long-context
Implements context extension strategies for transformer models, enabling them to handle inputs exceeding their original token limits.
Install
mkdir -p .claude/skills/long-context && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/4435" && unzip -o skill.zip -d .claude/skills/long-context && rm skill.zipInstalls to .claude/skills/long-context
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Extend context windows of transformer models using RoPE, YaRN, ALiBi, and position interpolation techniques. Use when processing long documents (32k-128k+ tokens), extending pre-trained models beyond original context limits, or implementing efficient positional encodings. Covers rotary embeddings, attention biases, interpolation methods, and extrapolation strategies for LLMs.Key capabilities
- →Implement RoPE
- →Implement ALiBi
- →Apply position interpolation
- →Extend context windows
- →Optimize memory usage
How it works
The skill uses positional encoding techniques like RoPE and YaRN to modify transformer attention mechanisms for longer sequences.
Inputs & outputs
When to use long-context
- →Process 128k+ token documents
- →Extend pre-trained LLaMA or Mistral context limits
- →Implement RoPE for custom model training
- →Fine-tune models for long-context tasks
About this skill
Long Context: Extending Transformer Context Windows
When to Use This Skill
Use Long Context techniques when you need to:
- Process long documents (32k, 64k, 128k+ tokens) with transformer models
- Extend context windows of pre-trained models (LLaMA, Mistral, etc.)
- Implement efficient positional encodings (RoPE, ALiBi)
- Train models with length extrapolation capabilities
- Deploy models that handle variable-length inputs efficiently
- Fine-tune existing models for longer contexts with minimal compute
Key Techniques: RoPE (Rotary Position Embeddings), YaRN, ALiBi (Attention with Linear Biases), Position Interpolation
Papers: RoFormer (arXiv 2104.09864), YaRN (arXiv 2309.00071), ALiBi (arXiv 2108.12409), Position Interpolation (arXiv 2306.15595)
Installation
# HuggingFace Transformers (includes RoPE, YaRN support)
pip install transformers torch
# For custom implementations
pip install einops # Tensor operations
pip install rotary-embedding-torch # Standalone RoPE
# Optional: FlashAttention for efficiency
pip install flash-attn --no-build-isolation
Quick Start
RoPE (Rotary Position Embeddings)
import torch
import torch.nn as nn
class RotaryEmbedding(nn.Module):
"""Rotary Position Embeddings (RoPE)."""
def __init__(self, dim, max_seq_len=8192, base=10000):
super().__init__()
# Compute inverse frequencies
inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float() / dim))
self.register_buffer("inv_freq", inv_freq)
self.max_seq_len = max_seq_len
def forward(self, seq_len, device):
# Position indices
t = torch.arange(seq_len, device=device).type_as(self.inv_freq)
# Compute frequencies
freqs = torch.outer(t, self.inv_freq) # (seq_len, dim/2)
# Compute sin and cos
emb = torch.cat((freqs, freqs), dim=-1) # (seq_len, dim)
return emb.cos(), emb.sin()
def rotate_half(x):
"""Rotate half the hidden dimensions."""
x1, x2 = x.chunk(2, dim=-1)
return torch.cat((-x2, x1), dim=-1)
def apply_rotary_pos_emb(q, k, cos, sin):
"""Apply rotary embeddings to queries and keys."""
# q, k shape: (batch, heads, seq_len, dim)
q_embed = (q * cos) + (rotate_half(q) * sin)
k_embed = (k * cos) + (rotate_half(k) * sin)
return q_embed, k_embed
# Usage
rope = RotaryEmbedding(dim=64, max_seq_len=8192)
cos, sin = rope(seq_len=2048, device='cuda')
# In attention layer
q_rotated, k_rotated = apply_rotary_pos_emb(query, key, cos, sin)
ALiBi (Attention with Linear Biases)
def get_alibi_slopes(num_heads):
"""Get ALiBi slope values for each attention head."""
def get_slopes_power_of_2(n):
start = 2 ** (-(2 ** -(math.log2(n) - 3)))
ratio = start
return [start * (ratio ** i) for i in range(n)]
if math.log2(num_heads).is_integer():
return get_slopes_power_of_2(num_heads)
else:
# Closest power of 2
closest_power = 2 ** math.floor(math.log2(num_heads))
slopes = get_slopes_power_of_2(closest_power)
# Add extra slopes
extra = get_slopes_power_of_2(2 * closest_power)
slopes.extend(extra[0::2][:num_heads - closest_power])
return slopes
def create_alibi_bias(seq_len, num_heads):
"""Create ALiBi attention bias."""
# Distance matrix
context_position = torch.arange(seq_len)
memory_position = torch.arange(seq_len)
relative_position = memory_position[None, :] - context_position[:, None]
# Get slopes
slopes = torch.tensor(get_alibi_slopes(num_heads))
# Apply slopes to distances
alibi = slopes[:, None, None] * relative_position[None, :, :]
return alibi # (num_heads, seq_len, seq_len)
# Usage in attention
num_heads = 8
seq_len = 2048
alibi_bias = create_alibi_bias(seq_len, num_heads).to('cuda')
# Add bias to attention scores
# attn_scores shape: (batch, num_heads, seq_len, seq_len)
attn_scores = attn_scores + alibi_bias
attn_weights = torch.softmax(attn_scores, dim=-1)
Position Interpolation for LLaMA
from transformers import LlamaForCausalLM, LlamaTokenizer
# Original context: 2048 tokens
model = LlamaForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf")
# Extend to 32k with position interpolation
# Modify RoPE base frequency
model.config.rope_scaling = {
"type": "linear",
"factor": 16.0 # 2048 * 16 = 32768
}
# Or use dynamic scaling
model.config.rope_scaling = {
"type": "dynamic",
"factor": 16.0
}
# Fine-tune with long documents (minimal steps needed)
# Position interpolation works out-of-the-box after this config change
Core Concepts
1. RoPE (Rotary Position Embeddings)
How it works:
- Encodes absolute position via rotation matrix
- Provides relative position dependency in attention
- Enables length extrapolation
Mathematical formulation:
q_m = (W_q * x_m) * e^(imθ)
k_n = (W_k * x_n) * e^(inθ)
where θ_j = base^(-2j/d) for j ∈ [0, d/2)
Advantages:
- Decaying inter-token dependency with distance
- Compatible with linear attention
- Better extrapolation than absolute position encodings
2. YaRN (Yet another RoPE extensioN)
Key innovation:
- NTK-aware interpolation (Neural Tangent Kernel)
- Attention temperature scaling
- Efficient context extension (10× less tokens vs baselines)
Parameters:
# YaRN configuration
yarn_config = {
"scale": 16, # Extension factor
"original_max_position": 2048, # Base context
"extrapolation_factor": 1.0, # NTK parameter
"attn_factor": 1.0, # Attention scaling
"beta_fast": 32, # High-frequency scale
"beta_slow": 1, # Low-frequency scale
}
Performance:
- Extends LLaMA to 128k tokens
- 2.5× less training steps than baselines
- State-of-the-art context window extension
3. ALiBi (Attention with Linear Biases)
Core idea:
- No positional embeddings added to tokens
- Apply distance penalty directly to attention scores
- Bias proportional to key-query distance
Formula:
attention_bias[i, j] = -m * |i - j|
where m = slope for each attention head
Advantages:
- 11% faster training vs sinusoidal embeddings
- 11% less memory usage
- Strong length extrapolation (train 1k, test 2k+)
- Inductive bias towards recency
4. Position Interpolation
Technique:
- Linearly down-scale position indices
- Interpolate within trained range (vs extrapolate beyond)
- Minimal fine-tuning required
Formula:
# Original: position indices [0, 1, 2, ..., L]
# Extended: position indices [0, 0.5, 1.0, ..., L/2]
# (for 2× extension)
scaled_position[i] = i / extension_factor
Results:
- LLaMA 7B-65B extended to 32k tokens
- 1000 fine-tuning steps sufficient
- 600× better stability than extrapolation
Method Comparison
| Method | Max Context | Training Needed | Memory | Extrapolation | Best For |
|---|---|---|---|---|---|
| RoPE | 8k-32k | Full pre-training | Moderate | Good | New models |
| YaRN | 32k-128k | Minimal (10× efficient) | Moderate | Excellent | Extending existing models |
| ALiBi | Unlimited | Full pre-training | Low (-11%) | Excellent | Training from scratch |
| Position Interpolation | 32k+ | Minimal (1k steps) | Moderate | Poor (by design) | Quick extension |
Implementation Patterns
HuggingFace Transformers Integration
from transformers import AutoModelForCausalLM, AutoConfig
# RoPE with YaRN scaling
config = AutoConfig.from_pretrained("mistralai/Mistral-7B-v0.1")
config.rope_scaling = {
"type": "yarn",
"factor": 8.0,
"original_max_position_embeddings": 8192,
"attention_factor": 1.0
}
model = AutoModelForCausalLM.from_config(config)
# Position interpolation (simpler)
config.rope_scaling = {
"type": "linear",
"factor": 4.0
}
# Dynamic scaling (adjusts based on input length)
config.rope_scaling = {
"type": "dynamic",
"factor": 8.0
}
Custom RoPE Implementation
class LongContextAttention(nn.Module):
"""Multi-head attention with RoPE."""
def __init__(self, hidden_size, num_heads, max_seq_len=32768):
super().__init__()
self.num_heads = num_heads
self.head_dim = hidden_size // num_heads
# Q, K, V projections
self.q_proj = nn.Linear(hidden_size, hidden_size)
self.k_proj = nn.Linear(hidden_size, hidden_size)
self.v_proj = nn.Linear(hidden_size, hidden_size)
self.o_proj = nn.Linear(hidden_size, hidden_size)
# RoPE
self.rotary_emb = RotaryEmbedding(
dim=self.head_dim,
max_seq_len=max_seq_len
)
def forward(self, hidden_states):
batch_size, seq_len, _ = hidden_states.shape
# Project to Q, K, V
q = self.q_proj(hidden_states)
k = self.k_proj(hidden_states)
v = self.v_proj(hidden_states)
# Reshape for multi-head
q = q.view(batch_size, seq_len, self.num_heads, self.head_dim).transpose(1, 2)
k = k.view(batch_size, seq_len, self.num_heads, self.head_dim).transpose(1, 2)
v = v.view(batch_size, seq_len, self.num_heads, self.head_dim).transpose(1, 2)
# Apply RoPE
cos, sin = self.rotary_emb(seq_len, device=hidden_states.device)
q, k = apply_rotary_pos_emb(q, k, cos, sin)
# Standard attention
attn_output = F.scaled_dot_product_attention(q, k, v)
# Reshape and project
attn_output = attn_output.transpose(1, 2).contiguous()
attn_output = attn_output.view(batch_size, seq_len, -1)
output = self.o_proj(attn_output)
return output
Fine-tuning for Long Context
Minimal Fine-tuning (Position Interpolation)
from transformers import Trainer, TrainingArguments
# Extend model config
model.config.max_position_em
---
*Content truncated.*
When not to use it
- →Scaling context without fine-tuning
- →Aggressive scaling without data
Prerequisites
Limitations
- →Requires fine-tuning for stability
- →Memory intensive
How it compares
It provides specific implementation patterns for context extension rather than just theoretical paper references.
Compared to similar skills
long-context side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| long-context (this skill) | 1 | 7mo | Review | Advanced |
| qiskit | 4 | 7mo | Review | Advanced |
| pennylane | 1 | 7mo | Review | Advanced |
| model-registry-maintainer | 0 | 8mo | Caution | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by davila7
View all by davila7 →You might also like
qiskit
davila7
Comprehensive quantum computing toolkit for building, optimizing, and executing quantum circuits. Use when working with quantum algorithms, simulations, or quantum hardware including (1) Building quantum circuits with gates and measurements, (2) Running quantum algorithms (VQE, QAOA, Grover), (3) Transpiling/optimizing circuits for hardware, (4) Executing on IBM Quantum or other providers, (5) Quantum chemistry and materials science, (6) Quantum machine learning, (7) Visualizing circuits and results, or (8) Any quantum computing development task.
pennylane
davila7
Cross-platform Python library for quantum computing, quantum machine learning, and quantum chemistry. Enables building and training quantum circuits with automatic differentiation, seamless integration with PyTorch/JAX/TensorFlow, and device-independent execution across simulators and quantum hardware (IBM, Amazon Braket, Google, Rigetti, IonQ, etc.). Use when working with quantum circuits, variational quantum algorithms (VQE, QAOA), quantum neural networks, hybrid quantum-classical models, molecular simulations, quantum chemistry calculations, or any quantum computing tasks requiring gradient-based optimization, hardware-agnostic programming, or quantum machine learning workflows.
model-registry-maintainer
massgen
Guide for maintaining the MassGen model and backend registry. This skill should be used when adding new models, updating model information (release dates, pricing, context windows), or ensuring the registry stays current with provider releases. Covers both the capabilities registry and the pricing/token manager.
hf-mcp
huggingface
Use Hugging Face Hub via MCP server tools. Search models, datasets, Spaces, papers. Get repo details, fetch documentation, run compute jobs, and use Gradio Spaces as AI tools. Available when connected to the HF MCP server.
nvalchemi-dynamics-api
NVIDIA
How to configure and run dynamics simulations, compose multi-stage pipelines (FusedStage, DistributedPipeline), use inflight batching, and manage data sinks.
quantum-espresso
Hello-QM
>