metadata

language:
  - en
license: other
license_name: llama3
model_name: Llama3 70B Instruct
arxiv: 2307.09288
base_model: meta-llama/Meta-Llama-3-70B-Instruct
inference: false
model_creator: Meta Llama3
model_type: llama
pipeline_tag: text-generation
quantized_by: Second State Inc.

Meta-Llama-3-70B-Instruct-GGUF

Original Model

meta-llama/Meta-Llama-3-70B-Instruct

Run with LlamaEdge

LlamaEdge version: v0.8.3 and above

Prompt template

Prompt type: llama-3-chat

Prompt string

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

{{ system_prompt }}<|eot_id|><|start_header_id|>user<|end_header_id|>

{{ user_message_1 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

{{ model_answer_1 }}<|eot_id|><|start_header_id|>user<|end_header_id|>

{{ user_message_2 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

Context size: 8192

Run as LlamaEdge service

wasmedge --dir .:. --nn-preload default:GGML:AUTO:Meta-Llama-3-70B-Instruct-Q5_K_M.gguf \
  llama-api-server.wasm \
  --prompt-template llama-3-chat \
  --ctx-size 8192 \
  --model-name Llama-3-70b

Run as LlamaEdge command app

wasmedge --dir .:. --nn-preload default:GGML:AUTO:Meta-Llama-3-70B-Instruct-Q5_K_M.gguf \
  llama-chat.wasm \
  --prompt-template llama-3-chat \
  --ctx-size 8192

Quantized GGUF Models

Name	Quant method	Bits	Size	Use case
Meta-Llama-3-70B-Instruct-Q2_K.gguf	Q2_K	2	26.4 GB	smallest, significant quality loss - not recommended for most purposes
Meta-Llama-3-70B-Instruct-Q3_K_L.gguf	Q3_K_L	3	37.1 GB	small, substantial quality loss
Meta-Llama-3-70B-Instruct-Q3_K_M.gguf	Q3_K_M	3	34.3 GB	very small, high quality loss
Meta-Llama-3-70B-Instruct-Q3_K_S.gguf	Q3_K_S	3	30.9 GB	very small, high quality loss
Meta-Llama-3-70B-Instruct-Q4_0.gguf	Q4_0	4	40 GB	legacy; small, very high quality loss - prefer using Q3_K_M
Meta-Llama-3-70B-Instruct-Q4_K_M.gguf	Q4_K_M	4	42.5 GB	medium, balanced quality - recommended
Meta-Llama-3-70B-Instruct-Q5_0.gguf	Q5_0	5	48.7 GB	legacy; medium, balanced quality - prefer using Q4_K_M
Meta-Llama-3-70B-Instruct-Q5_K_M.gguf	Q5_K_M	5	50 GB	large, very low quality loss - recommended
Meta-Llama-3-70B-Instruct-Q5_K_S.gguf	Q5_K_S	5	48.7 GB	large, low quality loss - recommended
Meta-Llama-3-70B-Instruct-Q6_K-00001-of-00002.gguf	Q6_K	6	32.1 GB	very large, extremely low quality loss
Meta-Llama-3-70B-Instruct-Q6_K-00002-of-00002.gguf	Q6_K	6	25.7 GB	very large, extremely low quality loss
Meta-Llama-3-70B-Instruct-Q8_0-00001-of-00003.gguf	Q8_0	8	32 GB	very large, extremely low quality loss - not recommended
Meta-Llama-3-70B-Instruct-Q8_0-00002-of-00003.gguf	Q8_0	8	32.1 GB	very large, extremely low quality loss - not recommended
Meta-Llama-3-70B-Instruct-Q8_0-00003-of-00003.gguf	Q8_0	8	10.9 GB	very large, extremely low quality loss - not recommended

The f16 GGUF model for the original model can be found in second-state/Meta-Llama-3-70B-Instruct-f16-GGUF

Quantized with llama.cpp b2715.