From b76885049f4c910c5eaa11b1207e4f147d0aab65 Mon Sep 17 00:00:00 2001 From: Mergen Nachin Date: Wed, 24 Apr 2024 18:08:10 -0700 Subject: [PATCH] Update llama2 readme file - main branch (#3340) Summary: Pull Request resolved: https://github.com/pytorch/executorch/pull/3340 Reviewed By: orionr, kimishpatel, cccclai Differential Revision: D56553088 Pulled By: mergennachin --- examples/models/llama2/README.md | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-) diff --git a/examples/models/llama2/README.md b/examples/models/llama2/README.md index 7284acfcec6..f3c3951b4cb 100644 --- a/examples/models/llama2/README.md +++ b/examples/models/llama2/README.md @@ -33,21 +33,22 @@ We evaluated WikiText perplexity using [LM Eval](https://github.com/EleutherAI/l Note that groupsize less than 128 was not enabled, since such model were still too large. This is because our current efforts have focused on enabling FP32 and support for FP16 is under way. What this implies for model size is that 1) embedding table is in FP32 and 2) quantized weights scales are FP32. +## Enablement + +We have verified running Llama 2 7B [mobile applications](#step-6-build-mobile-apps) efficiently on select devices including the iPhone 15 Pro, iPhone 15 Pro Max, Samsung Galaxy S22 and S24, and OnePlus 12. + +For Llama 3 8B, we have verified so far on iPhone 15 Pro Max and OnePlus 12 (with 16GB RAM). + ## Performance -Performance was measured on Samsung Galaxy S22, S24, One Plus 12 and iPhone 15 max Pro. Measurement performance is in terms of tokens/second. +Llama2 7B performance was measured on the Samsung Galaxy S22, S24, and OnePlus 12 devices. The performance measurement is expressed in terms of tokens per second using an [adb binary-based approach](#step-5-run-benchmark-on). |Device | Groupwise 4-bit (128) | Groupwise 4-bit (256) |--------| ---------------------- | --------------- -|Galaxy S22* | 8.15 tokens/second | 8.3 tokens/second | -|Galaxy S24* | 10.66 tokens/second | 11.26 tokens/second | -|One plus 12* | 11.55 tokens/second | 11.6 tokens/second | -|Galaxy S22** | 5.5 tokens/second | 5.9 tokens/second | -|iPhone 15 pro** | ~6 tokens/second | ~6 tokens/second | - -*: Measured via adb binary based [workflow](#step-5-run-benchmark-on) +|Galaxy S22 | 8.15 tokens/second | 8.3 tokens/second | +|Galaxy S24 | 10.66 tokens/second | 11.26 tokens/second | +|OnePlus 12 | 11.55 tokens/second | 11.6 tokens/second | -**: Measured via app based [workflow](#step-6-build-mobile-apps) # Instructions