Mirrors
/
hf_text-generation-inference
mirror of https://github.com/huggingface/text-generation-inference.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fix(models): Revert buggy support for AutoModel

This commit is contained in:
OlivierDehaene 2022-11-03 16:07:54 +01:00
parent b3b7ea0d74
commit 755fc0e403
7 changed files with 339 additions and 315 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
README.md
View File

@ -15,13 +15,11 @@ A Rust and gRPC server for text generation inference.
- [Safetensors](https://github.com/huggingface/safetensors) weight loading
- 45ms per token generation for BLOOM with 8xA100 80GB
## Officially supported models
## Supported models
- BLOOM
- BLOOM-560m
Other models are supported on a best-effort basis using `AutoModelForCausalLM.from_pretrained(<model>, torch_dtype=torch.float16, device_map="auto")`.
## Load Tests for BLOOM
See `k6/load_test.js`
@ -82,6 +80,7 @@ make router-dev
## TODO:
- [ ] Support AutoModelForSeq2SeqLM
- [ ] Add tests for the `server/model` logic
- [ ] Backport custom CUDA kernels to Transformers
- [ ] Install safetensors with pip

10
server/Makefile
View File

@ -9,11 +9,11 @@ gen-server:
install-transformers:
# Install specific version of transformers
rm transformers || true
rm transformers-46d37bece7d3ffdef97b1ee4a3170c0a0627d921 || true
curl -L -O https://github.com/huggingface/transformers/archive/46d37bece7d3ffdef97b1ee4a3170c0a0627d921.zip
unzip 46d37bece7d3ffdef97b1ee4a3170c0a0627d921.zip
rm 46d37bece7d3ffdef97b1ee4a3170c0a0627d921.zip
mv transformers-46d37bece7d3ffdef97b1ee4a3170c0a0627d921 transformers
rm transformers-7302a24535e8dc5637ea5b4e4572fc971d404098 || true
curl -L -O https://github.com/OlivierDehaene/transformers/archive/7302a24535e8dc5637ea5b4e4572fc971d404098.zip
unzip 7302a24535e8dc5637ea5b4e4572fc971d404098.zip
rm 7302a24535e8dc5637ea5b4e4572fc971d404098.zip
mv transformers-7302a24535e8dc5637ea5b4e4572fc971d404098 transformers
cd transformers && python setup.py install
install-safetensors:

14
server/text_generation/models/__init__.py
View File

@ -1,22 +1,16 @@
from text_generation.models.model import Model
from text_generation.models.bloom import BLOOMSharded
from text_generation.models.bloom import BLOOM, BLOOMSharded
__all__ = ["Model", "BLOOMSharded"]
__all__ = ["Model", "BLOOM", "BLOOMSharded"]
def get_model(model_name: str, sharded: bool, quantize: bool) -> Model:
if model_name.startswith("bigscience/bloom"):
if sharded:
return BLOOMSharded(model_name, quantize)
else:
if quantize:
raise ValueError("quantization is not supported for non-sharded BLOOM")
return Model(model_name)
return BLOOM(model_name)
else:
if sharded:
raise ValueError("sharded is only supported for BLOOM models")
if quantize:
raise ValueError("Quantization is only supported for BLOOM models")
return Model(model_name)
raise ValueError(f"model {model_name} is not supported yet")

303
server/text_generation/models/bloom.py
View File

@ -1,7 +1,7 @@
import torch
import torch.distributed
from typing import List, Optional
from typing import List, Optional, Tuple, Type
from accelerate import init_empty_weights
from safetensors import safe_open
@ -11,8 +11,10 @@ from transformers.models.bloom.parallel_layers import (
TensorParallelEmbedding,
TensorParallelRowLinear,
)
from transformers.modeling_outputs import CausalLMOutputWithPast
from text_generation.models import Model
from text_generation.models.types import Batch, GeneratedText
from text_generation.utils import (
initialize_torch_distributed,
weight_files,
@ -29,9 +31,306 @@ except Exception as e:
torch.manual_seed(0)
class BLOOMSharded(Model):
class BloomBatch(Batch):
@classmethod
def concatenate(cls, batches: List["Batch"]) -> "BloomBatch":
# Used for padding
total_batch_size = sum(batch.size for batch in batches)
max_sequence_length = max(batch.max_sequence_length for batch in batches)
# Batch attributes
input_ids = {"input_ids": None, "attention_mask": None, "past_key_values": []}
requests = []
all_input_lengths = []
all_input_ids = []
next_token_choosers = []
stopping_criterias = []
# 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)
all_input_lengths.extend(batch.all_input_lengths)
all_input_ids.extend(batch.all_input_ids)
next_token_choosers.extend(batch.next_token_choosers)
stopping_criterias.extend(batch.stopping_criterias)
# Slicing end index for this batch
end_index = start_index + batch.size
# We only concatenate batches that did at least one step
if batch.input_ids["input_ids"].shape[1] > 1:
raise ValueError("Batch input_ids should be of shape (batch_size, 1)")
# Initialize tensors
if i == 0:
input_ids["input_ids"] = torch.empty(
(total_batch_size, 1),
dtype=batch.input_ids["input_ids"].dtype,
device=batch.input_ids["input_ids"].device,
)
input_ids["attention_mask"] = torch.zeros(
(total_batch_size, max_sequence_length),
dtype=batch.input_ids["attention_mask"].dtype,
device=batch.input_ids["attention_mask"].device,
)
# input_ids["input_ids"] is always of shape [batch_size, 1]
# We do not need to pad it
input_ids["input_ids"][start_index:end_index] = batch.input_ids["input_ids"]
# We need to slice the attention mask to remove padding from previous steps
input_ids["attention_mask"][
start_index:end_index, -batch.max_sequence_length:
] = batch.input_ids["attention_mask"][:, -batch.max_sequence_length:]
for j, past in enumerate(batch.input_ids["past_key_values"]):
past_keys = past[0]
past_values = past[1]
_, head_dim, padded_sequence_length = past_keys.shape
# Reshape the tensors to make slicing easier
past_keys = past_keys.view(
batch.size, -1, head_dim, padded_sequence_length
)
past_values = past_values.view(
batch.size, -1, padded_sequence_length, head_dim
)
num_heads = past_keys.shape[1]
# Initialize tensors
# This will run only once per layer
if j == len(input_ids["past_key_values"]):
padded_past_keys = torch.zeros(
(
total_batch_size,
num_heads,
head_dim,
max_sequence_length - 1,
),
dtype=past_keys.dtype,
device=past_keys.device,
)
padded_past_values = torch.zeros(
(
total_batch_size,
num_heads,
max_sequence_length - 1,
head_dim,
),
dtype=past_values.dtype,
device=past_values.device,
)
input_ids["past_key_values"].append(
[padded_past_keys, padded_past_values]
)
# We slice the past keys and values to remove the padding from previous batches
input_ids["past_key_values"][j][0][
start_index:end_index, :, :, -(batch.max_sequence_length - 1):
] = past_keys[:, :, :, -(batch.max_sequence_length - 1):]
input_ids["past_key_values"][j][1][
start_index:end_index, :, -(batch.max_sequence_length - 1):, :
] = past_values[:, :, -(batch.max_sequence_length - 1):, :]
# If we are on the last batch, we need to reshape the tensors
if (i + 1) == len(batches):
input_ids["past_key_values"][j][0] = input_ids["past_key_values"][
j
][0].view(total_batch_size * num_heads, head_dim, -1)
input_ids["past_key_values"][j][1] = input_ids["past_key_values"][
j
][1].view(total_batch_size * num_heads, -1, head_dim)
start_index += batch.size
return cls(
batch_id=batches[0].batch_id,
requests=requests,
all_input_lengths=all_input_lengths,
input_ids=input_ids,
all_input_ids=all_input_ids,
next_token_choosers=next_token_choosers,
stopping_criterias=stopping_criterias,
size=total_batch_size,
max_sequence_length=max_sequence_length,
)
class BLOOM(Model):
def __init__(self, model_name: str):
if not model_name.startswith("bigscience/bloom"):
raise ValueError(f"Model {model_name} is not supported")
if torch.cuda.is_available():
self.device = torch.device("cuda")
dtype = torch.bfloat16 if torch.cuda.is_bf16_supported() else torch.float16
else:
self.device = torch.device("cpu")
dtype = torch.float32
self.tokenizer = AutoTokenizer.from_pretrained(model_name, padding_side="left")
self.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
self.model = AutoModelForCausalLM.from_pretrained(
model_name, torch_dtype=dtype, device_map="auto" if torch.cuda.is_available() else None
).eval()
self.num_heads = self.model.config.num_attention_heads
@property
def batch_type(self) -> Type[BloomBatch]:
return BloomBatch
def forward(
self, input_ids, attention_mask, past_key_values: Optional = None
) -> CausalLMOutputWithPast:
# Model Forward
return self.model.forward(
input_ids=input_ids,
attention_mask=attention_mask,
past_key_values=past_key_values,
use_cache=True,
)
def generate_token(
self, batch: BloomBatch
) -> Tuple[List[GeneratedText], Optional[BloomBatch]]:
# For some reason, inference_mode does not work well with GLOO which we use on CPU
context_manager = (
torch.no_grad if self.device.type == "cpu" else torch.inference_mode
)
with context_manager():
outputs = self.forward(**batch.input_ids)
# List of indices to cache
next_batch_keep_indices = []
next_batch_past_keep_indices = []
# New input_ids for next forward
next_batch_input_ids = []
next_batch_all_input_ids = []
next_all_input_lengths = []
next_batch_size = 0
next_batch_max_sequence_length = 0
# Finished requests
generated_texts: List[GeneratedText] = []
# Zipped iterator
iterator = zip(
batch.requests,
batch.all_input_lengths,
outputs.logits,
batch.next_token_choosers,
batch.stopping_criterias,
batch.all_input_ids,
)
# For each member of the batch
for i, (
request,
input_length,
logits,
next_token_chooser,
stopping_criteria,
all_tokens,
) in enumerate(iterator):
# Select next token
next_token = next_token_chooser(all_tokens, logits.unsqueeze(0)[:, -1])
# Append next token to all tokens
all_tokens = torch.cat([all_tokens, next_token])
# Evaluate stopping criteria
if stopping_criteria(all_tokens):
# Decode all tokens
output = self.tokenizer.decode(
all_tokens.squeeze(-1), skip_special_tokens=True
)
# Add to the list of finished generations with the original request
generated_texts.append(GeneratedText(request, output))
# add to the next batch
else:
next_batch_keep_indices.append(i)
# past_key_values is of shape [batch_size * num_heads, ...]
# so we need to take into account the `num_heads` stride here
next_batch_past_keep_indices.extend(
[j for j in range(i * self.num_heads, (i + 1) * self.num_heads)]
)
next_batch_input_ids.append(next_token)
next_batch_all_input_ids.append(all_tokens)
next_batch_size += 1
new_input_length = input_length + 1
next_all_input_lengths.append(new_input_length)
next_batch_max_sequence_length = max(
next_batch_max_sequence_length, new_input_length
)
# We finished all generations in the batch; there is no next batch
if not next_batch_keep_indices:
return generated_texts, None
# If we finished at least one generation
next_batch_input_ids = {"input_ids": torch.cat(next_batch_input_ids, dim=0)}
if generated_texts:
# Apply indices to attention mask, past key values and other items that need to be cached
next_batch_input_ids["attention_mask"] = batch.input_ids["attention_mask"][
next_batch_keep_indices
]
next_batch_input_ids["past_key_values"] = [
(
keys[next_batch_past_keep_indices],
values[next_batch_past_keep_indices],
)
for keys, values in outputs["past_key_values"]
]
next_batch_requests = [batch.requests[i] for i in next_batch_keep_indices]
next_batch_next_token_choosers = [
batch.next_token_choosers[i] for i in next_batch_keep_indices
]
next_batch_stopping_criterias = [
batch.stopping_criterias[i] for i in next_batch_keep_indices
]
else:
next_batch_input_ids["attention_mask"] = batch.input_ids["attention_mask"]
next_batch_input_ids["past_key_values"] = outputs["past_key_values"]
next_batch_requests = batch.requests
next_batch_next_token_choosers = batch.next_token_choosers
next_batch_stopping_criterias = batch.stopping_criterias
# Update attention_mask with padding as we added a new token to input_ids
next_batch_input_ids["attention_mask"] = torch.cat(
[
next_batch_input_ids["attention_mask"],
torch.ones((next_batch_size, 1)).to(self.device),
],
dim=1,
)
next_batch = BloomBatch(
batch_id=batch.batch_id,
requests=next_batch_requests,
all_input_lengths=next_all_input_lengths,
input_ids=next_batch_input_ids,
all_input_ids=next_batch_all_input_ids,
next_token_choosers=next_batch_next_token_choosers,
stopping_criterias=next_batch_stopping_criterias,
size=next_batch_size,
max_sequence_length=next_batch_max_sequence_length,
)
return generated_texts, next_batch
class BLOOMSharded(BLOOM):
def __init__(self, model_name: str, quantize: bool = False):
super(Model, self).__init__()
if not model_name.startswith("bigscience/bloom"):
raise ValueError(f"Model {model_name} is not supported")
self.process_group, self.rank, self.world_size = initialize_torch_distributed()
self.master = self.rank == 0
if torch.cuda.is_available():

171
server/text_generation/models/model.py
View File

@ -1,166 +1,19 @@
import torch
import torch.distributed
from typing import List, Tuple, Optional
from transformers import AutoTokenizer, AutoModelForCausalLM, AutoConfig
from transformers.modeling_outputs import CausalLMOutputWithPast
from abc import ABC, abstractmethod
from typing import List, Tuple, Optional, TypeVar, Type
from text_generation.models.types import Batch, GeneratedText
B = TypeVar("B", bound=Batch)
class Model:
def __init__(self, model_name: str):
if torch.cuda.is_available():
self.device = torch.device("cuda")
dtype = torch.float16
else:
self.device = torch.device("cpu")
dtype = torch.float32
self.tokenizer = AutoTokenizer.from_pretrained(model_name, padding_side="left")
self.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
self.model = AutoModelForCausalLM.from_pretrained(
model_name, torch_dtype=dtype, device_map="auto" if torch.cuda.is_available() else None
).eval()
self.num_heads = self.model.config.num_attention_heads
def forward(
self, input_ids, attention_mask, past_key_values: Optional = None
) -> CausalLMOutputWithPast:
# Model Forward
return self.model.forward(
input_ids=input_ids,
attention_mask=attention_mask,
past_key_values=past_key_values,
use_cache=True,
)
class Model(ABC):
@property
@abstractmethod
def batch_type(self) -> Type[B]:
raise NotImplementedError
@abstractmethod
def generate_token(
self, batch: Batch
) -> Tuple[List[GeneratedText], Optional[Batch]]:
# For some reason, inference_mode does not work well with GLOO which we use on CPU
context_manager = (
torch.no_grad if self.device.type == "cpu" else torch.inference_mode
)
with context_manager():
outputs = self.forward(**batch.input_ids)
# List of indices to cache
next_batch_keep_indices = []
next_batch_past_keep_indices = []
# New input_ids for next forward
next_batch_input_ids = []
next_batch_all_input_ids = []
next_all_input_lengths = []
next_batch_size = 0
next_batch_max_sequence_length = 0
# Finished requests
generated_texts: List[GeneratedText] = []
# Zipped iterator
iterator = zip(
batch.requests,
batch.all_input_lengths,
outputs.logits,
batch.next_token_choosers,
batch.stopping_criterias,
batch.all_input_ids,
)
# For each member of the batch
for i, (
request,
input_length,
logits,
next_token_chooser,
stopping_criteria,
all_tokens,
) in enumerate(iterator):
# Select next token
next_token = next_token_chooser(all_tokens, logits.unsqueeze(0)[:, -1])
# Append next token to all tokens
all_tokens = torch.cat([all_tokens, next_token])
# Evaluate stopping criteria
if stopping_criteria(all_tokens):
# Decode all tokens
output = self.tokenizer.decode(
all_tokens.squeeze(-1), skip_special_tokens=True
)
# Add to the list of finished generations with the original request
generated_texts.append(GeneratedText(request, output))
# add to the next batch
else:
next_batch_keep_indices.append(i)
# past_key_values is of shape [batch_size * num_heads, ...]
# so we need to take into account the `num_heads` stride here
next_batch_past_keep_indices.extend(
[j for j in range(i * self.num_heads, (i + 1) * self.num_heads)]
)
next_batch_input_ids.append(next_token)
next_batch_all_input_ids.append(all_tokens)
next_batch_size += 1
new_input_length = input_length + 1
next_all_input_lengths.append(new_input_length)
next_batch_max_sequence_length = max(
next_batch_max_sequence_length, new_input_length
)
# We finished all generations in the batch; there is no next batch
if not next_batch_keep_indices:
return generated_texts, None
# If we finished at least one generation
next_batch_input_ids = {"input_ids": torch.cat(next_batch_input_ids, dim=0)}
if generated_texts:
# Apply indices to attention mask, past key values and other items that need to be cached
next_batch_input_ids["attention_mask"] = batch.input_ids["attention_mask"][
next_batch_keep_indices
]
next_batch_input_ids["past_key_values"] = [
(
keys[next_batch_past_keep_indices],
values[next_batch_past_keep_indices],
)
for keys, values in outputs["past_key_values"]
]
next_batch_requests = [batch.requests[i] for i in next_batch_keep_indices]
next_batch_next_token_choosers = [
batch.next_token_choosers[i] for i in next_batch_keep_indices
]
next_batch_stopping_criterias = [
batch.stopping_criterias[i] for i in next_batch_keep_indices
]
else:
next_batch_input_ids["attention_mask"] = batch.input_ids["attention_mask"]
next_batch_input_ids["past_key_values"] = outputs["past_key_values"]
next_batch_requests = batch.requests
next_batch_next_token_choosers = batch.next_token_choosers
next_batch_stopping_criterias = batch.stopping_criterias
# Update attention_mask with padding as we added a new token to input_ids
next_batch_input_ids["attention_mask"] = torch.cat(
[
next_batch_input_ids["attention_mask"],
torch.ones((next_batch_size, 1)).to(self.device),
],
dim=1,
)
next_batch = Batch(
batch_id=batch.batch_id,
requests=next_batch_requests,
all_input_lengths=next_all_input_lengths,
input_ids=next_batch_input_ids,
all_input_ids=next_batch_all_input_ids,
next_token_choosers=next_batch_next_token_choosers,
stopping_criterias=next_batch_stopping_criterias,
size=next_batch_size,
max_sequence_length=next_batch_max_sequence_length,
)
return generated_texts, next_batch
self, batch: B
) -> Tuple[List[GeneratedText], Optional[B]]:
raise NotImplementedError

127
server/text_generation/models/types.py
View File

@ -1,5 +1,6 @@
import torch
from abc import abstractmethod
from dataclasses import dataclass
from typing import List, Dict
@ -70,131 +71,9 @@ class Batch:
)
@classmethod
@abstractmethod
def concatenate(cls, batches: List["Batch"]) -> "Batch":
# Used for padding
total_batch_size = sum(batch.size for batch in batches)
max_sequence_length = max(batch.max_sequence_length for batch in batches)
# Batch attributes
input_ids = {"input_ids": None, "attention_mask": None, "past_key_values": []}
requests = []
all_input_lengths = []
all_input_ids = []
next_token_choosers = []
stopping_criterias = []
# 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)
all_input_lengths.extend(batch.all_input_lengths)
all_input_ids.extend(batch.all_input_ids)
next_token_choosers.extend(batch.next_token_choosers)
stopping_criterias.extend(batch.stopping_criterias)
# Slicing end index for this batch
end_index = start_index + batch.size
# We only concatenate batches that did at least one step
if batch.input_ids["input_ids"].shape[1] > 1:
raise ValueError("Batch input_ids should be of shape (batch_size, 1)")
# Initialize tensors
if i == 0:
input_ids["input_ids"] = torch.empty(
(total_batch_size, 1),
dtype=batch.input_ids["input_ids"].dtype,
device=batch.input_ids["input_ids"].device,
)
input_ids["attention_mask"] = torch.zeros(
(total_batch_size, max_sequence_length),
dtype=batch.input_ids["attention_mask"].dtype,
device=batch.input_ids["attention_mask"].device,
)
# input_ids["input_ids"] is always of shape [batch_size, 1]
# We do not need to pad it
input_ids["input_ids"][start_index:end_index] = batch.input_ids["input_ids"]
# We need to slice the attention mask to remove padding from previous steps
input_ids["attention_mask"][
start_index:end_index, -batch.max_sequence_length :
] = batch.input_ids["attention_mask"][:, -batch.max_sequence_length :]
for j, past in enumerate(batch.input_ids["past_key_values"]):
past_keys = past[0]
past_values = past[1]
_, head_dim, padded_sequence_length = past_keys.shape
# Reshape the tensors to make slicing easier
past_keys = past_keys.view(
batch.size, -1, head_dim, padded_sequence_length
)
past_values = past_values.view(
batch.size, -1, padded_sequence_length, head_dim
)
num_heads = past_keys.shape[1]
# Initialize tensors
# This will run only once per layer
if j == len(input_ids["past_key_values"]):
padded_past_keys = torch.zeros(
(
total_batch_size,
num_heads,
head_dim,
max_sequence_length - 1,
),
dtype=past_keys.dtype,
device=past_keys.device,
)
padded_past_values = torch.zeros(
(
total_batch_size,
num_heads,
max_sequence_length - 1,
head_dim,
),
dtype=past_values.dtype,
device=past_values.device,
)
input_ids["past_key_values"].append(
[padded_past_keys, padded_past_values]
)
# We slice the past keys and values to remove the padding from previous batches
input_ids["past_key_values"][j][0][
start_index:end_index, :, :, -(batch.max_sequence_length - 1) :
] = past_keys[:, :, :, -(batch.max_sequence_length - 1) :]
input_ids["past_key_values"][j][1][
start_index:end_index, :, -(batch.max_sequence_length - 1) :, :
] = past_values[:, :, -(batch.max_sequence_length - 1) :, :]
# If we are on the last batch, we need to reshape the tensors
if (i + 1) == len(batches):
input_ids["past_key_values"][j][0] = input_ids["past_key_values"][
j
][0].view(total_batch_size * num_heads, head_dim, -1)
input_ids["past_key_values"][j][1] = input_ids["past_key_values"][
j
][1].view(total_batch_size * num_heads, -1, head_dim)
start_index += batch.size
return cls(
batch_id=batches[0].batch_id,
requests=requests,
all_input_lengths=all_input_lengths,
input_ids=input_ids,
all_input_ids=all_input_ids,
next_token_choosers=next_token_choosers,
stopping_criterias=stopping_criterias,
size=total_batch_size,
max_sequence_length=max_sequence_length,
)
raise NotImplementedError
@dataclass

4
server/text_generation/server.py
View File

@ -27,7 +27,7 @@ class TextGenerationService(generate_pb2_grpc.TextGenerationServiceServicer):
return generate_pb2.ClearCacheResponse()
async def Generate(self, request, context):
batch = Batch.from_pb(request.batch, self.model.tokenizer, self.model.device)
batch = self.model.batch_type.from_pb(request.batch, self.model.tokenizer, self.model.device)
generated_texts, next_batch = self.model.generate_token(batch)
self.cache.set(next_batch)
@ -51,7 +51,7 @@ class TextGenerationService(generate_pb2_grpc.TextGenerationServiceServicer):
batches.append(batch)
if len(batches) > 1:
batch = Batch.concatenate(batches)
batch = self.model.batch_type.concatenate(batches)
else:
batch = batches[0]