Skip to content

Commit 26f3206

Browse files
committed
Fix max new tokens
1 parent ff1133a commit 26f3206

8 files changed

+8
-8
lines changed

recipes/A5000_24GB_x8/fake-news-detector-en-1.5T.yaml

+1-1
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,7 @@ train_gradient_accumulation_steps: 2
1919
train_num_train_epochs: 4
2020
train_max_steps: 1000
2121
train_fp16: True
22-
inference_max_new_tokens: 2
22+
inference_max_new_tokens: 8
2323
evaluations:
2424
-
2525
prompt: "Donald Trump has never been President of the United States."

recipes/A5000_24GB_x8/fake-news-detector-en.yaml

+1-1
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,7 @@ train_gradient_accumulation_steps: 2
1919
train_num_train_epochs: 4
2020
train_max_steps: 1000
2121
train_fp16: True
22-
inference_max_new_tokens: 2
22+
inference_max_new_tokens: 8
2323
evaluations:
2424
-
2525
prompt: "Donald Trump has never been President of the United States."

recipes/A5000_24GB_x8/hate-speech-detector-en-1.5T.yaml

+1-1
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,7 @@ train_gradient_accumulation_steps: 4
1919
train_num_train_epochs: 4
2020
train_max_steps: 1000
2121
train_fp16: True
22-
inference_max_new_tokens: 2
22+
inference_max_new_tokens: 16
2323
evaluations:
2424
-
2525
prompt: "the white establishment can't have blk folx running around loving themselves and promoting our greatness"

recipes/A5000_24GB_x8/hate-speech-detector-en.yaml

+1-1
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,7 @@ train_gradient_accumulation_steps: 4
1919
train_num_train_epochs: 4
2020
train_max_steps: 1000
2121
train_fp16: True
22-
inference_max_new_tokens: 2
22+
inference_max_new_tokens: 16
2323
evaluations:
2424
-
2525
prompt: "the white establishment can't have blk folx running around loving themselves and promoting our greatness"

recipes/A5000_24GB_x8/karasu-sentiment-analyzer-ja.yaml

+1-1
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,7 @@ train_gradient_accumulation_steps: 2
1919
train_num_train_epochs: 4
2020
train_max_steps: 1000
2121
train_fp16: True
22-
inference_max_new_tokens: 2
22+
inference_max_new_tokens: 16
2323
evaluations:
2424
-
2525
prompt: "ぼけっとしてたらこんな時間。チャリあるから食べにでたいのに…"

recipes/A5000_24GB_x8/sentiment-analyzer-en.yaml

+1-1
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,7 @@ train_gradient_accumulation_steps: 2
1919
train_num_train_epochs: 4
2020
train_max_steps: 200
2121
train_fp16: True
22-
inference_max_new_tokens: 2
22+
inference_max_new_tokens: 16
2323
evaluations:
2424
-
2525
prompt: "Yes my laptop works So now i can abort my diplomthesis"

recipes/A5000_24GB_x8/sentiment-analyzer-ja.yaml

+1-1
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,7 @@ train_gradient_accumulation_steps: 2
1919
train_num_train_epochs: 4
2020
train_max_steps: 1000
2121
train_fp16: True
22-
inference_max_new_tokens: 2
22+
inference_max_new_tokens: 16
2323
evaluations:
2424
-
2525
prompt: "ぼけっとしてたらこんな時間。チャリあるから食べにでたいのに…"

recipes/A5000_24GB_x8/sentiment-analyzer-multilingual.yaml

+1-1
Original file line numberDiff line numberDiff line change
@@ -21,7 +21,7 @@ train_gradient_accumulation_steps: 2
2121
train_num_train_epochs: 4
2222
train_max_steps: 1000
2323
train_fp16: True
24-
inference_max_new_tokens: 2
24+
inference_max_new_tokens: 16
2525
evaluations:
2626
-
2727
prompt: "Yes my laptop works So now i can abort my diplomthesis"

0 commit comments

Comments
 (0)