Delete eval_llama3.sh
Browse files- eval_llama3.sh +0 -17
eval_llama3.sh
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
# MODEL=deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
|
| 2 |
-
# ValueError: User-specified max_model_len (32768) is greater than the derived max_model_len (max_position_embeddings=2048 or model_max_length=None in model's config.json). This may lead
|
| 3 |
-
# to incorrect model outputs or CUDA errors. To allow overriding this maximum, set the env var VLLM_ALLOW_LONG_MAX_MODEL_LEN=1
|
| 4 |
-
# Only needed for MicroLlama V1
|
| 5 |
-
# export VLLM_ALLOW_LONG_MAX_MODEL_LEN=1
|
| 6 |
-
|
| 7 |
-
NUM_GPUS=4
|
| 8 |
-
MODEL="/root/open-r1/data/meta-llama/Llama-3.2-1B-Instruct/checkpoint-900"
|
| 9 |
-
MODEL_ARGS="pretrained=$MODEL,dtype=float16,data_parallel_size=$NUM_GPUS,max_model_length=32768,gpu_memory_utilisation=0.8"
|
| 10 |
-
TASK=aime24
|
| 11 |
-
OUTPUT_DIR=data/evals/$MODEL
|
| 12 |
-
|
| 13 |
-
lighteval vllm $MODEL_ARGS "custom|$TASK|0|0" \
|
| 14 |
-
--custom-tasks src/open_r1/evaluate.py \
|
| 15 |
-
--use-chat-template \
|
| 16 |
-
--system-prompt="Please reason step by step, and put your final answer within \boxed{}." \
|
| 17 |
-
--output-dir $OUTPUT_DIR
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|