Upload eval_results/orpo-explorers/argilla-mistral-orpo-OpenHermesPreferences-50k-beta-0.2/main/gsm8k/results_2024-05-09T19-59-03.380354.json with huggingface_hub
Browse files
eval_results/orpo-explorers/argilla-mistral-orpo-OpenHermesPreferences-50k-beta-0.2/main/gsm8k/results_2024-05-09T19-59-03.380354.json
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"config_general": {
|
3 |
+
"lighteval_sha": "?",
|
4 |
+
"num_fewshot_seeds": 1,
|
5 |
+
"override_batch_size": 4,
|
6 |
+
"max_samples": null,
|
7 |
+
"job_id": "",
|
8 |
+
"start_time": 23496.854474168,
|
9 |
+
"end_time": 24293.695778991,
|
10 |
+
"total_evaluation_time_secondes": "796.8413048229995",
|
11 |
+
"model_name": "orpo-explorers/argilla-mistral-orpo-OpenHermesPreferences-50k-beta-0.2",
|
12 |
+
"model_sha": "8e456b857961e00f2e1b076756e84b97a3b88ea5",
|
13 |
+
"model_dtype": "torch.bfloat16",
|
14 |
+
"model_size": "13.99 GB",
|
15 |
+
"config": null
|
16 |
+
},
|
17 |
+
"results": {
|
18 |
+
"lighteval|gsm8k|5": {
|
19 |
+
"qem": 0.19257012888551933,
|
20 |
+
"qem_stderr": 0.010861483868509924,
|
21 |
+
"maj@8": 0.27748294162244125,
|
22 |
+
"maj@8_stderr": 0.012333447581047544
|
23 |
+
},
|
24 |
+
"all": {
|
25 |
+
"qem": 0.19257012888551933,
|
26 |
+
"qem_stderr": 0.010861483868509924,
|
27 |
+
"maj@8": 0.27748294162244125,
|
28 |
+
"maj@8_stderr": 0.012333447581047544
|
29 |
+
}
|
30 |
+
},
|
31 |
+
"versions": {
|
32 |
+
"lighteval|gsm8k|5": 0
|
33 |
+
},
|
34 |
+
"config_tasks": {
|
35 |
+
"lighteval|gsm8k": {
|
36 |
+
"name": "gsm8k",
|
37 |
+
"prompt_function": "gsm8k",
|
38 |
+
"hf_repo": "gsm8k",
|
39 |
+
"hf_subset": "main",
|
40 |
+
"metric": [
|
41 |
+
"quasi_exact_match_gsm8k",
|
42 |
+
"maj_at_8_gsm8k"
|
43 |
+
],
|
44 |
+
"hf_avail_splits": [
|
45 |
+
"train",
|
46 |
+
"test"
|
47 |
+
],
|
48 |
+
"evaluation_splits": [
|
49 |
+
"test"
|
50 |
+
],
|
51 |
+
"few_shots_split": null,
|
52 |
+
"few_shots_select": "random_sampling_from_train",
|
53 |
+
"generation_size": 256,
|
54 |
+
"stop_sequence": [
|
55 |
+
"Question:"
|
56 |
+
],
|
57 |
+
"output_regex": null,
|
58 |
+
"num_samples": null,
|
59 |
+
"frozen": false,
|
60 |
+
"suite": [
|
61 |
+
"lighteval"
|
62 |
+
],
|
63 |
+
"original_num_docs": 1319,
|
64 |
+
"effective_num_docs": 1319,
|
65 |
+
"trust_dataset": true,
|
66 |
+
"must_remove_duplicate_docs": null,
|
67 |
+
"version": 0
|
68 |
+
}
|
69 |
+
},
|
70 |
+
"summary_tasks": {
|
71 |
+
"lighteval|gsm8k|5": {
|
72 |
+
"hashes": {
|
73 |
+
"hash_examples": "0ed016e24e7512fd",
|
74 |
+
"hash_full_prompts": "faa09535944c3187",
|
75 |
+
"hash_input_tokens": "61e7bfb27a50fc48",
|
76 |
+
"hash_cont_tokens": "98d77946f06cd16d"
|
77 |
+
},
|
78 |
+
"truncated": 1319,
|
79 |
+
"non_truncated": 0,
|
80 |
+
"padded": 652,
|
81 |
+
"non_padded": 667,
|
82 |
+
"effective_few_shots": 5.0,
|
83 |
+
"num_truncated_few_shots": 0
|
84 |
+
}
|
85 |
+
},
|
86 |
+
"summary_general": {
|
87 |
+
"hashes": {
|
88 |
+
"hash_examples": "bc71463e88551d0e",
|
89 |
+
"hash_full_prompts": "24555d96d1b06364",
|
90 |
+
"hash_input_tokens": "fbc91ddb76fcfe8f",
|
91 |
+
"hash_cont_tokens": "6ff92edc86962a8d"
|
92 |
+
},
|
93 |
+
"truncated": 1319,
|
94 |
+
"non_truncated": 0,
|
95 |
+
"padded": 652,
|
96 |
+
"non_padded": 667,
|
97 |
+
"num_truncated_few_shots": 0
|
98 |
+
}
|
99 |
+
}
|