source stringclasses 470
values | url stringlengths 49 167 | file_type stringclasses 1
value | chunk stringlengths 1 512 | chunk_id stringlengths 5 9 |
|---|---|---|---|---|
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/big_models.md | https://huggingface.co/docs/transformers/en/big_models/#model-data-type | .md | PyTorch model weights are normally instantiated as torch.float32 and it can be an issue if you try to load a model as a different data type. For example, you'd need twice as much memory to load the weights in torch.float32 and then again to load them in your desired data type, like torch.float16.
> [!WARNING]
> Due t... | 52_5_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/big_models.md | https://huggingface.co/docs/transformers/en/big_models/#model-data-type | .md | > [!WARNING]
> Due to how PyTorch is designed, the `torch_dtype` parameter only supports floating data types.
To avoid wasting memory like this, explicitly set the `torch_dtype` parameter to the desired data type or set `torch_dtype="auto"` to load the weights with the most optimal memory pattern (the data type is au... | 52_5_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/big_models.md | https://huggingface.co/docs/transformers/en/big_models/#model-data-type | .md | gemma = AutoModelForCausalLM.from_pretrained("google/gemma-7b", torch_dtype=torch.float16)
```
</hfoption>
<hfoption id="auto dtype">
```py
from transformers import AutoModelForCausalLM
gemma = AutoModelForCausalLM.from_pretrained("google/gemma-7b", torch_dtype="auto")
```
</hfoption>
</hfoptions>
You can also... | 52_5_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/big_models.md | https://huggingface.co/docs/transformers/en/big_models/#model-data-type | .md | my_config = AutoConfig.from_pretrained("google/gemma-2b", torch_dtype=torch.float16)
model = AutoModel.from_config(my_config)
``` | 52_5_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/ | .md | <!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agr... | 53_0_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/ | .md | an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
--> | 53_0_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#efficient-training-on-cpu | .md | This guide focuses on training large models efficiently on CPU. | 53_1_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#mixed-precision-with-ipex | .md | Mixed precision uses single (fp32) and half-precision (bf16/fp16) data types in a model to accelerate training or inference while still preserving much of the single-precision accuracy. Modern CPUs such as 3rd, 4th, and 5th Gen Intel® Xeon® Scalable processors natively support bf16. 6th Gen Intel® Xeon® Scalable proces... | 53_2_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#mixed-precision-with-ipex | .md | To further maximize training performance, you can use Intel® Extension for PyTorch (IPEX), which is a library built on PyTorch and adds additional CPU instruction level architecture (ISA) level support such as Intel® Advanced Vector Extensions 512 Vector Neural Network Instructions (Intel® AVX512-VNNI), and Intel® Adva... | 53_2_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#mixed-precision-with-ipex | .md | boost on Intel CPUs. However, CPUs with only AVX2 (e.g., AMD or older Intel CPUs) are not guaranteed to have better performance under IPEX. | 53_2_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#mixed-precision-with-ipex | .md | Auto Mixed Precision (AMP) for CPU backends has been enabled since PyTorch 1.10. AMP support for bf16/fp16 on CPUs and bf16/fp16 operator optimization is also supported in IPEX and partially upstreamed to the main PyTorch branch. You can get better performance and user experience with IPEX AMP.
Check more detailed in... | 53_2_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#ipex-installation | .md | IPEX release is following PyTorch, to install via pip:
| PyTorch Version | IPEX version |
| :---------------: | :----------: |
| 2.5.0 | 2.5.0+cpu |
| 2.4.0 | 2.4.0+cpu |
| 2.3.0 | 2.3.0+cpu |
| 2.2.0 | 2.2.0+cpu |
Please run `pip list | grep... | 53_3_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#ipex-installation | .md | Please run `pip list | grep torch` to get your `pytorch_version`, so you can get the `IPEX version_name`.
```bash
pip install intel_extension_for_pytorch==<version_name> -f https://developer.intel.com/ipex-whl-stable-cpu
```
You can check the latest versions in [ipex-whl-stable-cpu](https://developer.intel.com/ipex-whl... | 53_3_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#usage-in-trainer | .md | To enable auto mixed precision with IPEX in Trainer, users should add `use_ipex`, `bf16` or `fp16`, and `no_cuda` in training command arguments.
Take an example of the use cases on [Transformers question-answering](https://github.com/huggingface/transformers/tree/main/examples/pytorch/question-answering)
- Training... | 53_4_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#usage-in-trainer | .md | --model_name_or_path google-bert/bert-base-uncased \
--dataset_name squad \
--do_train \
--do_eval \
--per_device_train_batch_size 12 \
--learning_rate 3e-5 \
--num_train_epochs 2 \
--max_seq_length 384 \
--doc_stride 128 \
--output_dir /tmp/debug_squad/ \
<b>--use_ipex</b> \
<b>--bf16</b> \
<b>--use_cpu</b></pre>
If... | 53_4_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#usage-in-trainer | .md | ```diff
training_args = TrainingArguments(
output_dir=args.output_path,
+ bf16=True,
+ use_ipex=True,
+ use_cpu=True,
**kwargs
)
``` | 53_4_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perf_train_cpu.md | https://huggingface.co/docs/transformers/en/perf_train_cpu/#practice-example | .md | Blog: [Accelerating PyTorch Transformers with Intel Sapphire Rapids](https://huggingface.co/blog/intel-sapphire-rapids) | 53_5_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/ | .md | <!---
Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or a... | 54_0_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/ | .md | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
--> | 54_0_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#model-training-anatomy | .md | To understand performance optimization techniques that one can apply to improve efficiency of model training
speed and memory utilization, it's helpful to get familiar with how GPU is utilized during training, and how compute
intensity varies depending on an operation performed.
Let's start by exploring a motivating ... | 54_1_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#model-training-anatomy | .md | we'll need to install a few libraries:
```bash
pip install transformers datasets accelerate nvidia-ml-py3
```
The `nvidia-ml-py3` library allows us to monitor the memory usage of the models from within Python. You might be familiar
with the `nvidia-smi` command in the terminal - this library allows to access the sa... | 54_1_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#model-training-anatomy | .md | Then, we create some dummy data: random token IDs between 100 and 30000 and binary labels for a classifier.
In total, we get 512 sequences each with length 512 and store them in a [`~datasets.Dataset`] with PyTorch format.
```py
>>> import numpy as np
>>> from datasets import Dataset | 54_1_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#model-training-anatomy | .md | >>> seq_len, dataset_size = 512, 512
>>> dummy_data = {
... "input_ids": np.random.randint(100, 30000, (dataset_size, seq_len)),
... "labels": np.random.randint(0, 2, (dataset_size)),
... }
>>> ds = Dataset.from_dict(dummy_data)
>>> ds.set_format("pt")
```
To print summary statistics for the GPU utilization a... | 54_1_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#model-training-anatomy | .md | >>> def print_gpu_utilization():
... nvmlInit()
... handle = nvmlDeviceGetHandleByIndex(0)
... info = nvmlDeviceGetMemoryInfo(handle)
... print(f"GPU memory occupied: {info.used//1024**2} MB.") | 54_1_4 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#model-training-anatomy | .md | >>> def print_summary(result):
... print(f"Time: {result.metrics['train_runtime']:.2f}")
... print(f"Samples/second: {result.metrics['train_samples_per_second']:.2f}")
... print_gpu_utilization()
```
Let's verify that we start with a free GPU memory:
```py
>>> print_gpu_utilization()
GPU memory occupied... | 54_1_5 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#model-training-anatomy | .md | ```
That looks good: the GPU memory is not occupied as we would expect before we load any models. If that's not the case on
your machine make sure to stop all processes that are using GPU memory. However, not all free GPU memory can be used by
the user. When a model is loaded to the GPU the kernels are also loaded, w... | 54_1_6 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#model-training-anatomy | .md | >>> torch.ones((1, 1)).to("cuda")
>>> print_gpu_utilization()
GPU memory occupied: 1343 MB.
```
We see that the kernels alone take up 1.3GB of GPU memory. Now let's see how much space the model uses. | 54_1_7 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#load-model | .md | First, we load the `google-bert/bert-large-uncased` model. We load the model weights directly to the GPU so that we can check
how much space just the weights use.
```py
>>> from transformers import AutoModelForSequenceClassification | 54_2_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#load-model | .md | >>> model = AutoModelForSequenceClassification.from_pretrained("google-bert/bert-large-uncased").to("cuda")
>>> print_gpu_utilization()
GPU memory occupied: 2631 MB.
```
We can see that the model weights alone take up 1.3 GB of GPU memory. The exact number depends on the specific
GPU you are using. Note that on newer... | 54_2_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#load-model | .md | GPU you are using. Note that on newer GPUs a model can sometimes take up more space since the weights are loaded in an
optimized fashion that speeds up the usage of the model. Now we can also quickly check if we get the same result
as with `nvidia-smi` CLI:
```bash
nvidia-smi
```
```bash
Tue Jan 11 08:58:05 2022
+-... | 54_2_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#load-model | .md | | NVIDIA-SMI 460.91.03 Driver Version: 460.91.03 CUDA Version: 11.2 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
... | 54_2_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#load-model | .md | |===============================+======================+======================|
| 0 Tesla V100-SXM2... On | 00000000:00:04.0 Off | 0 |
| N/A 37C P0 39W / 300W | 2631MiB / 16160MiB | 0% Default |
| | | N/A |
... | 54_2_4 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#load-model | .md | +-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
... | 54_2_5 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#load-model | .md | | 0 N/A N/A 3721 C ...nvs/codeparrot/bin/python 2629MiB |
+-----------------------------------------------------------------------------+
```
We get the same number as before and you can also see that we are using a V100 GPU with 16GB of memory. So now we can
start training the model and see how... | 54_2_6 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#load-model | .md | arguments:
```py
default_args = {
"output_dir": "tmp",
"eval_strategy": "steps",
"num_train_epochs": 1,
"log_level": "error",
"report_to": "none",
}
```
<Tip>
If you plan to run multiple experiments, in order to properly clear the memory between experiments, restart the Python
kernel between experiments.
</Tip> | 54_2_7 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#memory-utilization-at-vanilla-training | .md | Let's use the [`Trainer`] and train the model without using any GPU performance optimization techniques and a batch size of 4:
```py
>>> from transformers import TrainingArguments, Trainer, logging
>>> logging.set_verbosity_error() | 54_3_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#memory-utilization-at-vanilla-training | .md | >>> training_args = TrainingArguments(per_device_train_batch_size=4, **default_args)
>>> trainer = Trainer(model=model, args=training_args, train_dataset=ds)
>>> result = trainer.train()
>>> print_summary(result)
```
```
Time: 57.82
Samples/second: 8.86
GPU memory occupied: 14949 MB.
```
We see that already a relat... | 54_3_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#memory-utilization-at-vanilla-training | .md | ```
We see that already a relatively small batch size almost fills up our GPU's entire memory. However, a larger batch size
can often result in faster model convergence or better end performance. So ideally we want to tune the batch size to our
model's needs and not to the GPU limitations. What's interesting is that ... | 54_3_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-operations | .md | Transformers architecture includes 3 main groups of operations grouped below by compute-intensity.
1. **Tensor Contractions**
Linear layers and components of Multi-Head Attention all do batched **matrix-matrix multiplications**. These operations are the most compute-intensive part of training a transformer.
2. **... | 54_4_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-operations | .md | 2. **Statistical Normalizations**
Softmax and layer normalization are less compute-intensive than tensor contractions, and involve one or more **reduction operations**, the result of which is then applied via a map.
3. **Element-wise Operators**
These are the remaining operators: **biases, dropout, activations, a... | 54_4_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-operations | .md | This knowledge can be helpful to know when analyzing performance bottlenecks.
This summary is derived from [Data Movement Is All You Need: A Case Study on Optimizing Transformers 2020](https://arxiv.org/abs/2007.00072) | 54_4_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | We've seen that training the model uses much more memory than just putting the model on the GPU. This is because there
are many components during training that use GPU memory. The components on GPU memory are the following:
1. model weights
2. optimizer states
3. gradients
4. forward activations saved for gradient co... | 54_5_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | A typical model trained in mixed precision with AdamW requires 18 bytes per model parameter plus activation memory. For
inference there are no optimizer states and gradients, so we can subtract those. And thus we end up with 6 bytes per
model parameter for mixed precision inference, plus activation memory.
Let's look... | 54_5_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | Let's look at the details.
**Model Weights:**
- 4 bytes * number of parameters for fp32 training
- 6 bytes * number of parameters for mixed precision training (maintains a model in fp32 and one in fp16 in memory)
**Optimizer States:**
- 8 bytes * number of parameters for normal AdamW (maintains 2 states)
- 2 by... | 54_5_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | - 4 bytes * number of parameters for optimizers like SGD with momentum (maintains only 1 state)
**Gradients**
- 4 bytes * number of parameters for either fp32 or mixed precision training (gradients are always kept in fp32)
**Forward Activations**
- size depends on many factors, the key ones being sequence lengt... | 54_5_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | There are the input and output that are being passed and returned by the forward and the backward functions and the
forward activations saved for gradient computation.
**Temporary Memory**
Additionally, there are all kinds of temporary variables which get released once the calculation is done, but in the
moment the... | 54_5_4 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | moment these could require additional memory and could push to OOM. Therefore, when coding it's crucial to think
strategically about such temporary variables and sometimes to explicitly free those as soon as they are no longer needed.
**Functionality-specific memory**
Then, your software could have special memory n... | 54_5_5 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | needs to maintain multiple copies of inputs and outputs.
**`forward` vs `backward` Execution Speed**
For convolutions and linear layers there are 2x flops in the backward compared to the forward, which generally translates
into ~2x slower (sometimes more, because sizes in the backward tend to be more awkward). Acti... | 54_5_6 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | bandwidth-limited, and it’s typical for an activation to have to read more data in the backward than in the forward
(e.g. activation forward reads once, writes once, activation backward reads twice, gradOutput and output of the forward,
and writes once, gradInput).
As you can see, there are potentially a few places w... | 54_5_7 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/model_memory_anatomy.md | https://huggingface.co/docs/transformers/en/model_memory_anatomy/#anatomy-of-models-memory | .md | Now that you understand what affects GPU utilization and computation speed, refer to
the [Methods and tools for efficient training on a single GPU](perf_train_gpu_one) documentation page to learn about
performance optimization techniques. | 54_5_8 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/ | .md | <!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agr... | 55_0_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/ | .md | an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered ... | 55_0_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#perplexity-of-fixed-length-models | .md | [[open-in-colab]]
Perplexity (PPL) is one of the most common metrics for evaluating language models. Before diving in, we should note
that the metric applies specifically to classical language models (sometimes called autoregressive or causal language
models) and is not well defined for masked language models like BE... | 55_1_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#perplexity-of-fixed-length-models | .md | Perplexity is defined as the exponentiated average negative log-likelihood of a sequence. If we have a tokenized
sequence \\(X = (x_0, x_1, \dots, x_t)\\), then the perplexity of \\(X\\) is,
$$\text{PPL}(X) = \exp \left\{ {-\frac{1}{t}\sum_i^t \log p_\theta (x_i|x_{<i}) } \right\}$$ | 55_1_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#perplexity-of-fixed-length-models | .md | where \\(\log p_\theta (x_i|x_{<i})\\) is the log-likelihood of the ith token conditioned on the preceding tokens \\(x_{<i}\\) according to our model. Intuitively, it can be thought of as an evaluation of the model's ability to predict uniformly among the set of specified tokens in a corpus. Importantly, this means tha... | 55_1_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#perplexity-of-fixed-length-models | .md | This is also equivalent to the exponentiation of the cross-entropy between the data and model predictions. For more
intuition about perplexity and its relationship to Bits Per Character (BPC) and data compression, check out this
[fantastic blog post on The Gradient](https://thegradient.pub/understanding-evaluation-metr... | 55_1_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#calculating-ppl-with-fixed-length-models | .md | If we weren't limited by a model's context size, we would evaluate the model's perplexity by autoregressively
factorizing a sequence and conditioning on the entire preceding subsequence at each step, as shown below.
<img width="600" alt="Full decomposition of a sequence with unlimited context length" src="https://hug... | 55_2_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#calculating-ppl-with-fixed-length-models | .md | When working with approximate models, however, we typically have a constraint on the number of tokens the model can
process. The largest version of [GPT-2](model_doc/gpt2), for example, has a fixed length of 1024 tokens, so we
cannot calculate \\(p_\theta(x_t|x_{<t})\\) directly when \\(t\\) is greater than 1024.
Ins... | 55_2_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#calculating-ppl-with-fixed-length-models | .md | Instead, the sequence is typically broken into subsequences equal to the model's maximum input size. If a model's max
input size is \\(k\\), we then approximate the likelihood of a token \\(x_t\\) by conditioning only on the
\\(k-1\\) tokens that precede it rather than the entire context. When evaluating the model's pe... | 55_2_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#calculating-ppl-with-fixed-length-models | .md | log-likelihoods of each segment independently.
<img width="600" alt="Suboptimal PPL not taking advantage of full available context" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/ppl_chunked.gif"/>
This is quick to compute since the perplexity of each segment can be computed in o... | 55_2_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#calculating-ppl-with-fixed-length-models | .md | approximation of the fully-factorized perplexity and will typically yield a higher (worse) PPL because the model will
have less context at most of the prediction steps.
Instead, the PPL of fixed-length models should be evaluated with a sliding-window strategy. This involves repeatedly
sliding the context window so th... | 55_2_4 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#calculating-ppl-with-fixed-length-models | .md | sliding the context window so that the model has more context when making each prediction.
<img width="600" alt="Sliding window PPL taking advantage of all available context" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/ppl_sliding.gif"/>
This is a closer approximation to the t... | 55_2_5 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#calculating-ppl-with-fixed-length-models | .md | favorable score. The downside is that it requires a separate forward pass for each token in the corpus. A good
practical compromise is to employ a strided sliding window, moving the context by larger strides rather than sliding by
1 token a time. This allows computation to proceed much faster while still giving the mod... | 55_2_6 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | Let's demonstrate this process with GPT-2.
```python
from transformers import GPT2LMHeadModel, GPT2TokenizerFast
from accelerate.test_utils.testing import get_backend | 55_3_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | device, _, _ = get_backend() # automatically detects the underlying device type (CUDA, CPU, XPU, MPS, etc.)
model_id = "openai-community/gpt2-large"
model = GPT2LMHeadModel.from_pretrained(model_id).to(device)
tokenizer = GPT2TokenizerFast.from_pretrained(model_id)
```
We'll load in the WikiText-2 dataset and evaluat... | 55_3_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | this dataset is small and we're just doing one forward pass over the set, we can just load and encode the entire
dataset in memory.
```python
from datasets import load_dataset | 55_3_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | test = load_dataset("wikitext", "wikitext-2-raw-v1", split="test")
encodings = tokenizer("\n\n".join(test["text"]), return_tensors="pt")
```
With 🤗 Transformers, we can simply pass the `input_ids` as the `labels` to our model, and the average negative
log-likelihood for each token is returned as the loss. With our s... | 55_3_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | the tokens we pass to the model at each iteration. We don't want the log-likelihood for the tokens we're just treating
as context to be included in our loss, so we can set these targets to `-100` so that they are ignored. The following
is an example of how we could do this with a stride of `512`. This means that the mo... | 55_3_4 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | available to condition on).
```python
import torch
from tqdm import tqdm | 55_3_5 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | max_length = model.config.n_positions
stride = 512
seq_len = encodings.input_ids.size(1)
nll_sum = 0.0
n_tokens = 0
prev_end_loc = 0
for begin_loc in tqdm(range(0, seq_len, stride)):
end_loc = min(begin_loc + max_length, seq_len)
trg_len = end_loc - prev_end_loc # may be different from stride on last loop
input_ids =... | 55_3_6 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | with torch.no_grad():
outputs = model(input_ids, labels=target_ids)
# loss is calculated using CrossEntropyLoss which averages over valid labels
# N.B. the model only calculates loss over trg_len - 1 labels, because it internally shifts the labels
# to the left by 1.
neg_log_likelihood = outputs.loss | 55_3_7 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | # Accumulate the total negative log-likelihood and the total number of tokens
num_valid_tokens = (target_ids != -100).sum().item() # number of valid tokens in target_ids
batch_size = target_ids.size(0)
num_loss_tokens = num_valid_tokens - batch_size # subtract batch_size due to internal label shift
nll_sum += neg_log... | 55_3_8 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | avg_nll = nll_sum / n_tokens # average negative log-likelihood per token
ppl = torch.exp(avg_nll)
```
Running this with the stride length equal to the max input length is equivalent to the suboptimal, non-sliding-window
strategy we discussed above. The smaller the stride, the more context the model will have in maki... | 55_3_9 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/perplexity.md | https://huggingface.co/docs/transformers/en/perplexity/#example-calculating-perplexity-with-gpt-2-in--transformers | .md | and the better the reported perplexity will typically be.
When we run the above with `stride = 1024`, i.e. no overlap, the resulting PPL is `19.44`, which is about the same
as the `19.93` reported in the GPT-2 paper. By using `stride = 512` and thereby employing our striding window
strategy, this jumps down to `16.44... | 55_3_10 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/ | .md | <!--Copyright 2024 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agr... | 56_0_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/ | .md | an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered ... | 56_0_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#agents-and-tools | .md | [[open-in-colab]] | 56_1_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#what-is-an-agent | .md | Large Language Models (LLMs) trained to perform [causal language modeling](./tasks/language_modeling) can tackle a wide range of tasks, but they often struggle with basic tasks like logic, calculation, and search. When prompted in domains in which they do not perform well, they often fail to generate the answer we expe... | 56_2_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#what-is-an-agent | .md | An agent is a system that uses an LLM as its engine, and it has access to functions called *tools*.
These *tools* are functions for performing a task, and they contain all necessary description for the agent to properly use them.
The agent can be programmed to:
- devise a series of actions/tools and run them all at... | 56_2_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#code-agent | .md | This agent has a planning step, then generates python code to execute all its actions at once. It natively handles different input and output types for its tools, thus it is the recommended choice for multimodal tasks. | 56_3_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#react-agents | .md | This is the go-to agent to solve reasoning tasks, since the ReAct framework ([Yao et al., 2022](https://huggingface.co/papers/2210.03629)) makes it really efficient to think on the basis of its previous observations.
We implement two versions of ReactJsonAgent:
- [`ReactJsonAgent`] generates tool calls as a JSON in i... | 56_4_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#react-agents | .md | > [!TIP]
> Read [Open-source LLMs as LangChain Agents](https://huggingface.co/blog/open-source-llms-as-agents) blog post to learn more about ReAct agents.
<div class="flex justify-center">
<img
class="block dark:hidden"
src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/Ag... | 56_4_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#react-agents | .md | src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/Agent_ManimCE.gif"
/>
</div>
For example, here is how a ReAct Code ag... | 56_4_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#react-agents | .md | For example, here is how a ReAct Code agent would work its way through the following question.
```py3
>>> agent.run(
... "How many more blocks (also denoted as layers) in BERT base encoder than the encoder from the architecture proposed in Attention is All You Need?",
... )
=====New task=====
How many more blocks... | 56_4_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#react-agents | .md | ====Agent is executing the code below:
bert_blocks = search(query="number of blocks in BERT base encoder")
print("BERT blocks:", bert_blocks)
====
Print outputs:
BERT blocks: twelve encoder blocks | 56_4_4 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#react-agents | .md | ====Agent is executing the code below:
attention_layer = search(query="number of layers in Attention is All You Need")
print("Attention layers:", attention_layer)
====
Print outputs:
Attention layers: Encoder: The encoder is composed of a stack of N = 6 identical layers. Each layer has two sub-layers. The first is a mu... | 56_4_5 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#react-agents | .md | ====Agent is executing the code below:
bert_blocks = 12
attention_layers = 6
diff = bert_blocks - attention_layers
print("Difference in blocks:", diff)
final_answer(diff)
====
Print outputs:
Difference in blocks: 6
Final answer: 6
``` | 56_4_6 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | To initialize an agent, you need these arguments:
- an LLM to power your agent - the agent is not exactly the LLM, it’s more like the agent is a program that uses an LLM as its engine.
- a system prompt: what the LLM engine will be prompted with to generate its output
- a toolbox from which the agent pick tools to ex... | 56_5_0 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | - a parser to extract from the LLM output which tools are to call and with which arguments
Upon initialization of the agent system, the tool attributes are used to generate a tool description, then baked into the agent’s `system_prompt` to let it know which tools it can use and why.
To start with, please install th... | 56_5_1 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | ```bash
pip install transformers[agents]
```
Build your LLM engine by defining a `llm_engine` method which accepts a list of [messages](./chat_templating) and returns text. This callable also needs to accept a `stop` argument that indicates when to stop generating.
```python
from huggingface_hub import login, Infer... | 56_5_2 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | login("<YOUR_HUGGINGFACEHUB_API_TOKEN>")
client = InferenceClient(model="meta-llama/Meta-Llama-3-70B-Instruct") | 56_5_3 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | def llm_engine(messages, stop_sequences=["Task"]) -> str:
response = client.chat_completion(messages, stop=stop_sequences, max_tokens=1000)
answer = response.choices[0].message.content
return answer
```
You could use any `llm_engine` method as long as:
1. it follows the [messages format](./chat_templating) (`List[Dic... | 56_5_4 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | 2. it stops generating outputs at the sequences passed in the argument `stop_sequences`
Additionally, `llm_engine` can also take a `grammar` argument. In the case where you specify a `grammar` upon agent initialization, this argument will be passed to the calls to llm_engine, with the `grammar` that you defined upon ... | 56_5_5 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | You will also need a `tools` argument which accepts a list of `Tools` - it can be an empty list. You can also add the default toolbox on top of your `tools` list by defining the optional argument `add_base_tools=True`.
Now you can create an agent, like [`CodeAgent`], and run it. You can also create a [`TransformersEn... | 56_5_6 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | For convenience, since agentic behaviours generally require stronger models such as `Llama-3.1-70B-Instruct` that are harder to run locally for now, we also provide the [`HfApiEngine`] class that initializes a `huggingface_hub.InferenceClient` under the hood.
```python
from transformers import CodeAgent, HfApiEngine | 56_5_7 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | llm_engine = HfApiEngine(model="meta-llama/Meta-Llama-3-70B-Instruct")
agent = CodeAgent(tools=[], llm_engine=llm_engine, add_base_tools=True)
agent.run(
"Could you translate this sentence from French, say it out loud and return the audio.",
sentence="Où est la boulangerie la plus proche?",
)
```
This will be handy ... | 56_5_8 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | agent = CodeAgent(tools=[], add_base_tools=True)
agent.run(
"Could you translate this sentence from French, say it out loud and give me the audio.",
sentence="Où est la boulangerie la plus proche?",
)
```
Note that we used an additional `sentence` argument: you can pass text as additional arguments to the model.
Y... | 56_5_9 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | agent = ReactCodeAgent(tools=[], llm_engine=llm_engine, add_base_tools=True) | 56_5_10 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | agent.run("Why does Mike not know many people in New York?", audio="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/recording.mp3")
```
The prompt and output parser were automatically defined, but you can easily inspect them by calling the `system_prompt_template` on your ag... | 56_5_11 |
/Users/nielsrogge/Documents/python_projecten/transformers/docs/source/en/agents.md | https://huggingface.co/docs/transformers/en/agents/#how-can-i-build-an-agent | .md | print(agent.system_prompt_template)
```
It's important to explain as clearly as possible the task you want to perform.
Every [`~Agent.run`] operation is independent, and since an agent is powered by an LLM, minor variations in your prompt might yield completely different results.
You can also run an agent consecutive... | 56_5_12 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.