Add benchmark results (#76)

Author: qihqi

README.md

@@ -112,7 +112,7 @@ go to the deps/JetStream folder (downloaded during `install_everything.sh`)
 cd deps/JetStream
 wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
 export dataset_path=ShareGPT_V3_unfiltered_cleaned_split.json
-python benchmarks/benchmark_serving.py --tokenizer $tokenizer_path --num-prompts 2000  --dataset-path  $dataset_path --dataset sharegpt --save-request-outputs
+python benchmarks/benchmark_serving.py --tokenizer $tokenizer_path --num-prompts 2000  --dataset-path  $dataset_path --dataset sharegpt --save-request-outputs --warm-up=True
 ```
 Please look at `deps/JetStream/benchmarks/README.md` for more information.
 

benchmarks/summary.md

@@ -0,0 +1,64 @@
+# Benchmark results of various models
+
+
+## Llama 3 - 8B
+
+Date | Device  | dtype | batch size | cache length |max input length |max output length| throughput (token/s) 
+----| ------- | ------ |---------- | -------------|-----------------|------------------|----------------------
+2024-04-24 | TPU v5e-8 | bfloat16 | 128 | 2048 | 1024 | 1024 | 8249 
+2024-04-24 | TPU v5e-8 | int8 | 256 | 2048 | 1024 | 1024 | 10873
+
+
+## Gemma - 7B
+
+Date | Device  | dtype | batch size | cache length |max input length |max output length| throughput (token/s) 
+----| ------- | ------ |---------- | -------------|-----------------|------------------|----------------------
+2024-05-10 | TPU v5e-8 | bfloat16 | 96 | 2048 | 1024 | 1024 | 3236
+2024-05-10 | TPU v5e-8 | int8 | 128 | 2048 | 1024 | 1024 | 4695
+
+## Llama 2 - 7B
+
+Date | Device  | dtype | batch size | cache length |max input length |max output length| throughput (token/s) 
+----| ------- | ------ |---------- | -------------|-----------------|------------------|----------------------
+2024-03-28 | TPU v5e-8 | bfloat16 | 96 | 2048 | 1024 | 1024 | 3663
+2024-03-28 | TPU v5e-8 | int8 | 96 | 2048 | 1024 | 1024 | 4783 
+
+## Llama 2 - 13B
+
+Date | Device  | dtype | batch size | cache length |max input length |max output length| throughput (token/s) 
+----| ------- | ------ |---------- | -------------|-----------------|------------------|----------------------
+2024-03-28 | TPU v5e-8 | bfloat16 | 48 | 2048 | 1024 | 1024 | 2056
+2024-03-28 | TPU v5e-8 | int8 | 96 | 2048 | 1024 | 1024 | 3458 
+2024-03-28 | TPU v5e-8 | bfloat16 | 80 | 1280 | 1024 | 1024 | 2911
+2024-03-28 | TPU v5e-8 | int8 | 96 | 1280 | 1024 | 1024 | 3938
+
+**NOTE:** When cache length is less than the sum of max input length + max output length
+  we employ *Rolling window attention*. 
+
+
+# Instructions to reproduce:
+
+Please refer [README.md](README.md) for instructions in how to get the model weights.
+
+**NOTE** Different weights can produce different benchmark results (due to generating)
+different sentence length. For llama, we used the `-chat` versions of the weight.
+For Gemma we used the `-it` (instruction finetuned) version of the weights.
+
+## Run the server
+NOTE: the `--platform=tpu=8` need to specify number of tpu devices (which is 4 for v4-8 and 8 for v5light-8`)
+
+```bash
+python run_server.py --param_size=7b --batch_size= 128 --max_cache_length=2048 --quantize_weights=$quantize --quantize_kv_cache=$quantize --checkpoint_path=$output_ckpt_dir   --tokenizer_path=$tokenizer_path --platform=tpu=8 --model=$model_name
+```
+Now you can fire gRPC to it
+
+# Run benchmark
+go to the deps/JetStream folder (downloaded during `install_everything.sh`)
+
+```bash
+cd deps/JetStream
+wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
+export dataset_path=ShareGPT_V3_unfiltered_cleaned_split.json
+python benchmarks/benchmark_serving.py --tokenizer $tokenizer_path --num-prompts 2000  --dataset-path  $dataset_path --dataset sharegpt --save-request-outputs --warm-up=True
+```
+Please look at `deps/JetStream/benchmarks/README.md` for more information.