Donghao Huang
commited on
Commit
·
43e183b
1
Parent(s):
b05a046
completed tgi.sh
Browse files- .env.example +10 -0
- tgi.sh +3 -4
.env.example
CHANGED
|
@@ -91,3 +91,13 @@ CHUNK_OVERLAP=512
|
|
| 91 |
# telegram bot
|
| 92 |
TELEGRAM_API_TOKEN=
|
| 93 |
CHAT_API_URL=http://localhost:8080/chat_sync
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 91 |
# telegram bot
|
| 92 |
TELEGRAM_API_TOKEN=
|
| 93 |
CHAT_API_URL=http://localhost:8080/chat_sync
|
| 94 |
+
|
| 95 |
+
# template for env/tgi.conf
|
| 96 |
+
export PORT=64300
|
| 97 |
+
|
| 98 |
+
export NGROK_AUTHTOKEN=
|
| 99 |
+
export NGROK_EDGE=
|
| 100 |
+
|
| 101 |
+
export HUGGINGFACE_HUB_CACHE=$HOME/.cache/huggingface/hub/
|
| 102 |
+
export HUGGING_FACE_HUB_TOKEN=
|
| 103 |
+
ß
|
tgi.sh
CHANGED
|
@@ -7,12 +7,11 @@ pwd
|
|
| 7 |
|
| 8 |
uname -a
|
| 9 |
|
| 10 |
-
|
| 11 |
-
souce env/ngrok.conf
|
| 12 |
|
| 13 |
export MODEL_ID="meta-llama/Llama-2-7b-chat-hf"
|
| 14 |
-
export QUANTIZE
|
| 15 |
|
| 16 |
echo Running $MODEL_ID with TGI
|
| 17 |
|
| 18 |
-
text-generation-launcher --model-id $MODEL_ID --port $PORT --max-input-length 2048 --max-total-tokens 4096 --ngrok --ngrok-authtoken $NGROK_AUTHTOKEN --ngrok-edge NGROK_EDGE $QUANTIZE
|
|
|
|
| 7 |
|
| 8 |
uname -a
|
| 9 |
|
| 10 |
+
. env/tgi.conf
|
|
|
|
| 11 |
|
| 12 |
export MODEL_ID="meta-llama/Llama-2-7b-chat-hf"
|
| 13 |
+
export QUANTIZE="--quantize bitsandbytes-fp4"
|
| 14 |
|
| 15 |
echo Running $MODEL_ID with TGI
|
| 16 |
|
| 17 |
+
text-generation-launcher --model-id $MODEL_ID --port $PORT --max-input-length 2048 --max-total-tokens 4096 --ngrok --ngrok-authtoken $NGROK_AUTHTOKEN --ngrok-edge $NGROK_EDGE $QUANTIZE
|