generate_pb2.Request]
|
|
|
|
requests_idx_mapping: Dict[int, int]
|
|
|
|
|
|
|
|
# Decoder values
|
|
|
|
input_ids: torch.Tensor
|
|
|
|
attention_mask: torch.Tensor
|
|
|
|
position_ids: torch.Tensor
|
|
|
|
pixel_values: Optional[torch.Tensor]
|
|
|
|
image_hidden_states: Optional[torch.Tensor]
|
|
|
|
image_attention_mask: Optional[torch.Tensor]
|
|
|
|
past_key_values: Optional[List[Tuple]]
|
|
|
|
|
|
|
|
# All tokens
|
|
|
|
all_input_ids: List[torch.Tensor]
|
|
|
|
|
|
|
|
# Lengths of all generations present in the batch
|
|
|
|
input_lengths: List[int]
|
|
|
|
prefix_offsets: List[int]
|
|
|
|
read_offsets: List[int]
|
|
|
|
|
|
|
|
# Generation helpers
|
|
|
|
next_token_choosers: List[NextTokenChooser]
|
|
|
|
stopping_criterias: List[StoppingCriteria]
|
|
|
|
|
|
|
|
# Metadata used for padding
|
|
|
|
max_input_length: int
|
|
|
|
padding_right_offset: int
|
|
|
|
|
|
|
|
# Maximum number of tokens this batch will grow to
|
|
|
|
max_tokens: int
|
|
|
|
|
|
|
|
# Past metadata
|
|
|
|
keys_head_dim_last: bool = True
|
|
|
|
|
|
|
|
def to_pb(self) -> generate_pb2.CachedBatch:
|
|
|
|
return generate_pb2.CachedBatch(
|
|
|
|
id=self.batch_id,
|
|
|
|
request_ids=[r.id for r in self.requests],
|
|
|
|
size=len(self),
|
|
|
|
max_tokens=self.max_tokens,
|
|
|
|
)
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
def from_pb(
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
cls,
|
|
|
|
pb: generate_pb2.Batch,
|
|
|
|
tokenizer: PreTrainedTokenizerBase,
|
|
|
|
dtype: torch.dtype,
|
|
|
|
device: torch.device,
|
|
|
|
) -> "IdeficsCausalLMBatch":
|
|
|
|
raise NotImplementedError
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
def from_pb_processor(
|
2023-08-17 12:38:49 +00:00
|
|
|
cls,
|
|
|
|
pb: generate_pb2.Batch,
|
|
|
|
tokenizer: PreTrainedTokenizerBase,
|
2023-09-27 10:22:09 +00:00
|
|
|
processor: ProcessorMixin, # Hack
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
config,
|
2023-08-17 12:38:49 +00:00
|
|
|
dtype: torch.dtype,
|
|
|
|
device: torch.device,
|
|
|
|
) -> "IdeficsCausalLMBatch":
|
|
|
|
inputs = []
|
|
|
|
next_token_choosers = []
|
|
|
|
stopping_criterias = []
|
|
|
|
prefix_offsets = []
|
|
|
|
read_offsets = []
|
|
|
|
requests_idx_mapping = {}
|
|
|
|
|
|
|
|
# Parse batch
|
|
|
|
max_truncation = 0
|
|
|
|
padding_right_offset = 0
|
|
|
|
max_decode_tokens = 0
|
|
|
|
for i, r in enumerate(pb.requests):
|
|
|
|
requests_idx_mapping[r.id] = i
|
2024-05-31 11:51:42 +00:00
|
|
|
inputs.append(r.input_chunks.chunks)
|
2024-02-16 10:58:58 +00:00
|
|
|
next_token_choosers.append(
|
|
|
|
NextTokenChooser.from_pb(r.parameters, device, tokenizer)
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
stopping_criteria = StoppingCriteria.from_pb(
|
|
|
|
r.stopping_parameters, tokenizer
|
|
|
|
)
|
|
|
|
stopping_criterias.append(stopping_criteria)
|
|
|
|
max_truncation = max(max_truncation, r.truncate)
|
|
|
|
max_decode_tokens += stopping_criteria.max_new_tokens
|
|
|
|
padding_right_offset = max(
|
|
|
|
padding_right_offset, stopping_criteria.max_new_tokens
|
|
|
|
)
|
|
|
|
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
# TODO Check impact on idefics
|
2023-08-17 12:38:49 +00:00
|
|
|
prompts = []
|
|
|
|
for inp in inputs:
|
|
|
|
# Each input is encoded into a list, where each element of this input list is either a string or a URL
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
prompt = []
|
2024-05-31 11:51:42 +00:00
|
|
|
for chunk in inp:
|
|
|
|
chunk_type = chunk.WhichOneof("chunk")
|
|
|
|
if chunk_type == "text":
|
|
|
|
prompt.append(chunk.text)
|
|
|
|
elif chunk_type == "image":
|
|
|
|
image = Image.open(BytesIO(chunk.image.data))
|
|
|
|
prompt.append(image)
|
|
|
|
else:
|
|
|
|
raise RuntimeError(f"Invalid chunk type {chunk_type}")
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
prompts.append(prompt)
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
# The processor replaces the call to tokenizer, and
|
|
|
|
# a/ takes care of fetching images from the URL
|
|
|
|
# b/ generate the correct input_ids, attention_mask, pixel_values, image_attention_mask to feed to the model
|
|
|
|
tokenized_inputs = processor(
|
|
|
|
prompts,
|
|
|
|
return_tensors="pt",
|
|
|
|
padding=True,
|
|
|
|
truncation=True,
|
|
|
|
max_length=max_truncation,
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
# TODO Check impact on idefics
|
|
|
|
# add_end_of_utterance_token=False, # Already taken care of inside the prompts, so bypassing the processor's handling of this token
|
2023-08-17 12:38:49 +00:00
|
|
|
).to(device)
|
|
|
|
for _ in pb.requests:
|
|
|
|
input_len = tokenized_inputs["input_ids"].shape[1]
|
2023-09-27 10:22:09 +00:00
|
|
|
prefix_offsets.append(
|
|
|
|
input_len - 5
|
|
|
|
) # To decode without potential fallbacks errors
|
|
|
|
read_offsets.append(
|
|
|
|
input_len
|
|
|
|
) # To decode without potential fallbacks errors
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
input_lengths = tokenized_inputs["attention_mask"].sum(1)
|
|
|
|
max_input_length = input_lengths.max()
|
|
|
|
|
|
|
|
input_ids = tokenized_inputs["input_ids"]
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
pixel_values = tokenized_inputs.get("pixel_values", None)
|
2023-08-17 12:38:49 +00:00
|
|
|
image_hidden_states = None
|
|
|
|
# Allocate maximum attention_mask
|
|
|
|
attention_mask = input_ids.new_zeros(
|
|
|
|
(pb.size, max_input_length + padding_right_offset)
|
|
|
|
)
|
|
|
|
# Copy tokenizer attention_mask into fully allocated attention_mask
|
|
|
|
attention_mask[:, :max_input_length] = tokenized_inputs["attention_mask"]
|
|
|
|
# Do the same for image_attention_mask
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
if pixel_values is None:
|
|
|
|
image_attention_mask = None
|
|
|
|
else:
|
|
|
|
image_attention_mask = input_ids.new_zeros(
|
|
|
|
(
|
|
|
|
pb.size,
|
|
|
|
max_input_length + padding_right_offset,
|
|
|
|
pixel_values.size(1),
|
|
|
|
)
|
2023-09-27 10:22:09 +00:00
|
|
|
)
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
image_attention_mask[:, :max_input_length, :] = tokenized_inputs[
|
|
|
|
"image_attention_mask"
|
|
|
|
]
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
position_ids = tokenized_inputs["attention_mask"].long().cumsum(-1) - 1
|
|
|
|
position_ids.masked_fill_(tokenized_inputs["attention_mask"] == 0, 1)
|
2023-09-27 10:22:09 +00:00
|
|
|
all_input_ids = tokenized_inputs["input_ids"].T.split(
|
|
|
|
1, dim=1
|
|
|
|
) # It's input_ids but splitted into a tuple of tensors where each tensor is (seq_len, 1) size. It is then transformed into a list
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
max_tokens = len(inputs) * (max_input_length + max_decode_tokens)
|
|
|
|
|
|
|
|
return cls(
|
|
|
|
batch_id=pb.id,
|
|
|
|
requests=pb.requests,
|
|
|
|
requests_idx_mapping=requests_idx_mapping,
|
|
|
|
input_ids=input_ids,
|
|
|
|
attention_mask=attention_mask,
|
|
|
|
position_ids=position_ids,
|
|
|
|
pixel_values=pixel_values,
|
|
|
|
image_hidden_states=image_hidden_states,
|
|
|
|
image_attention_mask=image_attention_mask,
|
|
|
|
past_key_values=None,
|
|
|
|
all_input_ids=list(all_input_ids),
|
|
|
|
input_lengths=input_lengths.tolist(),
|
|
|
|
prefix_offsets=prefix_offsets,
|
|
|
|
read_offsets=read_offsets,
|
|
|
|
next_token_choosers=next_token_choosers,
|
|
|
|
stopping_criterias=stopping_criterias,
|
|
|
|
max_input_length=max_input_length.item(),
|
|
|
|
padding_right_offset=padding_right_offset,
|
|
|
|
max_tokens=max_tokens,
|
|
|
|
)
|
|
|
|
|
|
|
|
@tracer.start_as_current_span("filter")
|
|
|
|
def filter(self, request_ids: List[int]) -> Optional["IdeficsCausalLMBatch"]:
|
|
|
|
# It deletes requests from the batch. For instance when client lost connection
|
|
|
|
if len(request_ids) == 0:
|
|
|
|
raise ValueError("Batch must have at least one request")
|
|
|
|
if len(request_ids) == len(self):
|
|
|
|
return self
|
|
|
|
|
|
|
|
keep_indices = []
|
|
|
|
|
|
|
|
# New values after filtering
|
|
|
|
requests_idx_mapping = {}
|
|
|
|
requests = []
|
|
|
|
input_lengths = []
|
|
|
|
prefix_offsets = []
|
|
|
|
read_offsets = []
|
|
|
|
all_input_ids = []
|
|
|
|
max_input_length = 0
|
|
|
|
|
|
|
|
next_token_choosers = []
|
|
|
|
stopping_criterias = []
|
|
|
|
|
|
|
|
total_remaining_decode_tokens = 0
|
|
|
|
new_padding_right_offset = 0
|
|
|
|
|
|
|
|
for i, request_id in enumerate(request_ids):
|
|
|
|
idx = self.requests_idx_mapping[request_id]
|
|
|
|
requests_idx_mapping[request_id] = i
|
|
|
|
keep_indices.append(idx)
|
|
|
|
|
|
|
|
requests.append(self.requests[idx])
|
|
|
|
prefix_offsets.append(self.prefix_offsets[idx])
|
|
|
|
read_offsets.append(self.read_offsets[idx])
|
|
|
|
all_input_ids.append(self.all_input_ids[idx])
|
|
|
|
|
|
|
|
request_input_length = self.input_lengths[idx]
|
|
|
|
input_lengths.append(request_input_length)
|
|
|
|
max_input_length = max(max_input_length, request_input_length)
|
|
|
|
|
|
|
|
next_token_choosers.append(self.next_token_choosers[idx])
|
|
|
|
stopping_criteria = self.stopping_criterias[idx]
|
|
|
|
stopping_criterias.append(stopping_criteria)
|
|
|
|
remaining_decode_tokens = (
|
|
|
|
stopping_criteria.max_new_tokens - stopping_criteria.current_tokens
|
|
|
|
)
|
|
|
|
total_remaining_decode_tokens += remaining_decode_tokens
|
|
|
|
new_padding_right_offset = max(
|
|
|
|
new_padding_right_offset, remaining_decode_tokens
|
|
|
|
)
|
|
|
|
|
|
|
|
# Apply indices to input_ids, attention mask, past key values and other items that need to be cached
|
|
|
|
input_ids = self.input_ids[keep_indices]
|
|
|
|
position_ids = self.position_ids[keep_indices]
|
|
|
|
self.attention_mask = self.attention_mask[
|
|
|
|
keep_indices,
|
|
|
|
-(self.padding_right_offset + max_input_length) : (
|
|
|
|
self.attention_mask.shape[1] - self.padding_right_offset
|
|
|
|
)
|
|
|
|
+ new_padding_right_offset,
|
|
|
|
]
|
|
|
|
# Do the same for pixel_values and image_attention_mask
|
|
|
|
pixel_values = self.pixel_values[keep_indices]
|
|
|
|
self.image_attention_mask = self.image_attention_mask[
|
|
|
|
keep_indices,
|
|
|
|
-(self.padding_right_offset + max_input_length) : (
|
|
|
|
self.image_attention_mask.shape[1] - self.padding_right_offset
|
|
|
|
)
|
|
|
|
+ new_padding_right_offset,
|
2023-09-27 10:22:09 +00:00
|
|
|
:,
|
2023-08-17 12:38:49 +00:00
|
|
|
]
|
|
|
|
if self.image_hidden_states is None:
|
|
|
|
image_hidden_states = None
|
|
|
|
else:
|
|
|
|
image_hidden_states = self.image_hidden_states[keep_indices]
|
|
|
|
|
|
|
|
# Ensure that past_key_values tensors can be updated in-place
|
2024-07-26 14:29:09 +00:00
|
|
|
if type(self.past_key_values[0]) is tuple:
|
2023-08-17 12:38:49 +00:00
|
|
|
self.past_key_values = [list(layer) for layer in self.past_key_values]
|
|
|
|
|
|
|
|
# Update tensors in-place to allow incremental garbage collection
|
|
|
|
past_kv_length = max_input_length - 1
|
|
|
|
for layer in self.past_key_values:
|
|
|
|
past_keys, past_values = layer
|
|
|
|
if len(past_keys.shape) == 3:
|
|
|
|
# Force past to be of dim [self_size, num_heads, ...] for easy indexing
|
|
|
|
past_keys = past_keys.view(len(self), -1, *past_keys.shape[-2:])
|
|
|
|
past_values = past_values.view(len(self), -1, *past_values.shape[-2:])
|
|
|
|
if self.keys_head_dim_last:
|
|
|
|
layer[0] = past_keys[keep_indices, :, -past_kv_length:, :]
|
|
|
|
else:
|
|
|
|
layer[0] = past_keys[keep_indices, :, :, -past_kv_length:]
|
|
|
|
del past_keys
|
|
|
|
layer[1] = past_values[keep_indices, :, -past_kv_length:, :]
|
|
|
|
del past_values
|
|
|
|
|
|
|
|
max_tokens = len(request_ids) * max_input_length + total_remaining_decode_tokens
|
|
|
|
|
|
|
|
self.requests = requests
|
|
|
|
self.requests_idx_mapping = requests_idx_mapping
|
|
|
|
self.input_ids = input_ids
|
|
|
|
self.pixel_values = pixel_values
|
|
|
|
self.image_hidden_states = image_hidden_states
|
|
|
|
self.position_ids = position_ids
|
|
|
|
self.all_input_ids = all_input_ids
|
|
|
|
self.input_lengths = input_lengths
|
|
|
|
self.prefix_offsets = prefix_offsets
|
|
|
|
self.read_offsets = read_offsets
|
|
|
|
self.next_token_choosers = next_token_choosers
|
|
|
|
self.stopping_criterias = stopping_criterias
|
|
|
|
self.max_input_length = max_input_length
|
|
|
|
self.padding_right_offset = new_padding_right_offset
|
|
|
|
self.max_tokens = max_tokens
|
|
|
|
|
|
|
|
return self
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
@tracer.start_as_current_span("concatenate")
|
2023-09-27 10:22:09 +00:00
|
|
|
def concatenate(
|
|
|
|
cls, batches: List["IdeficsCausalLMBatch"]
|
|
|
|
) -> "IdeficsCausalLMBatch":
|
2023-08-17 12:38:49 +00:00
|
|
|
# It adds new requests to the batch
|
|
|
|
# Used for padding
|
|
|
|
total_batch_size = 0
|
|
|
|
max_input_length = 0
|
|
|
|
max_num_images = 0
|
|
|
|
padding_right_offset = 0
|
|
|
|
for batch in batches:
|
|
|
|
total_batch_size += len(batch)
|
|
|
|
max_input_length = max(max_input_length, batch.max_input_length)
|
|
|
|
max_num_images = max(max_num_images, batch.pixel_values.size(1))
|
|
|
|
padding_right_offset = max(padding_right_offset, batch.padding_right_offset)
|
|
|
|
|
|
|
|
# Batch attributes
|
|
|
|
requests = []
|
|
|
|
requests_idx_mapping = {}
|
|
|
|
input_lengths = []
|
|
|
|
prefix_offsets = []
|
|
|
|
read_offsets = []
|
|
|
|
all_input_ids = []
|
|
|
|
next_token_choosers = []
|
|
|
|
stopping_criterias = []
|
|
|
|
max_tokens = 0
|
|
|
|
|
|
|
|
# Batch tensors
|
|
|
|
input_ids = None
|
|
|
|
attention_mask = None
|
|
|
|
position_ids = None
|
|
|
|
pixel_values = None
|
|
|
|
image_hidden_states = None
|
|
|
|
image_attention_mask = None
|
|
|
|
past_key_values = []
|
|
|
|
|
|
|
|
# Used for slicing correctly inside the tensors
|
|
|
|
# Equivalent to a cumsum on batch sizes
|
|
|
|
start_index = 0
|
|
|
|
for i, batch in enumerate(batches):
|
|
|
|
requests.extend(batch.requests)
|
|
|
|
input_lengths.extend(batch.input_lengths)
|
|
|
|
prefix_offsets.extend(batch.prefix_offsets)
|
|
|
|
read_offsets.extend(batch.read_offsets)
|
|
|
|
all_input_ids.extend(batch.all_input_ids)
|
|
|
|
next_token_choosers.extend(batch.next_token_choosers)
|
|
|
|
stopping_criterias.extend(batch.stopping_criterias)
|
|
|
|
|
|
|
|
if i == 0:
|
|
|
|
requests_idx_mapping = batch.requests_idx_mapping
|
|
|
|
else:
|
|
|
|
# We need to offset the mapping for each batch by the cumulative batch size
|
|
|
|
for k, v in batch.requests_idx_mapping.items():
|
|
|
|
requests_idx_mapping[k] = v + start_index
|
|
|
|
|
|
|
|
# Slicing end index for this batch
|
|
|
|
end_index = start_index + len(batch)
|
|
|
|
|
|
|
|
# We only concatenate batches that did at least one step
|
|
|
|
if batch.past_key_values is None:
|
|
|
|
raise ValueError("only concatenate prefilled batches")
|
|
|
|
|
|
|
|
# Create empty tensor
|
|
|
|
# input_ids is always of shape [batch_size, 1]
|
|
|
|
# We do not need to pad it
|
|
|
|
if input_ids is None:
|
|
|
|
input_ids = batch.input_ids.new_empty((total_batch_size, 1))
|
|
|
|
# Copy to correct indices
|
|
|
|
input_ids[start_index:end_index] = batch.input_ids
|
|
|
|
|
|
|
|
# Create padded tensor
|
|
|
|
if attention_mask is None:
|
|
|
|
attention_mask = batch.attention_mask.new_zeros(
|
|
|
|
(total_batch_size, max_input_length + padding_right_offset),
|
|
|
|
)
|
|
|
|
|
|
|
|
curr_batch_max_num_images = batch.pixel_values.size(1)
|
|
|
|
if pixel_values is None:
|
2023-09-27 10:22:09 +00:00
|
|
|
pixel_values = batch.pixel_values.new_zeros(
|
|
|
|
(total_batch_size, max_num_images, 3, 224, 224)
|
|
|
|
)
|
2024-02-16 10:58:58 +00:00
|
|
|
pixel_values[start_index:end_index, :curr_batch_max_num_images] = (
|
|
|
|
batch.pixel_values
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
if image_attention_mask is None:
|
|
|
|
image_attention_mask = batch.image_attention_mask.new_zeros(
|
2023-09-27 10:22:09 +00:00
|
|
|
(
|
|
|
|
total_batch_size,
|
|
|
|
max_input_length + padding_right_offset,
|
|
|
|
max_num_images,
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
# We need to slice the attention mask to remove padding from previous steps
|
|
|
|
# and to remove unused allocated space
|
|
|
|
left_offset = max_input_length - batch.max_input_length
|
|
|
|
batch_left_offset = (
|
|
|
|
batch.attention_mask.shape[1]
|
|
|
|
- batch.max_input_length
|
|
|
|
- batch.padding_right_offset
|
|
|
|
)
|
|
|
|
attention_mask[
|
|
|
|
start_index:end_index,
|
|
|
|
left_offset:-padding_right_offset,
|
|
|
|
] = batch.attention_mask[
|
|
|
|
:,
|
|
|
|
batch_left_offset : -batch.padding_right_offset,
|
|
|
|
]
|
|
|
|
image_attention_mask[
|
|
|
|
start_index:end_index,
|
|
|
|
left_offset:-padding_right_offset,
|
2023-09-27 10:22:09 +00:00
|
|
|
:curr_batch_max_num_images,
|
2023-08-17 12:38:49 +00:00
|
|
|
] = batch.image_attention_mask[
|
2023-09-27 10:22:09 +00:00
|
|
|
:, batch_left_offset : -batch.padding_right_offset, :
|
2023-08-17 12:38:49 +00:00
|
|
|
]
|
|
|
|
|
|
|
|
# Create empty tensor
|
|
|
|
# position_ids is always of shape [batch_size, 1]
|
|
|
|
if position_ids is None:
|
|
|
|
position_ids = batch.position_ids.new_empty((total_batch_size, 1))
|
|
|
|
position_ids[start_index:end_index] = batch.position_ids
|
|
|
|
|
|
|
|
# Shenanigans to get dimensions because BLOOM outputs a past with a different shape
|
|
|
|
# BLOOM Keys: [batch_size * num_heads, head_dim, seq_length]
|
|
|
|
# BLOOM Values: [batch_size * num_heads, seq_length, head_dim]
|
|
|
|
# And ensure that we can update tensors in-place
|
2024-07-26 14:29:09 +00:00
|
|
|
if isinstance(batch.past_key_values[0], tuple):
|
2023-08-17 12:38:49 +00:00
|
|
|
batch.past_key_values = [
|
|
|
|
[t.view(len(batch), -1, *t.shape[-2:]) for t in layer]
|
|
|
|
for layer in batch.past_key_values
|
|
|
|
]
|
|
|
|
elif len(batch.past_key_values[0][0].shape) == 3:
|
|
|
|
for layer in batch.past_key_values:
|
|
|
|
for k, t in enumerate(layer):
|
|
|
|
layer[k] = t.view(len(batch), -1, *t.shape[-2:])
|
|
|
|
|
|
|
|
# Add eventual padding tokens that were added while concatenating
|
|
|
|
max_tokens += batch.max_tokens + (
|
|
|
|
max_input_length - batch.max_input_length
|
|
|
|
) * len(batch)
|
|
|
|
|
|
|
|
start_index = end_index
|
|
|
|
|
|
|
|
first_past_kvs = batches[0].past_key_values
|
|
|
|
_, num_heads, padded_sequence_length, head_dim = first_past_kvs[0][1].shape
|
|
|
|
|
|
|
|
padded_past_values_shape = (
|
|
|
|
total_batch_size,
|
|
|
|
num_heads,
|
|
|
|
max_input_length - 1,
|
|
|
|
head_dim,
|
|
|
|
)
|
|
|
|
|
|
|
|
if batches[0].keys_head_dim_last:
|
|
|
|
padded_past_keys_shape = padded_past_values_shape
|
|
|
|
else:
|
|
|
|
# seq_length is last for BLOOM
|
|
|
|
padded_past_keys_shape = (
|
|
|
|
total_batch_size,
|
|
|
|
num_heads,
|
|
|
|
head_dim,
|
|
|
|
max_input_length - 1,
|
|
|
|
)
|
|
|
|
|
|
|
|
# Iterate over attention layers
|
|
|
|
# Concatenate past key values layer by layer to allow incremental garbage collection
|
|
|
|
for j in range(len(first_past_kvs)):
|
|
|
|
padded_past_keys = first_past_kvs[j][0].new_zeros(padded_past_keys_shape)
|
|
|
|
start_index = 0
|
|
|
|
for batch in batches:
|
|
|
|
past_keys = batch.past_key_values[j][0]
|
|
|
|
# Clear reference to the original tensor
|
|
|
|
batch.past_key_values[j][0] = None
|
|
|
|
|
|
|
|
# Slicing end index for this batch
|
|
|
|
end_index = start_index + len(batch)
|
|
|
|
# We slice the keys to remove the padding from previous batches
|
|
|
|
past_seq_len = batch.max_input_length - 1
|
|
|
|
if batch.keys_head_dim_last:
|
2024-02-16 10:58:58 +00:00
|
|
|
padded_past_keys[start_index:end_index, :, -past_seq_len:, :] = (
|
|
|
|
past_keys[:, :, -past_seq_len:, :]
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
else:
|
|
|
|
# BLOOM case
|
2024-02-16 10:58:58 +00:00
|
|
|
padded_past_keys[start_index:end_index, :, :, -past_seq_len:] = (
|
|
|
|
past_keys[:, :, :, -past_seq_len:]
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
del past_keys
|
|
|
|
|
|
|
|
start_index = end_index
|
|
|
|
|
|
|
|
padded_past_values = first_past_kvs[j][1].new_zeros(
|
|
|
|
padded_past_values_shape
|
|
|
|
)
|
|
|
|
start_index = 0
|
|
|
|
for batch in batches:
|
|
|
|
past_values = batch.past_key_values[j][1]
|
|
|
|
# Clear reference to the original tensor
|
|
|
|
batch.past_key_values[j][1] = None
|
|
|
|
|
|
|
|
# Slicing end index for this batch
|
|
|
|
end_index = start_index + len(batch)
|
|
|
|
# We slice the past values to remove the padding from previous batches
|
|
|
|
past_seq_len = batch.max_input_length - 1
|
2024-02-16 10:58:58 +00:00
|
|
|
padded_past_values[start_index:end_index, :, -past_seq_len:, :] = (
|
|
|
|
past_values[:, :, -past_seq_len:, :]
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
del past_values
|
|
|
|
|
|
|
|
# Update values
|
|
|
|
start_index = end_index
|
|
|
|
|
|
|
|
past_key_values.append([padded_past_keys, padded_past_values])
|
|
|
|
|
|
|
|
return cls(
|
|
|
|
batch_id=batches[0].batch_id,
|
|
|
|
requests=requests,
|
|
|
|
requests_idx_mapping=requests_idx_mapping,
|
|
|
|
input_ids=input_ids,
|
|
|
|
attention_mask=attention_mask,
|
|
|
|
position_ids=position_ids,
|
|
|
|
pixel_values=pixel_values,
|
|
|
|
image_hidden_states=image_hidden_states,
|
|
|
|
image_attention_mask=image_attention_mask,
|
|
|
|
past_key_values=past_key_values,
|
|
|
|
all_input_ids=all_input_ids,
|
|
|
|
input_lengths=input_lengths,
|
|
|
|
prefix_offsets=prefix_offsets,
|
|
|
|
read_offsets=read_offsets,
|
|
|
|
next_token_choosers=next_token_choosers,
|
|
|
|
stopping_criterias=stopping_criterias,
|
|
|
|
max_input_length=max_input_length,
|
|
|
|
padding_right_offset=padding_right_offset,
|
|
|
|
keys_head_dim_last=batches[0].keys_head_dim_last,
|
|
|
|
max_tokens=max_tokens,
|
|
|
|
)
|
|
|
|
|
|
|
|
def __len__(self):
|
|
|
|
return len(self.requests)
|
|
|
|
|
|
|
|
|
|
|
|
class IdeficsCausalLM(Model):
|
|
|
|
def __init__(
|
|
|
|
self,
|
|
|
|
model_id: str,
|
|
|
|
revision: Optional[str] = None,
|
|
|
|
quantize: Optional[str] = None,
|
2024-10-02 09:22:13 +00:00
|
|
|
speculator: Optional[str] = None,
|
2023-08-17 12:38:49 +00:00
|
|
|
dtype: Optional[torch.dtype] = None,
|
|
|
|
trust_remote_code: bool = False,
|
|
|
|
):
|
2024-08-14 09:58:08 +00:00
|
|
|
self.quantize = quantize
|
2024-10-02 09:22:13 +00:00
|
|
|
self.process_group, rank, world_size = initialize_torch_distributed()
|
2023-08-17 12:38:49 +00:00
|
|
|
if torch.cuda.is_available():
|
2024-10-02 09:22:13 +00:00
|
|
|
device = torch.device(f"cuda:{rank}")
|
|
|
|
# 9b seems to work correctly enough in float16, but 80b seems
|
|
|
|
# to be really saturating for f16.
|
|
|
|
dtype = torch.float16 if dtype is None else dtype
|
|
|
|
elif SYSTEM == "ipex":
|
|
|
|
if hasattr(torch, "xpu") and torch.xpu.is_available():
|
|
|
|
device = torch.device(f"xpu:{rank}")
|
|
|
|
dtype = torch.float16 if dtype is None else dtype
|
|
|
|
else:
|
|
|
|
device = torch.device("cpu")
|
|
|
|
# Float16 doesn't exist on target.
|
|
|
|
dtype = torch.bfloat16 if dtype is None else dtype
|
2023-08-17 12:38:49 +00:00
|
|
|
else:
|
|
|
|
device = torch.device("cpu")
|
2023-09-19 15:19:28 +00:00
|
|
|
dtype = torch.float32 if dtype is None else dtype
|
2024-10-02 09:22:13 +00:00
|
|
|
self.device, self.dtype = device, dtype
|
|
|
|
|
|
|
|
config = AutoConfig.from_pretrained(
|
|
|
|
model_id,
|
|
|
|
revision=revision,
|
|
|
|
trust_remote_code=trust_remote_code,
|
|
|
|
)
|
|
|
|
config.quantize = quantize
|
|
|
|
config.speculator = speculator
|
|
|
|
config.vision_config.quantize = quantize
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(
|
|
|
|
model_id,
|
|
|
|
revision=revision,
|
|
|
|
padding_side="left",
|
|
|
|
truncation_side="left",
|
|
|
|
trust_remote_code=trust_remote_code,
|
|
|
|
)
|
|
|
|
self.processor = AutoProcessor.from_pretrained(
|
|
|
|
model_id,
|
|
|
|
revision=revision,
|
|
|
|
padding_side="left",
|
|
|
|
truncation_side="left",
|
|
|
|
trust_remote_code=trust_remote_code,
|
|
|
|
)
|
2024-10-02 09:22:13 +00:00
|
|
|
|
|
|
|
weights_loader = get_loader(
|
|
|
|
quantize=quantize, model_id=model_id, revision=revision
|
2023-08-17 12:38:49 +00:00
|
|
|
)
|
2024-10-02 09:22:13 +00:00
|
|
|
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,
|
|
|
|
weights_loader=weights_loader,
|
|
|
|
)
|
|
|
|
|
|
|
|
model = IdeficsForVisionText2Text(config, weights)
|
|
|
|
|
|
|
|
self.config = config
|
2023-08-17 12:38:49 +00:00
|
|
|
|
2024-10-02 09:22:13 +00:00
|
|
|
torch.distributed.barrier(group=self.process_group)
|
|
|
|
super().__init__(
|
2024-06-25 18:46:27 +00:00
|
|
|
model_id=model_id,
|
2023-08-17 12:38:49 +00:00
|
|
|
model=model,
|
|
|
|
tokenizer=tokenizer,
|
|
|
|
requires_padding=True,
|
|
|
|
dtype=dtype,
|
|
|
|
device=device,
|
2024-10-02 09:22:13 +00:00
|
|
|
rank=rank,
|
|
|
|
world_size=world_size,
|
2023-08-17 12:38:49 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
@property
|
|
|
|
def batch_type(self) -> Type[IdeficsCausalLMBatch]:
|
|
|
|
return IdeficsCausalLMBatch
|
|
|
|
|
|
|
|
def forward(
|
|
|
|
self,
|
|
|
|
input_ids,
|
|
|
|
attention_mask,
|
|
|
|
position_ids,
|
|
|
|
pixel_values,
|
|
|
|
image_hidden_states,
|
|
|
|
image_attention_mask,
|
|
|
|
past_key_values: Optional = None,
|
|
|
|
) -> Tuple[torch.Tensor, List[Tuple[torch.Tensor, torch.Tensor]]]:
|
|
|
|
# Model Forward
|
|
|
|
kwargs = {
|
|
|
|
"input_ids": input_ids,
|
|
|
|
"attention_mask": attention_mask,
|
|
|
|
"pixel_values": pixel_values,
|
|
|
|
"image_hidden_states": image_hidden_states,
|
|
|
|
"image_attention_mask": image_attention_mask,
|
|
|
|
"past_key_values": past_key_values,
|
|
|
|
"use_cache": True,
|
|
|
|
"return_dict": True,
|
|
|
|
}
|
|
|
|
if self.has_position_ids:
|
|
|
|
kwargs["position_ids"] = position_ids
|
|
|
|
|
2024-02-26 18:49:28 +00:00
|
|
|
outputs, speculative_logits = self.model.forward(**kwargs)
|
|
|
|
return (
|
|
|
|
outputs.logits,
|
|
|
|
speculative_logits,
|
|
|
|
outputs.past_key_values,
|
|
|
|
outputs.image_hidden_states,
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
@tracer.start_as_current_span("generate_token")
|
|
|
|
def generate_token(
|
|
|
|
self, batch: IdeficsCausalLMBatch
|
2023-12-14 14:59:38 +00:00
|
|
|
) -> Tuple[List[Generation], Optional[IdeficsCausalLMBatch], Tuple[int, int]]:
|
|
|
|
start = time.time_ns()
|
2023-08-17 12:38:49 +00:00
|
|
|
# slice the attention mask to the correct shape
|
|
|
|
attention_mask = batch.attention_mask[:, : -batch.padding_right_offset]
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
if batch.image_attention_mask is None:
|
|
|
|
image_attention_mask = None
|
2023-08-17 12:38:49 +00:00
|
|
|
else:
|
Adding Llava-Next (Llava 1.6) with full support. (#1709)
# What does this PR do?
- Changed all models to extract `embed_tokens` in order to enable llava
to separately call the embeddings and the core model layers.
- Added VlmCausalLM to inherit from FlashMistral in order to be
maximally supported. The only added logics sits on top and parses images
into pixel values, preallocates input_ids space for the image
embeddings, and passes them for the model.
- Added Clip for the vision tower.
- Didn't add flash for the vision tower since there's no padding anyway.
- Added heuristic (potentially incomplete) to calculate number of
features *before* calculating the clip patches (allows for easier logic
reuse of the LLM under the hood).
Still needs to be done:
- [x] Implement the image parsing in the controller side, to avoid
downloading n times per TP shard and also refusing requests too large
early and avoid issues where the truncation actually truncates the
image.
- [ ] Make sure it works with quantization properly.
- [x] Make sure it works with TP>1
<!--
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
-->
2024-04-09 19:32:00 +00:00
|
|
|
if batch.input_ids.size(1) == 1:
|
|
|
|
# THIS is a hack: when calling idefics.generate, the first time, we need the whole image_attention_mask (size bs x max_seq_len x max_num_images),
|
|
|
|
# but the subsequent times, we only need the last attention mask along the `max_seq_len` dimension
|
|
|
|
# this is due to the nature IDEFICS: it's an encoder decoder, and so when decoding, only the currently generated
|
|
|
|
# token need to attend to the encoder hidden states (i.e. the vision encoder)
|
|
|
|
# Also see seq2seq_lm.Seq2SeqLM.generate_token which has roughly the same logic
|
|
|
|
image_attention_mask = batch.image_attention_mask[
|
|
|
|
:, -(batch.padding_right_offset + 1)
|
|
|
|
].unsqueeze(1)
|
|
|
|
else:
|
|
|
|
image_attention_mask = batch.image_attention_mask[
|
|
|
|
:, : -batch.padding_right_offset
|
|
|
|
]
|
2023-08-17 12:38:49 +00:00
|
|
|
|
2024-02-26 18:49:28 +00:00
|
|
|
logits, speculative_logits, past, image_hidden_states = self.forward(
|
2023-08-17 12:38:49 +00:00
|
|
|
input_ids=batch.input_ids,
|
|
|
|
attention_mask=attention_mask,
|
|
|
|
position_ids=batch.position_ids,
|
|
|
|
pixel_values=batch.pixel_values,
|
|
|
|
image_hidden_states=batch.image_hidden_states,
|
|
|
|
image_attention_mask=image_attention_mask,
|
|
|
|
past_key_values=batch.past_key_values,
|
|
|
|
)
|
|
|
|
# Hardcoded remove image tokens
|
|
|
|
logits[:, 32000:32001] = torch.finfo(logits.dtype).min
|
|
|
|
|
2023-12-14 14:59:38 +00:00
|
|
|
start_decode = time.time_ns()
|
|
|
|
|
2023-08-17 12:38:49 +00:00
|
|
|
# Results
|
|
|
|
generations: List[Generation] = []
|
|
|
|
stopped = True
|
|
|
|
|
|
|
|
# Zipped iterator
|
|
|
|
iterator = zip(
|
|
|
|
batch.requests,
|
|
|
|
batch.input_lengths,
|
|
|
|
batch.prefix_offsets,
|
|
|
|
batch.read_offsets,
|
|
|
|
logits,
|
|
|
|
batch.next_token_choosers,
|
|
|
|
batch.stopping_criterias,
|
|
|
|
batch.all_input_ids,
|
|
|
|
)
|
|
|
|
|
|
|
|
# For each member of the batch
|
|
|
|
for i, (
|
|
|
|
request,
|
|
|
|
input_length,
|
|
|
|
prefix_offset,
|
|
|
|
read_offset,
|
|
|
|
logits,
|
|
|
|
next_token_chooser,
|
|
|
|
stopping_criteria,
|
|
|
|
all_input_ids,
|
|
|
|
) in enumerate(iterator):
|
|
|
|
# Select next token
|
|
|
|
next_token_id, logprobs = next_token_chooser(
|
|
|
|
all_input_ids.view(1, -1), logits[-1:, :]
|
|
|
|
)
|
|
|
|
|
|
|
|
# Append next token to all tokens
|
|
|
|
all_input_ids = torch.cat([all_input_ids, next_token_id])
|
|
|
|
new_input_length = input_length + 1
|
|
|
|
|
|
|
|
# Generated token
|
|
|
|
next_token_logprob = logprobs[-1, next_token_id]
|
|
|
|
next_token_id_squeezed = next_token_id.squeeze()
|
|
|
|
next_token_text, prefix_offset, read_offset = self.decode_token(
|
|
|
|
all_input_ids[:, 0], prefix_offset, read_offset
|
|
|
|
)
|
|
|
|
|
|
|
|
# Evaluate stopping criteria
|
|
|
|
stop, reason = stopping_criteria(
|
|
|
|
next_token_id_squeezed,
|
|
|
|
next_token_text,
|
|
|
|
)
|
|
|
|
|
|
|
|
if not stop:
|
|
|
|
stopped = False
|
|
|
|
|
|
|
|
# Shard generations
|
|
|
|
# All generations will be appended in the rust sharded client
|
|
|
|
if i % self.world_size == self.rank:
|
|
|
|
if stop:
|
|
|
|
# Decode generated tokens
|
2023-09-27 10:13:45 +00:00
|
|
|
output_text, _, _ = self.decode_token(
|
|
|
|
all_input_ids[:, 0],
|
2023-09-27 10:22:09 +00:00
|
|
|
prefix_offset=len(all_input_ids)
|
|
|
|
- stopping_criteria.current_tokens
|
|
|
|
- 1,
|
|
|
|
read_offset=len(all_input_ids)
|
|
|
|
- stopping_criteria.current_tokens,
|
|
|
|
skip_special_tokens=True,
|
2023-08-17 12:38:49 +00:00
|
|
|
)
|
|
|
|
# Get seed
|
|
|
|
if isinstance(next_token_chooser.choice, Sampling):
|
|
|
|
seed = next_token_chooser.choice.seed
|
|
|
|
else:
|
|
|
|
seed = None
|
|
|
|
|
|
|
|
generated_text = GeneratedText(
|
|
|
|
output_text, stopping_criteria.current_tokens, reason, seed
|
|
|
|
)
|
|
|
|
else:
|
|
|
|
generated_text = None
|
|
|
|
|
|
|
|
# Prefill
|
|
|
|
if stopping_criteria.current_tokens == 1 and request.prefill_logprobs:
|
|
|
|
# Remove generated token to only have prefill and add nan for first prompt token
|
|
|
|
prefill_logprobs = [float("nan")] + torch.log_softmax(
|
|
|
|
logits, -1
|
|
|
|
).gather(1, all_input_ids[1:]).squeeze(1)[
|
|
|
|
-new_input_length:-1
|
|
|
|
].tolist()
|
|
|
|
prefill_token_ids = all_input_ids[-new_input_length:-1]
|
|
|
|
prefill_texts = self.tokenizer.batch_decode(
|
|
|
|
prefill_token_ids,
|
|
|
|
clean_up_tokenization_spaces=False,
|
|
|
|
skip_special_tokens=False,
|
|
|
|
)
|
2023-12-11 11:46:30 +00:00
|
|
|
prefill_tokens = Tokens(
|
2023-12-11 13:49:52 +00:00
|
|
|
prefill_token_ids,
|
|
|
|
prefill_logprobs,
|
|
|
|
prefill_texts,
|
|
|
|
is_special=[],
|
2023-08-17 12:38:49 +00:00
|
|
|
)
|
|
|
|
else:
|
|
|
|
prefill_tokens = None
|
|
|
|
|
2023-09-27 10:22:09 +00:00
|
|
|
top_tokens = None
|
2023-08-28 09:43:47 +00:00
|
|
|
|
2023-08-17 12:38:49 +00:00
|
|
|
generation = Generation(
|
|
|
|
request.id,
|
|
|
|
prefill_tokens,
|
2023-12-11 11:46:30 +00:00
|
|
|
Tokens(
|
2023-12-11 13:49:52 +00:00
|
|
|
[next_token_id_squeezed],
|
|
|
|
[next_token_logprob],
|
|
|
|
[next_token_text],
|
|
|
|
[next_token_id_squeezed.item() in self.all_special_ids],
|
2023-12-11 11:46:30 +00:00
|
|
|
),
|
2023-08-17 12:38:49 +00:00
|
|
|
generated_text,
|
2023-09-27 10:22:09 +00:00
|
|
|
top_tokens,
|
2023-08-17 12:38:49 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
generations.append(generation)
|
|
|
|
|
|
|
|
# Update values
|
2024-02-15 09:28:10 +00:00
|
|
|
batch.next_token_choosers[i] = batch.next_token_choosers[i].advance_grammar(
|
|
|
|
next_token_id_squeezed.item()
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
batch.input_ids[i, 0] = next_token_id
|
|
|
|
batch.all_input_ids[i] = all_input_ids
|
|
|
|
batch.input_lengths[i] = new_input_length
|
|
|
|
batch.prefix_offsets[i] = prefix_offset
|
|
|
|
batch.read_offsets[i] = read_offset
|
|
|
|
batch.max_input_length = max(batch.max_input_length, new_input_length)
|
|
|
|
|
|
|
|
# We finished all generations in the batch; there is no next batch
|
|
|
|
if stopped:
|
2023-12-14 14:59:38 +00:00
|
|
|
forward_ns = start_decode - start
|
|
|
|
decode_ns = time.time_ns() - start_decode
|
|
|
|
return generations, None, (forward_ns, decode_ns)
|
2023-08-17 12:38:49 +00:00
|
|
|
|
|
|
|
# Slice unused values from prefill
|
|
|
|
batch.input_ids = batch.input_ids[:, :1]
|
|
|
|
|
|
|
|
# Update attention_mask as we added a new token to input_ids
|
|
|
|
batch.attention_mask[:, -batch.padding_right_offset] = 1
|
2024-02-16 10:58:58 +00:00
|
|
|
batch.image_attention_mask[:, -batch.padding_right_offset, :] = (
|
|
|
|
batch.image_attention_mask[:, -(batch.padding_right_offset + 1), :]
|
|
|
|
)
|
2023-08-17 12:38:49 +00:00
|
|
|
# Decrease right offset
|
|
|
|
batch.padding_right_offset -= 1
|
|
|
|
|
|
|
|
# Update position_ids
|
|
|
|
batch.position_ids = batch.position_ids[:, -1:] + 1
|
|
|
|
|
|
|
|
# Update past key values
|
|
|
|
batch.past_key_values = past
|
|
|
|
batch.image_hidden_states = image_hidden_states
|
|
|
|
|
2023-12-14 14:59:38 +00:00
|
|
|
forward_ns = start_decode - start
|
|
|
|
decode_ns = time.time_ns() - start_decode
|
|
|
|
return generations, batch, (forward_ns, decode_ns)
|