picocreator
commited on
Commit
•
a650975
1
Parent(s):
3ee0475
Upload 2 files
Browse files- Eagle-7B-1T-ac/results.json +70 -0
- Eagle-7B-1T-hs/results.json +68 -0
Eagle-7B-1T-ac/results.json
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"results": {
|
3 |
+
"arc_challenge": {
|
4 |
+
"acc,none": 0.42918088737201365,
|
5 |
+
"acc_stderr,none": 0.014464085894870677,
|
6 |
+
"acc_norm,none": 0.4761092150170648,
|
7 |
+
"acc_norm_stderr,none": 0.014594701798071666,
|
8 |
+
"alias": "arc_challenge"
|
9 |
+
}
|
10 |
+
},
|
11 |
+
"configs": {
|
12 |
+
"arc_challenge": {
|
13 |
+
"task": "arc_challenge",
|
14 |
+
"group": [
|
15 |
+
"ai2_arc"
|
16 |
+
],
|
17 |
+
"dataset_path": "allenai/ai2_arc",
|
18 |
+
"dataset_name": "ARC-Challenge",
|
19 |
+
"training_split": "train",
|
20 |
+
"validation_split": "validation",
|
21 |
+
"test_split": "test",
|
22 |
+
"doc_to_text": "Question: {{question}}\nAnswer:",
|
23 |
+
"doc_to_target": "{{choices.label.index(answerKey)}}",
|
24 |
+
"doc_to_choice": "{{choices.text}}",
|
25 |
+
"description": "",
|
26 |
+
"target_delimiter": " ",
|
27 |
+
"fewshot_delimiter": "\n\n",
|
28 |
+
"num_fewshot": 25,
|
29 |
+
"metric_list": [
|
30 |
+
{
|
31 |
+
"metric": "acc",
|
32 |
+
"aggregation": "mean",
|
33 |
+
"higher_is_better": true
|
34 |
+
},
|
35 |
+
{
|
36 |
+
"metric": "acc_norm",
|
37 |
+
"aggregation": "mean",
|
38 |
+
"higher_is_better": true
|
39 |
+
}
|
40 |
+
],
|
41 |
+
"output_type": "multiple_choice",
|
42 |
+
"repeats": 1,
|
43 |
+
"should_decontaminate": true,
|
44 |
+
"doc_to_decontamination_query": "Question: {{question}}\nAnswer:",
|
45 |
+
"metadata": {
|
46 |
+
"version": 1.0
|
47 |
+
}
|
48 |
+
}
|
49 |
+
},
|
50 |
+
"versions": {
|
51 |
+
"arc_challenge": 1.0
|
52 |
+
},
|
53 |
+
"n-shot": {
|
54 |
+
"arc_challenge": 25
|
55 |
+
},
|
56 |
+
"config": {
|
57 |
+
"model": "hf",
|
58 |
+
"model_args": "pretrained=RWKV/rwkv-5-world-7b,dtype=float16,trust_remote_code=True",
|
59 |
+
"batch_size": "auto",
|
60 |
+
"batch_sizes": [
|
61 |
+
8
|
62 |
+
],
|
63 |
+
"device": null,
|
64 |
+
"use_cache": null,
|
65 |
+
"limit": null,
|
66 |
+
"bootstrap_iters": 100000,
|
67 |
+
"gen_kwargs": null
|
68 |
+
},
|
69 |
+
"git_hash": "1554066c"
|
70 |
+
}
|
Eagle-7B-1T-hs/results.json
ADDED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"results": {
|
3 |
+
"hellaswag": {
|
4 |
+
"acc,none": 0.5274845648277235,
|
5 |
+
"acc_stderr,none": 0.00498223713340941,
|
6 |
+
"acc_norm,none": 0.7165903206532563,
|
7 |
+
"acc_norm_stderr,none": 0.004497325533959324,
|
8 |
+
"alias": "hellaswag"
|
9 |
+
}
|
10 |
+
},
|
11 |
+
"configs": {
|
12 |
+
"hellaswag": {
|
13 |
+
"task": "hellaswag",
|
14 |
+
"group": [
|
15 |
+
"multiple_choice"
|
16 |
+
],
|
17 |
+
"dataset_path": "hellaswag",
|
18 |
+
"training_split": "train",
|
19 |
+
"validation_split": "validation",
|
20 |
+
"process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n def _process_doc(doc):\n ctx = doc[\"ctx_a\"] + \" \" + doc[\"ctx_b\"].capitalize()\n out_doc = {\n \"query\": preprocess(doc[\"activity_label\"] + \": \" + ctx),\n \"choices\": [preprocess(ending) for ending in doc[\"endings\"]],\n \"gold\": int(doc[\"label\"]),\n }\n return out_doc\n\n return dataset.map(_process_doc)\n",
|
21 |
+
"doc_to_text": "{{query}}",
|
22 |
+
"doc_to_target": "{{label}}",
|
23 |
+
"doc_to_choice": "choices",
|
24 |
+
"description": "",
|
25 |
+
"target_delimiter": " ",
|
26 |
+
"fewshot_delimiter": "\n\n",
|
27 |
+
"num_fewshot": 10,
|
28 |
+
"metric_list": [
|
29 |
+
{
|
30 |
+
"metric": "acc",
|
31 |
+
"aggregation": "mean",
|
32 |
+
"higher_is_better": true
|
33 |
+
},
|
34 |
+
{
|
35 |
+
"metric": "acc_norm",
|
36 |
+
"aggregation": "mean",
|
37 |
+
"higher_is_better": true
|
38 |
+
}
|
39 |
+
],
|
40 |
+
"output_type": "multiple_choice",
|
41 |
+
"repeats": 1,
|
42 |
+
"should_decontaminate": false,
|
43 |
+
"metadata": {
|
44 |
+
"version": 1.0
|
45 |
+
}
|
46 |
+
}
|
47 |
+
},
|
48 |
+
"versions": {
|
49 |
+
"hellaswag": 1.0
|
50 |
+
},
|
51 |
+
"n-shot": {
|
52 |
+
"hellaswag": 10
|
53 |
+
},
|
54 |
+
"config": {
|
55 |
+
"model": "hf",
|
56 |
+
"model_args": "pretrained=RWKV/rwkv-5-world-7b,dtype=float16,trust_remote_code=True",
|
57 |
+
"batch_size": "auto",
|
58 |
+
"batch_sizes": [
|
59 |
+
8
|
60 |
+
],
|
61 |
+
"device": null,
|
62 |
+
"use_cache": null,
|
63 |
+
"limit": null,
|
64 |
+
"bootstrap_iters": 100000,
|
65 |
+
"gen_kwargs": null
|
66 |
+
},
|
67 |
+
"git_hash": "1554066c"
|
68 |
+
}
|