code stringlengths 66 870k | docstring stringlengths 19 26.7k | func_name stringlengths 1 138 | language stringclasses 1
value | repo stringlengths 7 68 | path stringlengths 5 324 | url stringlengths 46 389 | license stringclasses 7
values |
|---|---|---|---|---|---|---|---|
def concatenated_inputs(
batch: Dict[str, Union[List, torch.LongTensor]],
is_encoder_decoder: bool = False,
label_pad_token_id: int = -100,
padding_value: int = 0,
device: Optional[torch.device] = None,
) -> Dict[str, torch.LongTensor]:
"""Concatenate the chosen and r... | Concatenate the chosen and rejected inputs into a single tensor.
Args:
batch: A batch of data. Must contain the keys 'chosen_input_ids' and 'rejected_input_ids', which are tensors of shape (batch_size, sequence_length).
is_encoder_decoder: Whether the model is an encoder-decoder model.
... | concatenated_inputs | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def dpo_loss(
self,
policy_chosen_logps: torch.FloatTensor,
policy_rejected_logps: torch.FloatTensor,
reference_chosen_logps: torch.FloatTensor,
reference_rejected_logps: torch.FloatTensor,
) -> Tuple[torch.FloatTensor, torch.FloatTensor, torch.FloatTensor]:
"""Comput... | Compute the DPO loss for a batch of policy and reference model log probabilities.
Args:
policy_chosen_logps: Log probabilities of the policy model for the chosen responses. Shape: (batch_size,)
policy_rejected_logps: Log probabilities of the policy model for the rejected responses. Shap... | dpo_loss | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def get_batch_logps(
logits: torch.FloatTensor,
labels: torch.LongTensor,
average_log_prob: bool = False,
label_pad_token_id: int = -100,
is_encoder_decoder: bool = False,
) -> torch.FloatTensor:
"""Compute the log probabilities of the given labels under the given log... | Compute the log probabilities of the given labels under the given logits.
Args:
logits: Logits of the model (unnormalized). Shape: (batch_size, sequence_length, vocab_size)
labels: Labels for which to compute the log probabilities. Label tokens with a value of label_pad_token_id are ign... | get_batch_logps | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def concatenated_forward(
self, model: nn.Module, batch: Dict[str, Union[List, torch.LongTensor]]
) -> Tuple[torch.FloatTensor, torch.FloatTensor, torch.FloatTensor, torch.FloatTensor]:
"""Run the given model on the given batch of inputs, concatenating the chosen and rejected inputs together.
... | Run the given model on the given batch of inputs, concatenating the chosen and rejected inputs together.
We do this to avoid doing two forward passes, because it's faster for FSDP.
| concatenated_forward | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def get_batch_loss_metrics(
self,
model,
batch: Dict[str, Union[List, torch.LongTensor]],
train_eval: Literal["train", "eval"] = "train",
):
"""Compute the DPO loss and other metrics for the given batch of inputs for train or test.
CHANGE: 1. add sft loss
2. a... | Compute the DPO loss and other metrics for the given batch of inputs for train or test.
CHANGE: 1. add sft loss
2. all gather metrics
| get_batch_loss_metrics | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def get_batch_samples(self, model, batch: Dict[str, torch.LongTensor]) -> Tuple[str, str]:
"""Generate samples from the model and reference model for the given batch of inputs."""
# If one uses `generate_during_eval` with peft + bf16, we need to explictly call generate with
# the torch cuda amp... | Generate samples from the model and reference model for the given batch of inputs. | get_batch_samples | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def evaluation_loop(
self,
dataloader: DataLoader,
description: str,
prediction_loss_only: Optional[bool] = None,
ignore_keys: Optional[List[str]] = None,
metric_key_prefix: str = "eval",
) -> EvalLoopOutput:
"""
Overriding built-in evaluation loop to ... |
Overriding built-in evaluation loop to store metrics for each batch.
Prediction/evaluation loop, shared by `Trainer.evaluate()` and `Trainer.predict()`.
Works both with or without labels.
| evaluation_loop | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def log(self, logs: Dict[str, float]) -> None:
"""
Log `logs` on the various objects watching training, including stored metrics.
Args:
logs (`Dict[str, float]`):
The values to log.
"""
# logs either has 'loss' or 'eval_loss'
train_eval = "tra... |
Log `logs` on the various objects watching training, including stored metrics.
Args:
logs (`Dict[str, float]`):
The values to log.
| log | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def push_to_hub(self, commit_message: Optional[str] = "End of training", blocking: bool = True, **kwargs) -> str:
"""
Overwrite the `push_to_hub` method in order to force-add the tag "sft" when pushing the
model on the Hub. Please refer to `~transformers.Trainer.push_to_hub` for more details.
... |
Overwrite the `push_to_hub` method in order to force-add the tag "sft" when pushing the
model on the Hub. Please refer to `~transformers.Trainer.push_to_hub` for more details.
| push_to_hub | python | NVlabs/VILA | llava/trl/trainer/dpo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/dpo_trainer.py | Apache-2.0 |
def _step_safety_checker(
input_ids: List[torch.LongTensor],
attention_mask: List[torch.LongTensor],
labels: List[torch.LongTensor],
texts: List[str],
texts_labels: List[str],
):
"""
Check if the input data is valid for training.
Args:
inp... |
Check if the input data is valid for training.
Args:
input_ids (List[`torch.LongTensor`]):
List of tensors containing the input_ids
attention_mask (List[`torch.LongTensor`]):
List of tensors containing the attention_mask
labels (List[... | _step_safety_checker | python | NVlabs/VILA | llava/trl/trainer/iterative_sft_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/iterative_sft_trainer.py | Apache-2.0 |
def step(
self,
input_ids: Optional[List[torch.LongTensor]] = None,
attention_mask: Optional[List[torch.LongTensor]] = None,
labels: Optional[List[torch.LongTensor]] = None,
texts: Optional[List[str]] = None,
texts_labels: Optional[List[str]] = None,
):
"""
... |
Run an optimisation step given a list of input_ids, attention_mask, and labels or a list of text and text_labels.
Args:
input_ids (List[`torch.LongTensor`]):
List of tensors containing the input_ids (if not provided, text will be used)
attention_mask (List[`torch... | step | python | NVlabs/VILA | llava/trl/trainer/iterative_sft_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/iterative_sft_trainer.py | Apache-2.0 |
def __init__(
self,
config: PPOConfig = None,
model: PreTrainedModelWrapper = None,
ref_model: Optional[PreTrainedModelWrapper] = None,
tokenizer: PreTrainedTokenizerBase = None,
dataset: Optional[Union[torch.utils.data.Dataset, Dataset]] = None,
optimizer: Option... |
Initialize PPOTrainer.
Args:
config (`PPOConfig`):
Configuration object for PPOTrainer. Check the documentation of `PPOConfig` for more details.
model (`PreTrainedModelWrapper`):
Hugging Face transformer model with a value head.
ref_m... | __init__ | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def prepare_dataloader(self, dataset: Union[torch.utils.data.Dataset, Dataset], data_collator=None):
"""
Prepare the dataloader for training.
Args:
dataset (Union[`torch.utils.data.Dataset`, `datasets.Dataset`]):
PyTorch dataset or Hugging Face dataset. If a Hugging ... |
Prepare the dataloader for training.
Args:
dataset (Union[`torch.utils.data.Dataset`, `datasets.Dataset`]):
PyTorch dataset or Hugging Face dataset. If a Hugging Face dataset is passed, the dataset
will be preprocessed by removing the columns that are not us... | prepare_dataloader | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def generate(
self,
query_tensor: Union[torch.Tensor, List[torch.Tensor]],
length_sampler: Callable = None,
batch_size: int = 4,
return_prompt: bool = True,
generate_ref_response: bool = False,
**generation_kwargs,
):
"""
Generate response with... |
Generate response with the model given the query tensor.
call the `generate` method of the model.
Args:
query_tensor (`torch.LongTensor`):
A tensor of shape (`seq_len`) containing query tokens or a list of tensors of shape (`seq_len`).
length_sampler (`C... | generate | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def _step_safety_checker(
self,
batch_size: int,
queries: List[torch.LongTensor],
responses: List[torch.LongTensor],
scores: List[torch.FloatTensor],
masks: Optional[List[torch.LongTensor]] = None,
):
"""
Check if the input data is valid for training.
... |
Check if the input data is valid for training.
Args:
batch_size (int):
Batch size from the config file.
queries (List[`torch.LongTensor`]):
List of tensors containing the encoded queries of shape (`query_length`)
responses (List[`torc... | _step_safety_checker | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def step(
self,
queries: List[torch.LongTensor],
responses: List[torch.LongTensor],
scores: List[torch.FloatTensor],
response_masks: Optional[List[torch.LongTensor]] = None,
):
"""
Run a PPO optimisation step given a list of queries, model responses, and rewar... |
Run a PPO optimisation step given a list of queries, model responses, and rewards.
Args:
queries (List[`torch.LongTensor`]):
List of tensors containing the encoded queries of shape (`query_length`)
responses (List[`torch.LongTensor`]):
List of te... | step | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def _early_stop(self, policykl):
r"""
Handles the early stopping logic. If the policy KL is greater than the target KL, then the gradient is zeroed and
the optimization step is skipped.
This also handles the multi-gpu case where the policy KL is averaged across all processes.
Ar... |
Handles the early stopping logic. If the policy KL is greater than the target KL, then the gradient is zeroed and
the optimization step is skipped.
This also handles the multi-gpu case where the policy KL is averaged across all processes.
Args:
policy_kl (torch.Tensor):
... | _early_stop | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def gather_stats(self, stats):
"""
Gather stats from all processes. Useful in the context of distributed training.
Args:
stats (dict[str, Any]):
a dictionary of stats to be gathered. The stats should contain torch tensors.
Returns:
`dict[str, Any]`: ... |
Gather stats from all processes. Useful in the context of distributed training.
Args:
stats (dict[str, Any]):
a dictionary of stats to be gathered. The stats should contain torch tensors.
Returns:
`dict[str, Any]`: A dictionary of stats with the tensors gat... | gather_stats | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def batched_forward_pass(
self,
model: PreTrainedModelWrapper,
queries: torch.Tensor,
responses: torch.Tensor,
model_inputs: dict,
return_logits: bool = False,
response_masks: Optional[torch.Tensor] = None,
):
"""
Calculate model outputs in mul... |
Calculate model outputs in multiple batches.
Args:
queries (`torch.LongTensor`):
List of tensors containing the encoded queries, shape (`batch_size`, `query_length`)
responses (`torch.LongTensor`):
List of tensors containing the encoded responses... | batched_forward_pass | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def train_minibatch(
self,
old_logprobs: torch.FloatTensor,
values: torch.FloatTensor,
logprobs: torch.FloatTensor,
logits: torch.FloatTensor,
vpreds: torch.FloatTensor,
mask: torch.LongTensor,
advantages: torch.FloatTensor,
returns: torch.FloatTen... |
Train one PPO minibatch
Args:
logprobs (`torch.FloatTensor`):
Log probabilities of the model, shape [mini_batch_size, response_length]
values (`torch.FloatTensor`):
Values of the value head, shape [mini_batch_size, response_length]
qu... | train_minibatch | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def compute_rewards(
self,
scores: torch.FloatTensor,
logprobs: torch.FloatTensor,
ref_logprobs: torch.FloatTensor,
masks: torch.LongTensor,
):
"""
Compute per token rewards from scores and KL-penalty.
Args:
scores (`torch.FloatTensor`):
... |
Compute per token rewards from scores and KL-penalty.
Args:
scores (`torch.FloatTensor`):
Scores from the reward model, shape (`batch_size`)
logprobs (`torch.FloatTensor`):
Log probabilities of the model, shape (`batch_size`, `response_length`)
... | compute_rewards | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def loss(
self,
old_logprobs: torch.FloatTensor,
values: torch.FloatTensor,
logits: torch.FloatTensor,
vpreds: torch.FloatTensor,
logprobs: torch.FloatTensor,
mask: torch.LongTensor,
advantages: torch.FloatTensor,
returns: torch.FloatTensor,
):... |
Calculate policy and value losses.
Args:
old_logprobs (`torch.FloatTensor`):
Log probabilities of the model, shape (`batch_size`, `response_length`)
values (`torch.FloatTensor`):
Values of the value head, shape (`batch_size`, `response_length`)
... | loss | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def record_step_stats(self, kl_coef: float, **data):
"""
Record training step statistics.
Args:
kl_coef (`float`):
KL coefficient
data (`dict`):
Dictionary of training step data
Returns:
stats (`dict`):
... |
Record training step statistics.
Args:
kl_coef (`float`):
KL coefficient
data (`dict`):
Dictionary of training step data
Returns:
stats (`dict`):
Dictionary of training step statistics
| record_step_stats | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def log_stats(
self,
stats: dict,
batch: dict,
rewards: List[torch.FloatTensor],
columns_to_log: List[str] = ["query", "response"],
):
"""
A function that logs all the training stats. Call it at the end of each epoch.
Args:
stats (dict[str... |
A function that logs all the training stats. Call it at the end of each epoch.
Args:
stats (dict[str, Any]):
A dictionary of training stats.
batch (dict[str, Any]):
A dictionary of batch data, this contains the queries and responses.
... | log_stats | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def create_model_card(self, path: str, model_name: Optional[str] = "TRL Model") -> None:
"""Creates and saves a model card for a TRL model.
Args:
path (`str`): The path to save the model card to.
model_name (`str`, *optional*): The name of the model, defaults to `TRL Model`.
... | Creates and saves a model card for a TRL model.
Args:
path (`str`): The path to save the model card to.
model_name (`str`, *optional*): The name of the model, defaults to `TRL Model`.
| create_model_card | python | NVlabs/VILA | llava/trl/trainer/ppo_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/ppo_trainer.py | Apache-2.0 |
def __init__(
self,
model: Union[PreTrainedModel, nn.Module] = None,
args: Optional[RewardConfig] = None,
data_collator: Optional[DataCollator] = None,
train_dataset: Optional[Dataset] = None,
eval_dataset: Optional[Union[Dataset, Dict[str, Dataset]]] = None,
toke... |
Initialize RewardTrainer.
Args:
model (`transformers.PreTrainedModel`):
The model to train, preferably an `AutoModelForSequenceClassification`.
args (`RewardConfig`):
The arguments to use for training.
data_collator (`transformers.Dat... | __init__ | python | NVlabs/VILA | llava/trl/trainer/reward_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/reward_trainer.py | Apache-2.0 |
def _trl_activate_neftune(self, model):
r"""
Activates the neftune as presented in this code: https://github.com/neelsjain/NEFTune and paper: https://arxiv.org/abs/2310.05914
Since in transformers Trainer we do have an `_activate_neftune` method, we need to rename this method to avoid conflicts.... |
Activates the neftune as presented in this code: https://github.com/neelsjain/NEFTune and paper: https://arxiv.org/abs/2310.05914
Since in transformers Trainer we do have an `_activate_neftune` method, we need to rename this method to avoid conflicts.
| _trl_activate_neftune | python | NVlabs/VILA | llava/trl/trainer/sft_trainer.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/sft_trainer.py | Apache-2.0 |
def __init__(self, accelerator):
"""
Calculates the running mean and standard deviation of a data stream. Reference:
https://github.com/OpenLMLab/MOSS-RLHF/blob/40b91eb2f2b71b16919addede0341d2bef70825d/utils.py#L75
"""
self.mean = 0
self.std = 1
self.var = 1
... |
Calculates the running mean and standard deviation of a data stream. Reference:
https://github.com/OpenLMLab/MOSS-RLHF/blob/40b91eb2f2b71b16919addede0341d2bef70825d/utils.py#L75
| __init__ | python | NVlabs/VILA | llava/trl/trainer/utils.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/utils.py | Apache-2.0 |
def update(self, xs: torch.Tensor) -> Tuple[float, float]:
"""
Updates running moments from batch's moments computed across ranks
"""
if self.accelerator.use_distributed:
xs_mean, xs_var, xs_count = get_global_statistics(self.accelerator, xs)
else:
xs_coun... |
Updates running moments from batch's moments computed across ranks
| update | python | NVlabs/VILA | llava/trl/trainer/utils.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/utils.py | Apache-2.0 |
def get_global_statistics(accelerator, xs: torch.Tensor, mask=None, device="cpu") -> Tuple[float, float, int]:
"""
Computes element-wise mean and variance of the tensor across processes. Reference:
https://github.com/OpenLMLab/MOSS-RLHF/blob/40b91eb2f2b71b16919addede0341d2bef70825d/utils.py#L57C1-L73C75
... |
Computes element-wise mean and variance of the tensor across processes. Reference:
https://github.com/OpenLMLab/MOSS-RLHF/blob/40b91eb2f2b71b16919addede0341d2bef70825d/utils.py#L57C1-L73C75
| get_global_statistics | python | NVlabs/VILA | llava/trl/trainer/utils.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/utils.py | Apache-2.0 |
def neftune_post_forward_hook(module, input, output):
"""
Implements the NEFTune forward pass for the model using forward hooks. Note this works only for
torch.nn.Embedding layers. This method is slightly adapted from the original source code
that can be found here: https://github.com/neelsjain/NEFTune
... |
Implements the NEFTune forward pass for the model using forward hooks. Note this works only for
torch.nn.Embedding layers. This method is slightly adapted from the original source code
that can be found here: https://github.com/neelsjain/NEFTune
Simply add it to your model as follows:
```python
... | neftune_post_forward_hook | python | NVlabs/VILA | llava/trl/trainer/utils.py | https://github.com/NVlabs/VILA/blob/master/llava/trl/trainer/utils.py | Apache-2.0 |
def compute_file_md5sum(fname: Union[str, BinaryIO], chunksize: int = 1000000) -> str:
"""Compute the md5sum of a file in chunks.
Parameters
----------
fname : Union[str, BinaryIO]
Filename or file object
chunksize : int, optional
Chunk size in bytes, by default 1000000
Returns... | Compute the md5sum of a file in chunks.
Parameters
----------
fname : Union[str, BinaryIO]
Filename or file object
chunksize : int, optional
Chunk size in bytes, by default 1000000
Returns
-------
str
MD5 sum of the file
Examples
--------
>>> compute_fi... | compute_file_md5sum | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def compute_file_md5sum(fname: Union[str, BinaryIO], chunksize: int = 1000000) -> str:
"""Compute the md5sum of a file in chunks."""
md5 = hashlib.md5()
if isinstance(fname, str):
with open(fname, "rb") as f:
for chunk in iter(lambda: f.read(chunksize), b""):
md5.update(c... | Compute the md5sum of a file in chunks. | compute_file_md5sum | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def splitname(fname):
"""Returns the basename and extension of a filename"""
assert "." in fname, "Filename must have an extension"
# basename, extension = re.match(r"^((?:.*/)?.*?)(\..*)$", fname).groups()
# wids v1:
# basename, extension = os.path.splitext(fname)
# wids v2:
# fit MMDU mu... | Returns the basename and extension of a filename | splitname | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def group_by_key(names):
"""Group the file names by key.
Args:
names: A list of file names.
Returns:
A list of lists of indices, where each sublist contains indices of files
with the same key.
"""
groups = []
kmaps = {}
for i, fname in enumerate(names):
# Ig... | Group the file names by key.
Args:
names: A list of file names.
Returns:
A list of lists of indices, where each sublist contains indices of files
with the same key.
| group_by_key | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def default_decoder(sample: Dict[str, Any], format: Optional[Union[bool, str]] = True):
"""A default decoder for webdataset.
This handles common file extensions: .txt, .cls, .cls2,
.jpg, .png, .json, .npy, .mp, .pt, .pth, .pickle, .pkl.
These are the most common extensions used in webdataset.
F... | A default decoder for webdataset.
This handles common file extensions: .txt, .cls, .cls2,
.jpg, .png, .json, .npy, .mp, .pt, .pth, .pickle, .pkl.
These are the most common extensions used in webdataset.
For other extensions, users can provide their own decoder.
Args:
sample: sample, mo... | default_decoder | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def f(shard):
"""Given a URL, return a local name for the shard."""
if shard.startswith("pipe:"):
# uuencode the entire URL string
hex32 = base64.urlsafe_b64encode(hashlib.sha256(shard.encode()).digest())[:32].decode()
return os.path.join(dldir, "pipe__" + hex32)
... | Given a URL, return a local name for the shard. | f | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def interpret_transformations(transformations):
"""Interpret the transformations argument.
This takes care of transformations specified as string shortcuts
and returns a list of callables.
"""
if not isinstance(transformations, list):
transformations = [transformations]
result = []
... | Interpret the transformations argument.
This takes care of transformations specified as string shortcuts
and returns a list of callables.
| interpret_transformations | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def hash_dataset_name(input_string):
"""Compute a hash of the input string and return the first 16 characters of the hash."""
# Compute SHA256 hash of the input string
hash_object = hashlib.sha256(input_string.encode())
hash_digest = hash_object.digest()
# Encode the hash in base64
base64_encod... | Compute a hash of the input string and return the first 16 characters of the hash. | hash_dataset_name | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def __init__(
self,
shards,
*,
cache_size=int(1e12),
cache_dir=None,
lru_size=10,
dataset_name=None,
localname=None,
transformations="PIL",
keep=False,
base=None,
options=None,
):
"""Create a ShardListDataset.
... | Create a ShardListDataset.
Args:
shards: a list of (filename, length) pairs or a URL pointing to a JSON descriptor file
cache_size: the number of shards to keep in the cache
lru_size: the number of shards to keep in the LRU cache
localname: a function that maps U... | __init__ | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def check_cache_misses(self):
"""Check if the cache miss rate is too high."""
accesses, misses = self.get_stats()
if accesses > 100 and misses / accesses > 0.3:
# output a warning only once
self.check_cache_misses = lambda: None
print(f"Warning: ShardListDatas... | Check if the cache miss rate is too high. | check_cache_misses | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def get_shard(self, index):
"""Get the shard and index within the shard corresponding to the given index."""
# Find the shard corresponding to the given index.
shard_idx = np.searchsorted(self.cum_lengths, index, side="right")
# Figure out which index within the shard corresponds to the... | Get the shard and index within the shard corresponding to the given index. | get_shard | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def __getitem__(self, index):
"""Return the sample corresponding to the given index."""
shard, inner_idx, desc = self.get_shard(index)
sample = shard[inner_idx]
# Check if we're missing the cache too often.
self.check_cache_misses()
sample["__dataset__"] = desc.get("dat... | Return the sample corresponding to the given index. | __getitem__ | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def lengths_to_ranges(lengths):
"""Convert a list of lengths to a list of ranges."""
ranges = []
start = 0
for length in lengths:
ranges.append((start, start + length))
start += length
return ranges | Convert a list of lengths to a list of ranges. | lengths_to_ranges | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def intersect_range(a, b):
"""Return the intersection of the two half-open integer intervals."""
result = max(a[0], b[0]), min(a[1], b[1])
if result[0] >= result[1]:
return None
return result | Return the intersection of the two half-open integer intervals. | intersect_range | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def intersect_ranges(rangelist, r):
"""Return the intersection of the half-open integer interval r with the list of half-open integer intervals."""
result = []
for a in rangelist:
x = intersect_range(a, r)
if x is not None:
result.append(x)
return result | Return the intersection of the half-open integer interval r with the list of half-open integer intervals. | intersect_ranges | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def iterate_ranges(ranges, rng, indexshuffle=True, shardshuffle=True):
"""Iterate over the ranges in a random order."""
shard_indexes = list(range(len(ranges)))
if shardshuffle:
rng.shuffle(shard_indexes)
for i in shard_indexes:
lo, hi = ranges[i]
sample_indexes = list(range(lo, ... | Iterate over the ranges in a random order. | iterate_ranges | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def DistributedChunkedSampler(
dataset: Dataset,
*,
num_replicas: Optional[int] = None,
num_samples: Optional[int] = None,
rank: Optional[int] = None,
shuffle: bool = True,
shufflefirst: bool = False,
seed: int = 0,
drop_last: bool = None,
chunksize: int = 1000000,
) -> ChunkedSa... | Return a ChunkedSampler for the current worker in distributed training.
Reverts to a simple ChunkedSampler if not running in distributed mode.
Since the split among workers takes place before the chunk shuffle,
workers end up with a fixed set of shards they need to download. The
more workers, the fewe... | DistributedChunkedSampler | python | NVlabs/VILA | llava/wids/wids.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids.py | Apache-2.0 |
def keep_most_recent_files(pattern, maxsize=int(1e12), maxfiles=1000, debug=False):
"""Keep the most recent files in a directory, deleting the rest.
The maxsize is the maximum size of the directory in bytes. The maxfiles is
the maximum number of files to keep. The files are sorted by modification
time,... | Keep the most recent files in a directory, deleting the rest.
The maxsize is the maximum size of the directory in bytes. The maxfiles is
the maximum number of files to keep. The files are sorted by modification
time, and the most recent files are kept. If the directory is already
smaller than maxsize, ... | keep_most_recent_files | python | NVlabs/VILA | llava/wids/wids_cleanup.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_cleanup.py | Apache-2.0 |
def create_cleanup_background_process(pattern, maxsize=int(1e12), maxfiles=1000, every=60):
"""Create a background process that keeps a directory below a certain size."""
def cleanup_worker(every):
# use a lock file to ensure that only one cleanup worker is running
lockfile = os.path.join(os.pa... | Create a background process that keeps a directory below a certain size. | create_cleanup_background_process | python | NVlabs/VILA | llava/wids/wids_cleanup.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_cleanup.py | Apache-2.0 |
def download_file_no_log(remote, local, handlers=default_cmds):
"""Download a file from a remote url to a local path.
The remote url can be a pipe: url, in which case the remainder of
the url is treated as a command template that is executed to perform the download.
"""
if remote.startswith("pipe:"... | Download a file from a remote url to a local path.
The remote url can be a pipe: url, in which case the remainder of
the url is treated as a command template that is executed to perform the download.
| download_file_no_log | python | NVlabs/VILA | llava/wids/wids_dl.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_dl.py | Apache-2.0 |
def urldir(url):
"""Return the directory part of a url."""
parsed_url = urlparse(url)
path = parsed_url.path
directory = os.path.dirname(path)
return parsed_url._replace(path=directory).geturl() | Return the directory part of a url. | urldir | python | NVlabs/VILA | llava/wids/wids_index.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_index.py | Apache-2.0 |
def urlfile(url):
"""Return the file part of a url."""
parsed_url = urlparse(url)
path = parsed_url.path
filename = os.path.basename(path)
return filename | Return the file part of a url. | urlfile | python | NVlabs/VILA | llava/wids/wids_index.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_index.py | Apache-2.0 |
def main_create(args):
"""Create a full shard index for a list of files."""
# set default output file name
if args.output is None:
args.output = "shardindex.json"
if args.name is None:
first = os.path.splitext(args.files[0])[0]
args.name = shorten_name(first)
print("sett... | Create a full shard index for a list of files. | main_create | python | NVlabs/VILA | llava/wids/wids_index.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_index.py | Apache-2.0 |
def main_info(args):
"""Show info about an index file."""
if args.table:
print("file\tname\tnbytes\tnsamples\tbase\tlast\tdatasets")
for filename in args.filenames:
data = load_remote_dsdesc_raw(filename)
print(
filename,
data.get("name"),
... | Show info about an index file. | main_info | python | NVlabs/VILA | llava/wids/wids_index.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_index.py | Apache-2.0 |
def main():
"""Commands for manipulating the shard index."""
# Create the top-level parser
parser = argparse.ArgumentParser(description="Command line tool with subcommands for file operations.")
subparsers = parser.add_subparsers(dest="command", required=True, help="Subcommands")
# Create the parse... | Commands for manipulating the shard index. | main | python | NVlabs/VILA | llava/wids/wids_index.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_index.py | Apache-2.0 |
def __init__(self, capacity: int, release_handler=None):
"""Initialize a new LRU cache with the given capacity."""
self.capacity = capacity
self.cache = OrderedDict()
self.release_handler = release_handler | Initialize a new LRU cache with the given capacity. | __init__ | python | NVlabs/VILA | llava/wids/wids_lru.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_lru.py | Apache-2.0 |
def __getitem__(self, key):
"""Return the value associated with the given key, or None."""
if key not in self.cache:
return None
self.cache.move_to_end(key)
return self.cache[key] | Return the value associated with the given key, or None. | __getitem__ | python | NVlabs/VILA | llava/wids/wids_lru.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_lru.py | Apache-2.0 |
def __setitem__(self, key, value):
"""Associate the given value with the given key."""
if key in self.cache:
self.cache.move_to_end(key)
self.cache[key] = value
if len(self.cache) > self.capacity:
key, value = self.cache.popitem(last=False)
if self.rel... | Associate the given value with the given key. | __setitem__ | python | NVlabs/VILA | llava/wids/wids_lru.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_lru.py | Apache-2.0 |
def __delitem__(self, key):
"""Remove the given key from the cache."""
if key in self.cache:
if self.release_handler is not None:
value = self.cache[key]
self.release_handler(key, value)
del self.cache[key] | Remove the given key from the cache. | __delitem__ | python | NVlabs/VILA | llava/wids/wids_lru.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_lru.py | Apache-2.0 |
def keep_while_reading(fname, fd, phase, delay=0.0):
"""This is a possible cleanup callback for cleanup_callback of MIndexedTar.
It assumes that as long as there are some readers for a file,
more readers may be trying to open it.
Note that on Linux, unlinking the file doesn't matter after
it has b... | This is a possible cleanup callback for cleanup_callback of MIndexedTar.
It assumes that as long as there are some readers for a file,
more readers may be trying to open it.
Note that on Linux, unlinking the file doesn't matter after
it has been mmapped. The contents will only be deleted when
all ... | keep_while_reading | python | NVlabs/VILA | llava/wids/wids_mmtar.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_mmtar.py | Apache-2.0 |
def urlmerge(base, url):
"""Merge a base URL and a relative URL.
The function fills in any missing part of the url from the base,
except for params, query, and fragment, which are taken only from the 'url'.
For the pathname component, it merges the paths like os.path.join:
an absolute path in 'url'... | Merge a base URL and a relative URL.
The function fills in any missing part of the url from the base,
except for params, query, and fragment, which are taken only from the 'url'.
For the pathname component, it merges the paths like os.path.join:
an absolute path in 'url' overrides the base path, otherw... | urlmerge | python | NVlabs/VILA | llava/wids/wids_specs.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_specs.py | Apache-2.0 |
def check_shards(l):
"""Check that a list of shards is well-formed.
This checks that the list is a list of dictionaries, and that
each dictionary has a "url" and a "nsamples" key.
"""
assert isinstance(l, list)
for shard in l:
assert isinstance(shard, dict)
assert "url" in shard... | Check that a list of shards is well-formed.
This checks that the list is a list of dictionaries, and that
each dictionary has a "url" and a "nsamples" key.
| check_shards | python | NVlabs/VILA | llava/wids/wids_specs.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_specs.py | Apache-2.0 |
def set_all(l, k, v):
"""Set a key to a value in a list of dictionaries."""
if v is None:
return
for x in l:
if k not in x:
x[k] = v | Set a key to a value in a list of dictionaries. | set_all | python | NVlabs/VILA | llava/wids/wids_specs.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_specs.py | Apache-2.0 |
def load_remote_dsdesc_raw(source):
"""Load a remote or local dataset description in JSON format."""
if isinstance(source, str):
with tempfile.TemporaryDirectory() as tmpdir:
dlname = os.path.join(tmpdir, "dataset.json")
with download_and_open(source, dlname) as f:
... | Load a remote or local dataset description in JSON format. | load_remote_dsdesc_raw | python | NVlabs/VILA | llava/wids/wids_specs.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_specs.py | Apache-2.0 |
def resolve_dsdesc(dsdesc, *, options=None, base=None):
"""Resolve a dataset description.
This rebases the shards as necessary and loads any remote references.
Dataset descriptions are JSON files. They must have the following format;
{
"wids_version": 1,
# optional immediate shardlist... | Resolve a dataset description.
This rebases the shards as necessary and loads any remote references.
Dataset descriptions are JSON files. They must have the following format;
{
"wids_version": 1,
# optional immediate shardlist
"shardlist": [
{"url": "http://example.com... | resolve_dsdesc | python | NVlabs/VILA | llava/wids/wids_specs.py | https://github.com/NVlabs/VILA/blob/master/llava/wids/wids_specs.py | Apache-2.0 |
def quantize_module_config(module_name):
"""
Generate quantization config to apply W8A8 quantization for specified module.
Rest of the model is kept in FP32 precision.
"""
config = LinearQuantizerConfig(
global_config=ModuleLinearQuantizerConfig(
milestones=[0, 1000, 1000, 0],
... |
Generate quantization config to apply W8A8 quantization for specified module.
Rest of the model is kept in FP32 precision.
| quantize_module_config | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/activation_quantization.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/activation_quantization.py | MIT |
def quantize_cumulative_config(skip_conv_layers, skip_einsum_layers):
"""
Generate quantization config to apply W8A8 quantization.
Skipped layers are kept in W8A32 precision.
"""
logger.info(f"Skipping {len(skip_conv_layers)} conv layers and {len(skip_einsum_layers)} einsum layers")
w8config = M... |
Generate quantization config to apply W8A8 quantization.
Skipped layers are kept in W8A32 precision.
| quantize_cumulative_config | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/activation_quantization.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/activation_quantization.py | MIT |
def quantize(model, config, calibration_data):
"""
Apply post training activation quantization to specified model, using calibration data
"""
submodules = dict(model.named_modules(remove_duplicate=True))
layer_norm_modules = [key for key, val in submodules.items() if isinstance(val, LayerNormANE)]
... |
Apply post training activation quantization to specified model, using calibration data
| quantize | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/activation_quantization.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/activation_quantization.py | MIT |
def recipe_overrides_for_inference_speedup(conv_layers, skipped_conv):
"""
Quantize the slowest conv layers, even if in skipped set based on PSNR, for good inference speedup
"""
for layer in conv_layers:
if "up_blocks" in layer and "resnets" in layer and "conv1" in layer:
if layer in... |
Quantize the slowest conv layers, even if in skipped set based on PSNR, for good inference speedup
| recipe_overrides_for_inference_speedup | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/activation_quantization.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/activation_quantization.py | MIT |
def recipe_overrides_for_quality(conv_layers, skipped_conv):
"""
Do not quantize out projection layers to avoid quantizing outputs of preceding concat layers.
Quantizing output of concat layers can lead to quality degradation, due to sharing of scales
across concat inputs, which can have varied ranges. ... |
Do not quantize out projection layers to avoid quantizing outputs of preceding concat layers.
Quantizing output of concat layers can lead to quality degradation, due to sharing of scales
across concat inputs, which can have varied ranges. Since this is a constraint enforced during
model conversion, it ... | recipe_overrides_for_quality | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/activation_quantization.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/activation_quantization.py | MIT |
def register_input_log_hook(unet, inputs):
"""
Register forward pre hook to save model inputs
"""
def hook(_, input):
input_copy = deepcopy(input)
input_copy = tuple(i.to('cpu') for i in input_copy)
inputs.append(input_copy)
# Return inputs unmodified
return inp... |
Register forward pre hook to save model inputs
| register_input_log_hook | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/activation_quantization.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/activation_quantization.py | MIT |
def register_input_preprocessing_hook(pipe):
"""
Register forward pre hook to convert UNet inputs from HuggingFace StableDiffusionPipeline
to match expected model inputs in UNet2DConditionModel defined in unet.py
"""
def hook(_, args, kwargs):
sample = args[0]
timestep = args[1]
... |
Register forward pre hook to convert UNet inputs from HuggingFace StableDiffusionPipeline
to match expected model inputs in UNet2DConditionModel defined in unet.py
| register_input_preprocessing_hook | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/activation_quantization.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/activation_quantization.py | MIT |
def prepare_pipe(pipe, unet):
"""
Create a new pipeline from `pipe` with `unet` as the noise predictor
"""
new_pipe = deepcopy(pipe)
unet.to(new_pipe.unet.device)
new_pipe.unet = unet
pre_hook_handle = register_input_preprocessing_hook(new_pipe)
return new_pipe, pre_hook_handle |
Create a new pipeline from `pipe` with `unet` as the noise predictor
| prepare_pipe | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/activation_quantization.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/activation_quantization.py | MIT |
def split_einsum(q, k, v, mask, heads, dim_head):
""" Attention Implementation backing AttentionImplementations.SPLIT_EINSUM
- Implements https://machinelearning.apple.com/research/neural-engine-transformers
- Recommended for ANE
- Marginally slower on GPU
"""
mh_q = [
q[:, head_idx * d... | Attention Implementation backing AttentionImplementations.SPLIT_EINSUM
- Implements https://machinelearning.apple.com/research/neural-engine-transformers
- Recommended for ANE
- Marginally slower on GPU
| split_einsum | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/attention.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/attention.py | MIT |
def split_einsum_v2(q, k, v, mask, heads, dim_head):
""" Attention Implementation backing AttentionImplementations.SPLIT_EINSUM_V2
- Implements https://machinelearning.apple.com/research/neural-engine-transformers
- Recommended for ANE
- Marginally slower on GPU
- Chunks the query sequence to avoid... | Attention Implementation backing AttentionImplementations.SPLIT_EINSUM_V2
- Implements https://machinelearning.apple.com/research/neural-engine-transformers
- Recommended for ANE
- Marginally slower on GPU
- Chunks the query sequence to avoid large intermediate tensors and improves ANE performance
... | split_einsum_v2 | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/attention.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/attention.py | MIT |
def original(q, k, v, mask, heads, dim_head):
""" Attention Implementation backing AttentionImplementations.ORIGINAL
- Not recommended for ANE
- Recommended for GPU
"""
bs = q.size(0)
mh_q = q.view(bs, heads, dim_head, -1)
mh_k = k.view(bs, heads, dim_head, -1)
mh_v = v.view(bs, heads, ... | Attention Implementation backing AttentionImplementations.ORIGINAL
- Not recommended for ANE
- Recommended for GPU
| original | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/attention.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/attention.py | MIT |
def _verify_output_correctness_of_chunks(full_model,
first_chunk_model=None,
second_chunk_model=None,
pipeline_model=None,):
""" Verifies the end-to-end output correctness of full (original) mo... | Verifies the end-to-end output correctness of full (original) model versus chunked models
| _verify_output_correctness_of_chunks | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/chunk_mlprogram.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/chunk_mlprogram.py | MIT |
def _load_prog_from_mlmodel(model):
""" Load MIL Program from an MLModel
"""
model_spec = model.get_spec()
start_ = time.time()
logger.info(
"Loading MLModel object into a MIL Program object (including the weights).."
)
prog = _milproto_to_pymil(
model_spec=model_spec,
... | Load MIL Program from an MLModel
| _load_prog_from_mlmodel | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/chunk_mlprogram.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/chunk_mlprogram.py | MIT |
def _get_op_idx_split_location(prog: Program):
""" Find the op that approximately bisects the graph as measure by weights size on each side
"""
main_block = prog.functions["main"]
main_block.operations = list(main_block.operations)
total_size_in_mb = 0
for op in main_block.operations:
i... | Find the op that approximately bisects the graph as measure by weights size on each side
| _get_op_idx_split_location | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/chunk_mlprogram.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/chunk_mlprogram.py | MIT |
def _make_first_chunk_prog(prog, op_idx):
""" Build first chunk by declaring early outputs and removing unused subgraph
"""
block = prog.functions["main"]
boundary_vars = _get_first_chunk_outputs(block, op_idx)
# Due to possible numerical issues, cast any fp16 var to fp32
new_boundary_vars = _a... | Build first chunk by declaring early outputs and removing unused subgraph
| _make_first_chunk_prog | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/chunk_mlprogram.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/chunk_mlprogram.py | MIT |
def _make_second_chunk_prog(prog, op_idx):
""" Build second chunk by rebuilding a pristine MIL Program from MLModel
"""
block = prog.functions["main"]
block.opset_version = ct.target.iOS16
# First chunk outputs are second chunk inputs (e.g. skip connections)
boundary_vars = _get_first_chunk_out... | Build second chunk by rebuilding a pristine MIL Program from MLModel
| _make_second_chunk_prog | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/chunk_mlprogram.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/chunk_mlprogram.py | MIT |
def _macos_version():
"""
Returns macOS version as a tuple of integers. On non-Macs, returns an empty tuple.
"""
if sys.platform == "darwin":
try:
ver_str = subprocess.run(["sw_vers", "-productVersion"], stdout=subprocess.PIPE).stdout.decode('utf-8').strip('\n')
return tu... |
Returns macOS version as a tuple of integers. On non-Macs, returns an empty tuple.
| _macos_version | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/coreml_model.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/coreml_model.py | MIT |
def get_resource_type(resources_dir: str) -> str:
"""
Detect resource type based on filepath extensions.
returns:
`packages`: for .mlpackage resources
'compiled`: for .mlmodelc resources
"""
directories = [f for f in os.listdir(resources_dir) if os.path.isdir(os.path.... |
Detect resource type based on filepath extensions.
returns:
`packages`: for .mlpackage resources
'compiled`: for .mlmodelc resources
| get_resource_type | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/coreml_model.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/coreml_model.py | MIT |
def _load_mlpackage(submodule_name,
mlpackages_dir,
model_version,
compute_unit,
sources=None):
"""
Load Core ML (mlpackage) models from disk (As exported by torch2coreml.py)
"""
# if sources not provided, attempt to i... |
Load Core ML (mlpackage) models from disk (As exported by torch2coreml.py)
| _load_mlpackage | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/coreml_model.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/coreml_model.py | MIT |
def _load_mlpackage_controlnet(mlpackages_dir, model_version, compute_unit):
""" Load Core ML (mlpackage) models from disk (As exported by torch2coreml.py)
"""
model_name = model_version.replace("/", "_")
logger.info(f"Loading controlnet_{model_name} mlpackage")
fname = f"ControlNet_{model_name}.m... | Load Core ML (mlpackage) models from disk (As exported by torch2coreml.py)
| _load_mlpackage_controlnet | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/coreml_model.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/coreml_model.py | MIT |
def __init__(self,
num_channels,
clip_mag=None,
eps=1e-5,
elementwise_affine=True):
"""
Args:
num_channels: Number of channels (C) where the expected input data format is BC1S. S stands for sequence length.
... |
Args:
num_channels: Number of channels (C) where the expected input data format is BC1S. S stands for sequence length.
clip_mag: Optional float value to use for clamping the input range before layer norm is applied.
If specified, helps red... | __init__ | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/layer_norm.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/layer_norm.py | MIT |
def _ensure_numerical_range_and_cast(val, low, high, np_dtype):
'''
For some cases, the computed quantized data might exceed the data range.
For instance, after rounding and addition, we might get `128` for the int8 quantization.
This utility function ensures the val in the data range be... |
For some cases, the computed quantized data might exceed the data range.
For instance, after rounding and addition, we might get `128` for the int8 quantization.
This utility function ensures the val in the data range before doing the cast.
| _ensure_numerical_range_and_cast | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/mixed_bit_compression_pre_analysis.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/mixed_bit_compression_pre_analysis.py | MIT |
def get_coreml_pipe(pytorch_pipe,
mlpackages_dir,
model_version,
compute_unit,
delete_original_pipe=True,
scheduler_override=None,
controlnet_models=None,
force_zeros_for_empty_pro... |
Initializes and returns a `CoreMLStableDiffusionPipeline` from an original
diffusers PyTorch pipeline
sources: 'packages' or 'compiled' forces creation of model from specified sources. sources must be in mlpackages_dir
| get_coreml_pipe | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/pipeline.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/pipeline.py | MIT |
def get_image_path(args, **override_kwargs):
""" mkdir output folder and encode metadata in the filename
"""
out_folder = os.path.join(args.o, "_".join(args.prompt.replace("/", "_").rsplit(" ")))
os.makedirs(out_folder, exist_ok=True)
out_fname = f"randomSeed_{override_kwargs.get('seed', None) or a... | mkdir output folder and encode metadata in the filename
| get_image_path | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/pipeline.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/pipeline.py | MIT |
def compute_psnr(a, b):
""" Compute Peak-Signal-to-Noise-Ratio across two numpy.ndarray objects
"""
max_b = np.abs(b).max()
sumdeltasq = 0.0
sumdeltasq = ((a - b) * (a - b)).sum()
sumdeltasq /= b.size
sumdeltasq = np.sqrt(sumdeltasq)
eps = 1e-5
eps2 = 1e-10
psnr = 20 * np.log1... | Compute Peak-Signal-to-Noise-Ratio across two numpy.ndarray objects
| compute_psnr | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def report_correctness(original_outputs, final_outputs, log_prefix):
""" Report PSNR values across two compatible tensors
"""
original_psnr = compute_psnr(original_outputs, original_outputs)
final_psnr = compute_psnr(original_outputs, final_outputs)
dB_change = final_psnr - original_psnr
logger... | Report PSNR values across two compatible tensors
| report_correctness | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def quantize_weights(args):
""" Quantize weights to args.quantize_nbits using a palette (look-up table)
"""
for model_name in ["text_encoder", "text_encoder_2", "unet", "refiner", "control-unet"]:
logger.info(f"Quantizing {model_name} to {args.quantize_nbits}-bit precision")
out_path = _get_... | Quantize weights to args.quantize_nbits using a palette (look-up table)
| quantize_weights | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def _compile_coreml_model(source_model_path, output_dir, final_name):
""" Compiles Core ML models using the coremlcompiler utility from Xcode toolchain
"""
target_path = os.path.join(output_dir, f"{final_name}.mlmodelc")
if os.path.exists(target_path):
logger.warning(
f"Found existin... | Compiles Core ML models using the coremlcompiler utility from Xcode toolchain
| _compile_coreml_model | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def bundle_resources_for_swift_cli(args):
"""
- Compiles Core ML models from mlpackage into mlmodelc format
- Download tokenizer resources for the text encoder
"""
resources_dir = os.path.join(args.o, "Resources")
if not os.path.exists(resources_dir):
os.makedirs(resources_dir, exist_ok=... |
- Compiles Core ML models from mlpackage into mlmodelc format
- Download tokenizer resources for the text encoder
| bundle_resources_for_swift_cli | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def patched_make_causal_mask(input_ids_shape, dtype, device, past_key_values_length: int = 0):
""" Patch to replace torch.finfo(dtype).min with -1e4
"""
bsz, tgt_len = input_ids_shape
mask = torch.full((tgt_len, tgt_len), torch.tensor(-1e4, device=device), device=device)
mask_cond = torch.arange(mas... | Patch to replace torch.finfo(dtype).min with -1e4
| patched_make_causal_mask | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def convert_text_encoder(text_encoder, tokenizer, submodule_name, args):
""" Converts the text encoder component of Stable Diffusion
"""
text_encoder = text_encoder.to(dtype=torch.float32)
out_path = _get_out_path(args, submodule_name)
if os.path.exists(out_path):
logger.info(
f"... | Converts the text encoder component of Stable Diffusion
| convert_text_encoder | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def convert_vae_decoder(pipe, args):
""" Converts the VAE Decoder component of Stable Diffusion
"""
out_path = _get_out_path(args, "vae_decoder")
if os.path.exists(out_path):
logger.info(
f"`vae_decoder` already exists at {out_path}, skipping conversion."
)
return
... | Converts the VAE Decoder component of Stable Diffusion
| convert_vae_decoder | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def convert_vae_decoder_sd3(args):
""" Converts the VAE component of Stable Diffusion 3
"""
out_path = _get_out_path(args, "vae_decoder")
if os.path.exists(out_path):
logger.info(
f"`vae_decoder` already exists at {out_path}, skipping conversion."
)
return
# ... | Converts the VAE component of Stable Diffusion 3
| convert_vae_decoder_sd3 | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
def convert_vae_encoder(pipe, args):
""" Converts the VAE Encoder component of Stable Diffusion
"""
out_path = _get_out_path(args, "vae_encoder")
if os.path.exists(out_path):
logger.info(
f"`vae_encoder` already exists at {out_path}, skipping conversion."
)
return
... | Converts the VAE Encoder component of Stable Diffusion
| convert_vae_encoder | python | apple/ml-stable-diffusion | python_coreml_stable_diffusion/torch2coreml.py | https://github.com/apple/ml-stable-diffusion/blob/master/python_coreml_stable_diffusion/torch2coreml.py | MIT |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.