← All Models | EleutherAI - Non-profit AI research lab focused on open-source LLMs:

EleutherAI/gpt-neo-2.7B

📊 Model Parameters

Total Parameters 2,779,965,440
Context Length 2,048
Hidden Size 2560
Layers 32
Attention Heads 20
KV Heads 20

💾 Memory Requirements

FP32 (Full) 10.36 GB
FP16 (Half) 5.18 GB
INT8 (Quantized) 2.59 GB
INT4 (Quantized) 1.29 GB

🔑 KV Cache (Inference)

Per Token (FP16) 327.68 KB
Max Context FP32 1.25 GB
Max Context FP16 640.0 MB
Max Context INT8 320.0 MB

⚙️ Model Configuration

Core Architecture

Vocabulary Size50,257
Hidden Size2,560
Number of Layers32
Attention Heads20
FFN Intermediate SizeNot set

Context & Position

Max Context Length2,048
Sliding Window Size256

Attention Configuration

Attention Dropout0%
Tied EmbeddingsYes

Activation & Normalization

Activation Functiongelu_new
RMSNorm Epsilon1e-05

Dropout (Training)

Residual Dropout0%
Embedding Dropout0%

Special Tokens

BOS Token ID50,256
EOS Token ID50256
Pad Token IDNot set

Data Type

Model DtypeNot set
Layer Types:
Attention
MLP/FFN
Normalization
Embedding