Skip to content

Commit 49074a0

Browse files
fightfatfightfat
authored
Fix benchmark paths to use relative paths (#4740)
1 parent 88a1e6c commit 49074a0

19 files changed

Lines changed: 54 additions & 54 deletions

‎tests/config/benchmark/config/pt/DeepSeek-V3.yaml‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -2,7 +2,7 @@
22
stage: dsv3_pretrain
33

44
### data
5-
input_dir: /root/efficient_benchmark/dataset/DeepSeek-V3
5+
input_dir: ./dataset/DeepSeek-V3
66
max_seq_len: 8192
77
dataloader_num_workers: 8
88
ignore_data_skip: 0
@@ -11,7 +11,7 @@ pad_token_id: 1
1111

1212
### model
1313
model_name_or_path: ./config/
14-
tokenizer_name_or_path: /root/efficient_benchmark/huggingface/DeepSeek-V3
14+
tokenizer_name_or_path: ./huggingface/DeepSeek-V3
1515
use_expert_parallel: true
1616

1717
### train

‎tests/config/benchmark/config/pt/DeepSeek-V4.yaml‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,9 @@
11
### data
22
train_dataset_type: erniekit
33
eval_dataset_type: erniekit
4-
train_dataset_path: /root/efficient_benchmark/dataset/DeepSeek-V4/reasoning_decay.jsonl
4+
train_dataset_path: ./dataset/DeepSeek-V4/reasoning_decay.jsonl
55
train_dataset_prob: "1.0"
6-
eval_dataset_path: /root/efficient_benchmark/dataset/DeepSeek-V4/reasoning_decay.jsonl
6+
eval_dataset_path: ./dataset/DeepSeek-V4/reasoning_decay.jsonl
77
eval_dataset_prob: "1.0"
88
packing: true
99
truncate_packing: true
@@ -16,7 +16,7 @@ use_template: false
1616
dataset_num_proc: 8
1717

1818
### model
19-
model_name_or_path: /root/efficient_benchmark/huggingface/DeepSeek-V4-Flash-Base
19+
model_name_or_path: ./huggingface/DeepSeek-V4-Flash-Base
2020

2121
### train
2222
stage: PT

‎tests/config/benchmark/config/pt/ERNIE45-21B.yaml‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,7 @@ stage: ernie_pretrain
33

44
# data
55
dataset_type: pretrain
6-
input_dir: "1.0 /root/efficient_benchmark/dataset/eb45_industrycorpus2_94k/eb45_industrycorpus2_94k"
6+
input_dir: "1.0 ./dataset/eb45_industrycorpus2_94k/eb45_industrycorpus2_94k"
77
split: "998,1,1"
88
max_seq_len: 8192
99
dataloader_num_workers: 8
@@ -13,7 +13,7 @@ ignore_data_skip: 0
1313
pad_token_id: 0
1414

1515
# model
16-
model_name_or_path: /root/efficient_benchmark/huggingface/ERNIE-4.5-21B-A3B-Base-PT/
16+
model_name_or_path: ./huggingface/ERNIE-4.5-21B-A3B-Base-PT/
1717
tokenizer_name_or_path: PaddleFormers/examples/experiments/ernie_pretrain/ernie/src/tokenizers/tokenizer_model
1818
moe_router_bias_update_rate: 0.001
1919
moe_with_send_router_loss: False

‎tests/config/benchmark/config/pt/ERNIE45-300B.yaml‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,7 @@ stage: ernie_pretrain
33

44
# data
55
dataset_type: pretrain
6-
input_dir: "1.0 /root/efficient_benchmark/dataset/eb45_industrycorpus2_94k/eb45_industrycorpus2_94k"
6+
input_dir: "1.0 ./dataset/eb45_industrycorpus2_94k/eb45_industrycorpus2_94k"
77
split: "998,1,1"
88
data_load_process_num: 40
99
max_seq_len: 8192
@@ -15,7 +15,7 @@ ignore_data_skip: 0
1515
pad_token_id: 0
1616

1717
# model
18-
model_name_or_path: /root/efficient_benchmark/huggingface/ERNIE-4.5-300B-A47B/
18+
model_name_or_path: ./huggingface/ERNIE-4.5-300B-A47B/
1919
tokenizer_name_or_path: PaddleFormers/examples/experiments/ernie_pretrain/ernie/src/tokenizers/tokenizer_model
2020
moe_router_bias_update_rate: 0.001
2121
moe_with_send_router_loss: False

‎tests/config/benchmark/config/pt/GLM4.5-Air.yaml‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,9 @@
11
### data
22
train_dataset_type: erniekit
33
eval_dataset_type: erniekit
4-
train_dataset_path: ./data/pt/train.jsonl
4+
train_dataset_path: ./dataset/data/pt_data/train.jsonl
55
train_dataset_prob: "1.0"
6-
eval_dataset_path: ./data/pt/eval.jsonl
6+
eval_dataset_path: ./dataset/data/pt_data/eval.jsonl
77
eval_dataset_prob: "1.0"
88
packing: true
99
truncate_packing: true
@@ -16,7 +16,7 @@ prefetch_factor: 24
1616
pad_token_id: 151329
1717

1818
### model
19-
model_name_or_path: /root/efficient_benchmark/huggingface/GLM-4.5-Air
19+
model_name_or_path: ./huggingface/GLM-4.5-Air
2020
_attn_implementation: flashmask
2121

2222
### finetuning

‎tests/config/benchmark/config/pt/GLM4.5-Air_128k.yaml‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,8 +1,8 @@
11
train_dataset_type: messages
22
eval_dataset_type: messages
3-
train_dataset_path: /root/efficient_benchmark/dataset/data_60to64k/LVEval_60to64k_messages.jsonl
3+
train_dataset_path: ./dataset/data_60to64k/LVEval_60to64k_messages.jsonl
44
train_dataset_prob: "1.0"
5-
eval_dataset_path: /root/efficient_benchmark/dataset/data_60to64k/LVEval_60to64k_messages.jsonl
5+
eval_dataset_path: ./dataset/data_60to64k/LVEval_60to64k_messages.jsonl
66
eval_dataset_prob: "1.0"
77

88
packing: true
@@ -14,7 +14,7 @@ random_shuffle: true
1414
dataloader_shuffle: false
1515
use_template: false
1616

17-
model_name_or_path: /root/efficient_benchmark/huggingface/GLM-4.5-Air
17+
model_name_or_path: ./huggingface/GLM-4.5-Air
1818

1919
# base
2020
stage: PT

‎tests/config/benchmark/config/pt/GLM4.5-Air_64k.yaml‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,8 +1,8 @@
11
train_dataset_type: messages
22
eval_dataset_type: messages
3-
train_dataset_path: /root/efficient_benchmark/dataset/data_60to64k/LVEval_60to64k_messages.jsonl
3+
train_dataset_path: ./dataset/data_60to64k/LVEval_60to64k_messages.jsonl
44
train_dataset_prob: "1.0"
5-
eval_dataset_path: /root/efficient_benchmark/dataset/data_60to64k/LVEval_60to64k_messages.jsonl
5+
eval_dataset_path: ./dataset/data_60to64k/LVEval_60to64k_messages.jsonl
66
eval_dataset_prob: "1.0"
77

88
max_seq_len: 65536
@@ -16,7 +16,7 @@ padding_free: true
1616
pad_token_id: 151329
1717

1818
### model
19-
model_name_or_path: /root/efficient_benchmark/huggingface/GLM-4.5-Air
19+
model_name_or_path: ./huggingface/GLM-4.5-Air
2020
_attn_implementation: flashmask
2121

2222
### finetuning

‎tests/config/benchmark/config/pt/GLM4.5-Air_FP8.yaml‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,9 @@
11
### data
22
train_dataset_type: erniekit
33
eval_dataset_type: erniekit
4-
train_dataset_path: ./data/pt/train.jsonl
4+
train_dataset_path: ./dataset/data/pt_data/train.jsonl
55
train_dataset_prob: "1.0"
6-
eval_dataset_path: ./data/pt/eval.jsonl
6+
eval_dataset_path: ./dataset/data/pt_data/eval.jsonl
77
eval_dataset_prob: "1.0"
88
packing: true
99
truncate_packing: true
@@ -16,7 +16,7 @@ prefetch_factor: 24
1616
pad_token_id: 151329
1717

1818
### model
19-
model_name_or_path: /root/efficient_benchmark/huggingface/GLM-4.5-Air
19+
model_name_or_path: ./huggingface/GLM-4.5-Air
2020
_attn_implementation: flashmask
2121

2222
### finetuning

‎tests/config/benchmark/config/pt/Qwen3-30B-A3B-Base-64k.yaml‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,9 @@
11
### data
22
train_dataset_type: messages
33
eval_dataset_type: messages
4-
train_dataset_path: /root/efficient_benchmark/dataset/data_60to64k/LVEval_60to64k_messages.jsonl
4+
train_dataset_path: ./dataset/data_60to64k/LVEval_60to64k_messages.jsonl
55
train_dataset_prob: "1.0"
6-
eval_dataset_path: /root/efficient_benchmark/dataset/data_60to64k/LVEval_60to64k_messages.jsonl
6+
eval_dataset_path: ./dataset/data_60to64k/LVEval_60to64k_messages.jsonl
77
eval_dataset_prob: "1.0"
88

99
eval_iters: 10
@@ -23,7 +23,7 @@ dataloader_num_workers: 24
2323
prefetch_factor: 24
2424

2525
### model
26-
model_name_or_path: /root/efficient_benchmark/huggingface/Qwen/Qwen3-30B-A3B
26+
model_name_or_path: ./huggingface/Qwen/Qwen3-30B-A3B
2727
_attn_implementation: flashmask
2828
use_qk_norm: true
2929

‎tests/config/benchmark/config/pt/Qwen3-30B-A3B-Base.yaml‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,9 @@
11
## data
22
train_dataset_type: erniekit
33
eval_dataset_type: erniekit
4-
train_dataset_path: ./data/pt/train.jsonl
4+
train_dataset_path: ./dataset/data/pt_data/train.jsonl
55
train_dataset_prob: "1.0"
6-
eval_dataset_path: ./data/pt/eval.jsonl
6+
eval_dataset_path: ./dataset/data/pt_data/eval.jsonl
77
eval_dataset_prob: "1.0"
88

99
# dataset_type: "pretrain"
@@ -25,7 +25,7 @@ dataloader_num_workers: 24
2525
prefetch_factor: 24
2626

2727
### model
28-
model_name_or_path: /root/efficient_benchmark/huggingface/Qwen/Qwen3-30B-A3B
28+
model_name_or_path: ./huggingface/Qwen/Qwen3-30B-A3B
2929
_attn_implementation: flashmask
3030
use_qk_norm: true
3131

0 commit comments

Comments
 (0)