text stringlengths 1 1.02k | class_index int64 0 10.8k | source stringlengths 85 188 |
|---|---|---|
def to_json_string(self):
"""
Serializes this instance to a JSON formatted string.
Returns:
str: JSON formatted string representing the configuration instance.
"""
return json.dumps(self.__dict__, indent=2) + "\n"
# Copied from transformers.utils.quantization_con... | 211 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# Remove all the attributes that were updated, without modifying the input dict
unused_kwargs = {key: value for key, value in kwargs.items() if key not in to_remove}
return unused_kwargs | 211 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class QuantizedCacheConfig(CacheConfig):
"""
Configuration class for quantized cache settings. | 212 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Attributes:
backend (`str`, *optional*, defaults to `"quanto"`):
Backend to use when performing quantization, Can be one of [`quanto`, `HQQ`]
nbits (`Optional[int]`, *optional*, defaults to 4):
Number of bits, can be 2 or 4 for the `quanto` backend and one of [1, 2, 3, 4, 8] for ... | 212 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Length of the residual cache which will always be stored in original presicion.
Defaults to 128.
compute_dtype (`torch.dtype`, *optional*, defaults to `torch.float16`):
The defualt dtype used for computations in the model. Keys and Values will be cast to this dtype after dequantization.
... | 212 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def __init__(
self,
backend: str = "quanto",
nbits: Optional[int] = 4,
axis_key: Optional[int] = 0,
axis_value: Optional[int] = 0,
q_group_size: Optional[int] = 64,
residual_length: Optional[int] = 128,
compute_dtype: Optional[torch.dtype] = torch.float16,... | 212 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
incorrect_arg_msg = (
"Some of the keys in `cache_config` are defined incorrectly. `{key}` should be {correct_value}` "
"but found {found_value}"
)
# Check that the values are reasonable in general (nbits, axis)
# Later in QuantizedCache init we check if they are supporte... | 212 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
incorrect_arg_msg.format(
key="residual_length",
correct_value="a positive integer",
found_value=self.residual_length,
),
) | 212 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
if self.axis_key not in [0, 1, -1]:
raise ValueError(
incorrect_arg_msg.format(
key="axis_key",
correct_value="`1` or `0`, `-1`",
found_value=self.axis_key,
),
)
if self.axis_value not in [0, 1, ... | 212 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class StaticCacheConfig(CacheConfig):
"""
Configuration class for static cache settings.
"""
cache_implementation = "static"
def __init__(self, batch_size: int, max_cache_len: int, device="cpu"):
self.batch_size = batch_size
self.max_cache_len = max_cache_len
self.device = ... | 213 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
if self.max_cache_len <= 0:
raise ValueError(
incorrect_arg_msg.format(
key="max_cache_len",
correct_value="> 0",
found_value=self.max_cache_len,
),
) | 213 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class DynamicCache(Cache):
"""
A cache that grows dynamically as more tokens are generated. This is the default for generative models.
It stores the Key and Value states as a list of tensors, one for each layer. The expected shape for each tensor is
`[batch_size, num_heads, seq_len, head_dim]`.
Ex... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
@deprecate_kwarg("num_hidden_layers", version="4.47.0")
def __init__(self, num_hidden_layers: Optional[int] = None) -> None:
super().__init__()
self._seen_tokens = 0 # Used in `generate` to keep tally of how many tokens the cache has seen
self.key_cache: List[torch.Tensor] = []
self... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def __iter__(self):
"""
Support for backwards-compatible `past_key_value` iteration, e.g. `for x in past_key_value:` to iterate over
keys and values
"""
for layer_idx in range(len(self)):
yield (self.key_cache[layer_idx], self.value_cache[layer_idx])
def __len__(... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Parameters:
key_states (`torch.Tensor`):
The new key states to cache.
value_states (`torch.Tensor`):
The new value states to cache.
layer_idx (`int`):
The index of the layer to cache the states for.
cache_kwargs (`Dict[str, ... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# Update the cache
if key_states is not None:
if len(self.key_cache) <= layer_idx:
# There may be skipped layers, fill them with empty lists
for _ in range(len(self.key_cache), layer_idx):
self.key_cache.append([])
self.value_ca... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def get_seq_length(self, layer_idx: Optional[int] = 0) -> int:
"""Returns the sequence length of the cached states. A layer index can be optionally passed."""
# TODO: deprecate this function in favor of `cache_position`
is_empty_layer = (
len(self.key_cache) == 0 # no cache in any l... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def to_legacy_cache(self) -> Tuple[Tuple[torch.Tensor], Tuple[torch.Tensor]]:
"""Converts the `DynamicCache` instance into the its equivalent in the legacy cache format. Used for
backward compatibility."""
legacy_cache = ()
for layer_idx in range(len(self)):
legacy_cache += (... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
@classmethod
@deprecate_kwarg("num_hidden_layers", version="4.47.0")
def from_legacy_cache(
cls, past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None, num_hidden_layers: int = None
) -> "DynamicCache":
"""Converts a cache in the legacy cache format into an equivalent `DynamicCac... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def crop(self, max_length: int):
"""Crop the past key values up to a new `max_length` in terms of tokens. `max_length` can also be
negative to remove `max_length` tokens. This is used in assisted decoding and contrastive search."""
# In case it is negative
if max_length < 0:
... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
@deprecate_kwarg("num_hidden_layers", version="4.47.0")
def batch_split(
self, full_batch_size: int, split_size: int, num_hidden_layers: int = None
) -> List["DynamicCache"]:
"""Split the current instance into a list of `DynamicCache` by the batch size. This will be used by
`_split_model... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
@classmethod
@deprecate_kwarg("num_hidden_layers", version="4.47.0")
def from_batch_splits(cls, splits: List["DynamicCache"], num_hidden_layers: int = None) -> "DynamicCache":
"""This is the opposite of the above `batch_split()` method. This will be used by `stack_model_outputs` in
`generation.u... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def batch_repeat_interleave(self, repeats: int):
"""Repeat the cache `repeats` times in the batch dimension. Used in contrastive search."""
for layer_idx in range(len(self)):
self.key_cache[layer_idx] = self.key_cache[layer_idx].repeat_interleave(repeats, dim=0)
self.value_cache[... | 214 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class OffloadedCache(DynamicCache):
"""
A drop-in replacement for DynamicCache that conserves GPU memory at the expense of more CPU memory.
Useful for generating from models with very long context.
In addition to the default CUDA stream, where all forward() computations happen,
this class uses anot... | 215 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def __init__(self) -> None:
if not torch.cuda.is_available():
raise RuntimeError("OffloadedCache can only be used with a GPU")
super().__init__()
self.original_device = []
self.prefetch_stream = torch.cuda.Stream()
self.beam_idx = None # used to delay beam search ope... | 215 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def evict_previous_layer(self, layer_idx: int):
"Moves the previous layer cache to the CPU"
if len(self) > 2:
# We do it on the default stream so it occurs after all earlier computations on these tensors are done
prev_layer_idx = (layer_idx - 1) % len(self)
self.key_c... | 215 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def __getitem__(self, layer_idx: int) -> List[Tuple[torch.Tensor]]:
"Gets the cache for this layer to the device. Prefetches the next and evicts the previous layer."
if layer_idx < len(self):
# Evict the previous layer if necessary
torch.cuda.current_stream().synchronize()
... | 215 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
self.prefetch_layer((layer_idx + 1) % len(self))
return (key_tensor, value_tensor)
else:
raise KeyError(f"Cache only has {len(self)} layers, attempted to access layer with index {layer_idx}") | 215 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def reorder_cache(self, beam_idx: torch.LongTensor):
"""Saves the beam indices and reorders the cache when the tensor is back to its device."""
# We delay this operation until the tensors are back to their original
# device because performing torch.index_select on the CPU is very slow
de... | 215 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def update(
self,
key_states: torch.Tensor,
value_states: torch.Tensor,
layer_idx: int,
cache_kwargs: Optional[Dict[str, Any]] = None,
) -> Tuple[torch.Tensor, torch.Tensor]:
"""
Updates the cache with the new `key_states` and `value_states` for the layer `lay... | 215 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# Update the cache
if len(self.key_cache) < layer_idx:
raise ValueError("OffloadedCache does not support model usage where layers are skipped. Use DynamicCache.")
elif len(self.key_cache) == layer_idx:
self.key_cache.append(key_states)
self.value_cache.append(value_st... | 215 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class QuantizedCache(DynamicCache):
"""
A quantizer cache similar to what is described in the [KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache paper](https://arxiv.org/abs/2402.02750).
It allows the model to generate longer sequence length without allocating too much memory for Key and Value c... | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
It stores Keys and Values a list of quantized tensors (tuples in case we need to store metadata), one for each layer. Additionally, it stores the Key and
Value in original precision states as a list of tensors, one for each layer. The size of each tensor
is `[batch_size, num_heads, seq_len - residual_length, he... | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def update(
self,
key_states: torch.Tensor,
value_states: torch.Tensor,
layer_idx: int,
cache_kwargs: Optional[Dict[str, Any]] = None,
) -> Tuple[torch.Tensor, torch.Tensor]:
# Update the number of seen tokens
if layer_idx == 0:
self._seen_tokens +... | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
if len(self.key_cache) < layer_idx:
raise ValueError("QuantizedCache does not support model usage where layers are skipped. Use DynamicCache.")
elif len(self.key_cache) == layer_idx:
self._quantized_key_cache.append(self._quantize(key_states.contiguous(), axis=self.axis_key))
... | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
values_to_return = [dequant_value, self.value_cache[layer_idx], value_states] | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
keys_to_return = torch.cat(keys_to_return, dim=-2)
values_to_return = torch.cat(values_to_return, dim=-2)
if (
self.key_cache[layer_idx].dim() == 4
and self.key_cache[layer_idx].shape[-2] + 1 >= self.residual_length
):
self._quantized_k... | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
self.value_cache[layer_idx] = torch.cat([self.value_cache[layer_idx], value_states], dim=-2) | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
return keys_to_return, values_to_return
def get_seq_length(self, layer_idx: Optional[int] = 0) -> int:
"""Returns the sequence length of the cached states. A layer index can be optionally passed."""
if len(self.key_cache) <= layer_idx:
return 0
# since we cannot get the seq_leng... | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def _dequantize(self, q_tensor):
"""Dequantizes back the tensor that was quantized by `self._quantize()`"""
raise NotImplementedError("Make sure to implement `_dequantize` in a subclass.") | 216 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class QuantoQuantizedCache(QuantizedCache):
"""
Quantized Cache class that uses `quanto` as a backend to perform quantization. Current implementation supports `int2` and `int4` dtypes only.
Parameters:
cache_config (`QuantizedCacheConfig`):
A configuration containing all the arguments t... | 217 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
>>> # Prepare a cache class and pass it to model's forward
>>> cache_config = QuantizedCacheConfig(nbits=4)
>>> past_key_values = QuantoQuantizedCache(cache_config=cache_config)
>>> outputs = model(**inputs, past_key_values=past_key_values, use_cache=True)
>>> outputs.past_key_values # a... | 217 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
if is_optimum_quanto_available():
optimum_quanto_version = version.parse(importlib.metadata.version("optimum-quanto"))
if optimum_quanto_version <= version.parse("0.2.5"):
raise ImportError(
f"You need optimum-quanto package version to be greater or equal than... | 217 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
self.qtype = qint4 if self.nbits == 4 else qint2
self.optimizer = MaxOptimizer() # hardcode as it's the only one for per-channel quantization
def _quantize(self, tensor, axis):
# We have two different API since in optimum-quanto, we don't use AffineQuantizer anymore
if is_optimum_quanto_av... | 217 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class HQQQuantizedCache(QuantizedCache):
"""
Quantized Cache class that uses `HQQ` as a backend to perform quantization. Current implementation supports `int2`, `int4`, `int8` dtypes.
Parameters:
cache_config (`QuantizedCacheConfig`):
A configuration containing all the arguments to be u... | 218 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
>>> # Prepare a cache class and pass it to model's forward
>>> cache_config = QuantizedCacheConfig(nbits=4, axis_key=1, axis_value=1)
>>> past_key_values = HQQQuantizedCache(cache_config=cache_config)
>>> outputs = model(**inputs, past_key_values=past_key_values, use_cache=True)
>>> outp... | 218 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
if self.axis_value not in [0, 1]:
raise ValueError(f"`axis_value` for `HQQ` backend has to be one of [`0`, `1`] but got {self.axis_value}")
self.quantizer = HQQQuantizer
def _quantize(self, tensor, axis):
qtensor, meta = self.quantizer.quantize(
tensor,
axis=axi... | 218 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class SinkCache(Cache):
"""
A cache that as described in the [Attention Sinks paper](https://arxiv.org/abs/2309.17453). It allows the model to
generate beyond the length of its context window, without losing fluency in the conversation. As it discards past
tokens, the model will lose the ability to gene... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
>>> model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
>>> tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
>>> inputs = tokenizer(text="My name is Qwen2", return_tensors="pt")
>>> # Prepare a cache class and pass it to model's forward
>>> pas... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def __init__(self, window_length: int, num_sink_tokens: int) -> None:
super().__init__()
self.key_cache: List[torch.Tensor] = []
self.value_cache: List[torch.Tensor] = []
self.window_length = window_length
self.num_sink_tokens = num_sink_tokens
self.cos_sin_rerotation_cac... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def _get_rerotation_cos_sin(
self, key_states: torch.Tensor, cos: torch.Tensor, sin: torch.Tensor
) -> Tuple[torch.Tensor, torch.Tensor]:
if key_states.shape[-2] not in self.cos_sin_rerotation_cache:
# Upcast to float32 temporarily for better accuracy
cos = cos.to(torch.float... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
self.cos_sin_rerotation_cache[key_states.shape[-2]] = (
rerotation_cos.to(key_states.dtype).unsqueeze(0),
rerotation_sin.to(key_states.dtype).unsqueeze(0),
)
return self.cos_sin_rerotation_cache[key_states.shape[-2]]
def get_seq_length(self, layer_idx: Optional[i... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def update(
self,
key_states: torch.Tensor,
value_states: torch.Tensor,
layer_idx: int,
cache_kwargs: Optional[Dict[str, Any]] = None,
) -> Tuple[torch.Tensor, torch.Tensor]:
"""
Updates the cache with the new `key_states` and `value_states` for the layer `lay... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Return:
A tuple containing the updated key and value states.
"""
# Optional kwargs for `SinkCache` -- needed on models using RoPE. `partial_rotation_size` is used on models
# with partially rotated position embeddings, like Phi or Persimmon.
sin = cache_kwargs.get("sin")
... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# Update the sin/cos cache, which holds sin/cos values for all possible positions
if using_rope and layer_idx == 0:
# BC: some models still pass `sin`/`cos` with 2 dims. In those models, they are the full sin/cos. Remove
# after all RoPE models have a llama-like cache utilization.
... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# [bsz, num_heads, seq_len, head_dim]
if len(self.key_cache) <= layer_idx:
# Empty cache
self.key_cache.append(key_states)
self.value_cache.append(value_states)
elif key_states.shape[-2] + self.get_seq_length(layer_idx) < self.window_length:
# Growing cac... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# On RoPE models, we need to recompute the Key rotation as the tokens are shifted
if using_rope:
rerotation_cos, rerotation_sin = self._get_rerotation_cos_sin(
key_states, self._cos_cache[: self.window_length], self._sin_cache[: self.window_length]
)
... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# Concatenate sink tokens, shifted & rotated tokens (if needed), and new tokens
sink_keys = self.key_cache[layer_idx][:, :, : self.num_sink_tokens]
self.key_cache[layer_idx] = torch.cat([sink_keys, keys_to_keep, key_states], dim=-2)
sink_values = self.value_cache[layer_idx][:, :, : ... | 219 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class StaticCache(Cache):
"""
Static Cache class to be used with `torch.compile(model)` and `torch.export()`. | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Parameters:
config (`PretrainedConfig`):
The configuration file defining the shape-related attributes required to initialize the static cache.
batch_size (`int`):
The batch size with which the model will be used. Note that a new instance must be instantiated if a
smal... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Mapping between the layers and its device. This is required when you are manually initializing the cache and the model is splitted between differents gpus.
You can know which layers mapped to which device by checking the associated device_map: `model.hf_device_map`. | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Example:
```python
>>> from transformers import AutoTokenizer, AutoModelForCausalLM, StaticCache
>>> model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-chat-hf")
>>> tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-chat-hf")
>>> inputs = tokeni... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# TODO (joao): remove `=None` in non-optional arguments in v4.46. Remove from `OBJECTS_TO_IGNORE` as well.
def __init__(
self,
config: PretrainedConfig,
batch_size: int = None,
max_cache_len: int = None,
device: torch.device = None,
dtype: torch.dtype = torch.float32,... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# Some model define a custom `head_dim` != config.hidden_size // config.num_attention_heads
self.head_dim = (
config.head_dim if hasattr(config, "head_dim") else config.hidden_size // config.num_attention_heads
)
self.dtype = dtype
self.num_key_value_heads = (
co... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
self.key_cache: List[torch.Tensor] = []
self.value_cache: List[torch.Tensor] = []
# Note: There will be significant perf decrease if switching to use 5D tensors instead.
cache_shape = (self.max_batch_size, self.num_key_value_heads, self.max_cache_len, self.head_dim)
for idx in range(conf... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# 2. `torch.export()` requires mutations to be registered as buffers.
if not is_torchdynamo_compiling():
self.register_buffer(f"key_cache_{idx}", torch.zeros(cache_shape, dtype=dtype, device=layer_device))
self.register_buffer(f"value_cache_{idx}", torch.zeros(cache_shape, dt... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def update(
self,
key_states: torch.Tensor,
value_states: torch.Tensor,
layer_idx: int,
cache_kwargs: Optional[Dict[str, Any]] = None,
) -> Tuple[torch.Tensor, torch.Tensor]:
"""
Updates the cache with the new `key_states` and `value_states` for the layer `lay... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Return:
A tuple containing the updated key and value states.
"""
cache_position = cache_kwargs.get("cache_position")
k_out = self.key_cache[layer_idx]
v_out = self.value_cache[layer_idx]
key_states = key_states.to(k_out.dtype)
value_states = value_states.to(... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
if cache_position is None:
k_out.copy_(key_states)
v_out.copy_(value_states)
else:
# Note: here we use `tensor.index_copy_(dim, index, tensor)` that is equivalent to
# `tensor[:, :, index] = tensor`, but the first one is compile-friendly and it does explicitly an ... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def get_seq_length(self, layer_idx: Optional[int] = 0) -> int:
"""Returns the sequence length of the cached states that were seen by the model."""
# Occupied cache == any slot in the 3rd dim (sequence length) holds a non-zero value. To save on compute, let's
# limit the check to the first batch ... | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
@property
def batch_size(self):
logger.warning_once(
f"The 'batch_size' attribute of {self.__class__.__name__} is deprecated and will be removed in "
"v4.49. Use the more precisely named 'self.max_batch_size' attribute instead."
)
return self.max_batch_size | 220 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class SlidingWindowCache(StaticCache):
"""
Sliding Window Cache class to be used with `torch.compile` for models like Mistral that support sliding window attention.
Every time when we try to update the cache, we compute the `indices` based on `cache_position >= self.config.sliding_window - 1`,
if true(w... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
indices = (slicing + to_shift[-1].int()-1) % self.config.sliding_window
tensor([ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18,
19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36,
37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54,
... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Parameters:
config (`PretrainedConfig`):
The configuration file defining the shape-related attributes required to initialize the static cache.
batch_size (`int`):
The batch size with which the model will be used. Note that a new instance must be instantiated if a
smal... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
You can know which layers mapped to which device by checking the associated device_map: `model.hf_device_map`. | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Example:
```python
>>> from transformers import AutoTokenizer, AutoModelForCausalLM, SlidingWindowCache
>>> model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
>>> tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
>>... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
>>> # Prepare a cache class and pass it to model's forward
>>> # Leave empty space for 10 new tokens, which can be used when calling forward iteratively 10 times to generate
>>> max_generated_length = inputs.input_ids.shape[1] + 10
>>> past_key_values = SlidingWindowCache(config=model.config, ba... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# TODO (joao): remove `=None` in non-optional arguments in v4.46. Remove from `OBJECTS_TO_IGNORE` as well.
def __init__(
self,
config: PretrainedConfig,
batch_size: int = None,
max_cache_len: int = None,
device: torch.device = None,
dtype: torch.dtype = torch.float32,... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
max_cache_len=max_cache_len,
device=device,
dtype=dtype,
max_batch_size=max_batch_size,
layer_device_map=layer_device_map,
) | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def update(
self,
key_states: torch.Tensor,
value_states: torch.Tensor,
layer_idx: int,
cache_kwargs: Optional[Dict[str, Any]] = None,
) -> Tuple[torch.Tensor]:
cache_position = cache_kwargs.get("cache_position")
k_out = self.key_cache[layer_idx]
v_out... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# assume this only happens in prefill phase when prompt length > sliding_window_size (= max_cache_len)
if cache_position.shape[0] > self.max_cache_len:
k_out = key_states[:, :, -self.max_cache_len :, :]
v_out = value_states[:, :, -self.max_cache_len :, :]
# Assumption: caches... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
slicing = torch.ones(self.max_cache_len, dtype=torch.long, device=value_states.device).cumsum(0)
cache_position = cache_position.clamp(0, self.max_cache_len - 1)
to_shift = cache_position >= self.max_cache_len - 1
indices = (slicing + to_shift[-1].int() - 1) % self.max_cache_len
k_out =... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
self.key_cache[layer_idx] += k_out
self.value_cache[layer_idx] += v_out
return k_out, v_out
def get_max_cache_shape(self) -> Optional[int]:
return self.max_cache_len
def reset(self):
for layer_idx in range(len(self.key_cache)):
# In-place ops prevent breaking the s... | 221 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class EncoderDecoderCache(Cache):
"""
Base, abstract class for all encoder-decoder caches. Can be used to hold combinations of self-attention and
cross-attention caches.
Example:
```python
>>> from transformers import AutoProcessor, AutoModelForCausalLM, DynamicCache, EncoderDecoderCac... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
>>> # Prepare cache classes for encoder and decoder and pass it to model's forward
>>> self_attention_cache = DynamicCache()
>>> cross_attention_cache = DynamicCache()
>>> past_key_values = EncoderDecoderCache(self_attention_cache, cross_attention_cache)
>>> outputs = model(**inputs, pas... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def __getitem__(self, layer_idx: int) -> List[Tuple[torch.Tensor]]:
"""
Support for backwards-compatible `past_key_value` indexing, e.g. `past_key_value[0][0].shape[2]` to get the
sequence length.
"""
if layer_idx < len(self):
return (
self.self_attent... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def to_legacy_cache(self) -> Tuple[Tuple[torch.Tensor], Tuple[torch.Tensor]]:
"""Converts the `EncoderDecoderCache` instance into its equivalent in the legacy cache format."""
legacy_cache = ()
if len(self.cross_attention_cache) > 0:
for self_attn, cross_attn in zip(
... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
@classmethod
def from_legacy_cache(
cls, past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None
) -> "EncoderDecoderCache":
"""Converts a cache in the legacy cache format into an equivalent `EncoderDecoderCache`."""
cache = cls(
self_attention_cache=DynamicCache(),... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def get_seq_length(self, layer_idx: Optional[int] = 0) -> int:
"""Returns the sequence length of the cached states. A layer index can be optionally passed."""
# check if empty list because in case of static cache it will be a tensors and we can't check `if not torch.Tensor`
return self.self_atte... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def reset(self):
if hasattr(self.self_attention_cache, "reset"):
self.self_attention_cache.reset()
if hasattr(self.cross_attention_cache, "reset"):
self.cross_attention_cache.reset()
elif not hasattr(self.self_attention_cache, "reset") and not hasattr(self.cross_attention... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def reorder_cache(self, beam_idx: torch.LongTensor):
"""Reorders the cache for beam search, given the selected beam indices."""
self.self_attention_cache.reorder_cache(beam_idx)
self.cross_attention_cache.reorder_cache(beam_idx)
def check_dynamic_cache(self, method: str):
if not (
... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
# TODO(gante, sanchit-gandhi): move following functionality into `.generate`
def crop(self, maximum_length: int):
"""Crop the past key values up to a new `maximum_length` in terms of tokens. `maximum_length` can also be
negative to remove `maximum_length` tokens. This is used in assisted decoding an... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
@deprecate_kwarg("num_hidden_layers", version="4.47.0")
def batch_split(
self, full_batch_size: int, split_size: int, num_hidden_layers: int = None
) -> "List[EncoderDecoderCache]":
"""Split the current instance into a list of `DynamicCache` by the batch size. This will be used by
`_spli... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
@classmethod
@deprecate_kwarg("num_hidden_layers", version="4.47.0")
def from_batch_splits(
cls, splits: List["EncoderDecoderCache"], num_hidden_layers: int = None
) -> "EncoderDecoderCache":
"""This is the opposite of the above `batch_split()` method. This will be used by `stack_model_outpu... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
layer_keys = torch.cat([current.cross_attention_cache.key_cache[idx] for current in splits], dim=0)
layer_values = torch.cat([current.cross_attention_cache.value_cache[idx] for current in splits], dim=0)
cross_attention_cache.update(layer_keys, layer_values, idx)
return cls(self_attentio... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
def batch_select_indices(self, indices: torch.Tensor):
"""Only keep the `indices` in the batch dimension of the cache. Used in contrastive search."""
self.check_dynamic_cache(self.batch_select_indices.__name__)
self.self_attention_cache.batch_select_indices(indices)
self.cross_attention_... | 222 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
class HybridCache(Cache):
"""
Hybrid Cache class to be used with `torch.compile` for Gemma2 models that alternate between a local sliding window attention
and global attention in every other layer. Under the hood, Hybrid Cache leverages ["SlidingWindowCache"] for sliding window attention
and ["StaticCac... | 223 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Parameters:
config (`PretrainedConfig):
The configuration file defining the shape-related attributes required to initialize the static cache.
batch_size (`int`):
The batch size with which the model will be used. Note that a new instance must be instantiated if a
small... | 223 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
You can know which layers mapped to which device by checking the associated device_map: `model.hf_device_map`. | 223 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Example:
```python
>>> from transformers import AutoTokenizer, AutoModelForCausalLM, HybridCache
>>> model = AutoModelForCausalLM.from_pretrained("google/gemma-2-2b")
>>> tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-2b")
>>> inputs = tokenizer(text="My name is Gem... | 223 | /Users/nielsrogge/Documents/python_projecten/transformers/src/transformers/cache_utils.py |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.