transformers – skill Khác cho agent AI

Transformers

Overview

The Hugging Face Transformers library provides access to thousands of pre-trained models for tasks across NLP, computer vision, audio, and multimodal domains. Use this skill to load models, perform inference, and fine-tune on custom data.

Installation

Targets Transformers 5.18.0, verified against its released source on 2026-10-01. Native CPU checks use Python 3.11, Torch 2.14.1, Datasets 5.0.1, Accelerate 1.15.0, PEFT 0.21.2, and Hub 1.33.0. The Torch extra requires Torch >=2.5. Install in a dedicated environment:

uv venv --python 3.11 .venv-transformers
uv pip install --python .venv-transformers/bin/python "transformers[torch]==5.18.0" "torch==2.14.1" "huggingface-hub==1.33.0" "datasets==5.0.1" "accelerate==1.15.0" "peft==0.21.2"

Use .venv-transformers/Scripts/python.exe on Windows. Select an appropriate Torch build for the target hardware before installation. Hub 2.1.1 is newer, but Datasets 5.0.1 requires Hub <2; upgrading every package independently makes this training stack unsatisfiable. The separate esm SDK currently requires Transformers <5 and belongs in another environment.

Optional dependencies (install only for the selected workflow): Pillow 12.3.0 for images; torchvision matched to Torch and timm 1.0.30 for models that require them; librosa 1.0.0 and soundfile 0.14.0 for audio preprocessing (librosa requires Python >=3.12); FFmpeg for encoded audio file inputs; pytesseract plus Tesseract for OCR document pipelines; bitsandbytes 0.50.2 for supported quantization backends. See [model loading](references/models.md) before choosing precision or quantization.

Verification used tiny random models, synthetic input, and local save/reload only. Hub pretrained examples throughout this skill are illustrative: public checkpoint metadata and configurations were reviewed, but no weights or datasets were downloaded and no hosted inference or uploads were run. Optional export/distributed/hardware paths are source-checked, not end-to-end tested. See [review evidence](references/review.md).

Check your version:

import transformers
print(transformers.__version__)

Authentication

Many models on the Hugging Face Hub are gated or private. Authenticate before loading them.

Recommended: CLI login (uses $HF_TOKEN_PATH, defaulting to $HF_HOME/token, normally ~/.cache/huggingface/token):

hf auth login

Python:

from huggingface_hub import login
login()  # Interactive prompt; do not hardcode tokens in scripts

Servers / CI: set HF_TOKEN in the environment (never commit tokens to git or shell profiles):

export HF_TOKEN="..."  # Read token from a secret manager, not source code

Get tokens at: https://huggingface.co/settings/tokens

Security: Never paste tokens into notebooks, repos, or shared configs. Prefer hf auth login over exporting tokens in .bashrc or .zshrc.

Use the narrowest token scope that works: read for private or gated model downloads, write only for uploads. If a long-running environment should not send the stored token on every Hub request, set HF_HUB_DISABLE_IMPLICIT_TOKEN=1 and pass a token only where authentication is required.

Transformers v5

Transformers v5 is PyTorch-only (TensorFlow and JAX backends were removed). For upgrades from v4, see the [v5 migration guide](https://github.com/huggingface/transformers/blob/v5.18.0/MIGRATION_GUIDE_V5.md). Transformers 5.18.0 accepts Hub >=1.31,<3; preserve the tighter constraint of Datasets when training.

Gated or custom architectures: accept the model license on the Hub, then load with trust_remote_code=True only when required custom code has been reviewed; pin its full immutable commit with revision (and code_revision for a separate code repository). Gating and custom code are independent: a gated built-in architecture does not require remote code.

Cache location: set HF_HOME for all Hugging Face caches, or HF_HUB_CACHE just for Hub files. Use HF_HUB_OFFLINE=1 only after required model snapshots are already cached.

Quick Start

Use the Pipeline API for fast inference without manual configuration:

from transformers import pipeline

# Text generation (prefer max_new_tokens for causal LMs)
generator = pipeline("text-generation", model="Qwen/Qwen2.5-1.5B")
result = generator("The future of AI is", max_new_tokens=50)

# Text classification
classifier = pipeline("text-classification", model="distilbert/distilbert-base-uncased-finetuned-sst-2-english")
result = classifier("This movie was excellent!")

# Generative question answering: verify responses against the supplied context.
qa = pipeline("text-generation", model="Qwen/Qwen2.5-0.5B-Instruct")
result = qa([{ "role": "user", "content": "Context: AI means artificial intelligence. What does AI mean?" }], max_new_tokens=32, do_sample=False)

Core Capabilities

1. Pipelines for Quick Inference

Use for simple, optimized inference across many tasks. Supports text generation, classification, NER, image classification, object detection, audio classification, and more. In v5, question-answering, summarization, translation*, text2text-generation, image-to-text, and visual-question-answering pipelines are removed. Use direct task models when exact extractive/seq2seq semantics are needed; generative text/VLM pipelines are different tasks, not equivalent replacements.

When to use: Quick prototyping, simple inference tasks, no custom preprocessing needed.

See references/pipelines.md for comprehensive task coverage and optimization.

2. Model Loading and Management

Load pre-trained models with fine-grained control over configuration, device placement, and precision.

When to use: Custom model initialization, advanced device management, model inspection.

See references/models.md for loading patterns and best practices.

3. Text Generation

Generate text with LLMs using various decoding strategies (greedy, beam search, sampling) and control parameters (temperature, top-k, top-p).

When to use: Creative text generation, code generation, conversational AI, text completion.

For chat or instruction-tuned checkpoints, format messages with that checkpoint's tokenizer.apply_chat_template rather than hand-written role delimiters. Prefer tokenize=True; if formatting with tokenize=False and tokenizing afterward, set add_special_tokens=False to avoid duplicated BOS/EOS tokens. Use add_generation_prompt=True to start a new assistant reply, and preserve the same template when preparing fine-tuning data.

See references/generation.md for generation strategies and parameters.

4. Training and Fine-Tuning

Fine-tune pre-trained models on custom datasets using the Trainer API with automatic mixed precision, distributed training, and logging.

When to use: Task-specific model adaptation, domain adaptation, improving model performance.

See references/training.md for training workflows and best practices.

5. Tokenization

Convert text to tokens and token IDs for model input, with padding, truncation, and special token handling.

When to use: Custom preprocessing pipelines, understanding model inputs, batch processing.

See references/tokenizers.md for tokenization details.

Common Patterns

Pattern 1: Simple Inference

For straightforward tasks, use pipelines:

pipe = pipeline("task-name", model="model-id")
output = pipe(input_data)

Pattern 2: Custom Model Usage

For advanced control, load model and tokenizer separately:

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("model-id")
model = AutoModelForCausalLM.from_pretrained("model-id", device_map="auto")

inputs = tokenizer("text", return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=100)
result = tokenizer.decode(outputs[0])

Pattern 3: Fine-Tuning

For task adaptation, use Trainer:

from transformers import Trainer, TrainingArguments

training_args = TrainingArguments(
    output_dir="./results",
    num_train_epochs=3,
    per_device_train_batch_size=8,
    report_to="none",
)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=train_dataset,
    processing_class=tokenizer,
)

trainer.train()

Research validation

Record package versions, checkpoint and dataset revisions, label order, preprocessing, split units, seeds, and generation settings. Split by patient, subject, document family, or time when observations are dependent; fit preprocessing and tune hyperparameters on training/validation data only. Report truncation and excluded records. A softmax score is not calibrated certainty, and decoding choices do not establish factual accuracy. Compare against held-out baselines and inspect failures before scientific use. Test adapters such as SHAP against the actual

… (nội dung đầy đủ có trong gói tải về)