Added a build script for dolphin-2_6-phi-2

docker/scripts/build-dolphin-2_6-phi-2.sh

@@ -0,0 +1,26 @@
+#!/bin/bash -e
+
+## Note: Phi is only available in main branch and hasnt been released yet. So, make sure to build TensorRT-LLM from main branch.
+
+cd /root/TensorRT-LLM-examples/phi
+
+## Build TensorRT for [Dolphin Phi Finetuned](https://huggingface.co/cognitivecomputations/dolphin-2_6-phi-2) ChatML format with `fp16`
+
+git lfs install
+phi_path=$(huggingface-cli download --repo-type model cognitivecomputations/dolphin-2_6-phi-2)
+name=dolphin-2_6-phi-2
+python3 build.py --dtype=float16                    \
+                 --log_level=verbose                \
+                 --use_gpt_attention_plugin float16 \
+                 --use_gemm_plugin float16          \
+                 --max_batch_size=1                 \
+                 --max_input_len=1024               \
+                 --max_output_len=1024              \
+                 --output_dir=$name                 \
+                 --model_dir="$phi_path" >&1 | tee build.log
+
+dest=/root/scratch-space/models
+mkdir -p "$dest/$name/tokenizer"
+cp -r "$name" "$dest"
+(cd "$phi_path" && cp config.json tokenizer_config.json vocab.json merges.txt "$dest/$name/tokenizer")
+cp -r "$phi_path" "$dest/phi-orig-model"

docker/scripts/build-models.sh

@@ -4,4 +4,4 @@ test -f /etc/shinit_v2 && source /etc/shinit_v2
 
 ./build-whisper.sh
 # ./build-mistral.sh
-./build-phi-2.sh
+./build-dolphin-2_6-phi-2.sh

docker/scripts/run-whisperfusion.sh

@@ -6,8 +6,8 @@ cd WhisperFusion
 if [ "$1" != "mistral" ]; then
   exec python3 main.py --phi \
                   --whisper_tensorrt_path /root/whisper_small_en \
-                  --phi_tensorrt_path /root/phi-2 \
-                  --phi_tokenizer_path /root/phi-2
+                  --phi_tensorrt_path /root/dolphin-2_6-phi-2 \
+                  --phi_tokenizer_path /root/dolphin-2_6-phi-2/tokenizer
 else
   exec python3 main.py --mistral \
                   --whisper_tensorrt_path /root/models/whisper_small_en \