[CI] Fixes too many TOKENIZER in nightly bench llama (#1858)

deepjavalibrary · May 2, 2024 · e4221ba · e4221ba
1 parent 068d929
commit e4221ba
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 3 deletions.
diff --git a/tests/integration/benchmark/nightly/g5-12xl.txt b/tests/integration/benchmark/nightly/g5-12xl.txt
@@ -10,7 +10,7 @@ option.rolling_batch=$ENGINE
 option.model_id=s3://djl-llm/llama-2-7b-hf/
 option.tensor_parallel_degree=max
 [aws_curl]
-TOKENIZER=TOKENIZER=TheBloke/Llama-2-7B-fp16 ./awscurl -c 32 -N 10 \
+TOKENIZER=TheBloke/Llama-2-7B-fp16 ./awscurl -c 32 -N 10 \
 -X POST http://127.0.0.1:8080/invocations   \
 --connect-timeout 60   -H "Content-type: application/json"   \
 -d '{"inputs":"The new movie that got Oscar this year","parameters":{"max_new_tokens":256, "do_sample":true}}'   \
@@ -27,7 +27,7 @@ option.rolling_batch=$ENGINE
 option.model_id=s3://djl-llm/llama-3-8b-hf/
 option.tensor_parallel_degree=max
 [aws_curl]
-TOKENIZER=TOKENIZER=TheBloke/Llama-2-13B-fp16 ./awscurl -c 32 -N 10 \
+TOKENIZER=TheBloke/Llama-2-13B-fp16 ./awscurl -c 32 -N 10 \
 -X POST http://127.0.0.1:8080/invocations   \
 --connect-timeout 60   -H "Content-type: application/json"   \
 -d '{"inputs":"The new movie that got Oscar this year","parameters":{"max_new_tokens":256, "do_sample":true}}'   \

diff --git a/tests/integration/benchmark/nightly/g5-2xl.txt b/tests/integration/benchmark/nightly/g5-2xl.txt
@@ -11,7 +11,7 @@ option.model_id=NousResearch/Hermes-2-Pro-Mistral-7B
 option.tensor_parallel_degree=max
 option.max_model_len=8192
 [aws_curl]
-TOKENIZER=NousResearch/Hermes-2-Pro-Mistral-7B ./awscurl -c 32 -N 10 \
+TOKENIZER=TheBloke/Mistral-7B-Instruct-v0.2-AWQ ./awscurl -c 32 -N 10 \
 -X POST http://127.0.0.1:8080/invocations   \
 --connect-timeout 60   -H "Content-type: application/json"   \
 -d '{"inputs":"The new movie that got Oscar this year","parameters":{"max_new_tokens":256, "do_sample":true}}'   \