Upload eval_results/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B/main/gpqa/results_2025-05-05T15-29-20.168420.json with huggingface_hub
Browse files
eval_results/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B/main/gpqa/results_2025-05-05T15-29-20.168420.json
ADDED
@@ -0,0 +1,138 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"config_general": {
|
3 |
+
"lighteval_sha": "?",
|
4 |
+
"num_fewshot_seeds": 1,
|
5 |
+
"max_samples": null,
|
6 |
+
"job_id": 0,
|
7 |
+
"start_time": 3620506.368588403,
|
8 |
+
"end_time": 3621489.119445614,
|
9 |
+
"total_evaluation_time_secondes": "982.7508572111838",
|
10 |
+
"model_name": "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B",
|
11 |
+
"model_sha": "",
|
12 |
+
"model_dtype": null,
|
13 |
+
"model_size": null,
|
14 |
+
"generation_parameters": {
|
15 |
+
"early_stopping": null,
|
16 |
+
"repetition_penalty": null,
|
17 |
+
"frequency_penalty": null,
|
18 |
+
"length_penalty": null,
|
19 |
+
"presence_penalty": null,
|
20 |
+
"max_new_tokens": 32768,
|
21 |
+
"min_new_tokens": null,
|
22 |
+
"seed": null,
|
23 |
+
"stop_tokens": null,
|
24 |
+
"temperature": 0.6,
|
25 |
+
"top_k": null,
|
26 |
+
"min_p": null,
|
27 |
+
"top_p": 0.95,
|
28 |
+
"truncate_prompt": null,
|
29 |
+
"response_format": null
|
30 |
+
}
|
31 |
+
},
|
32 |
+
"results": {
|
33 |
+
"lighteval|gpqa:diamond|0": {
|
34 |
+
"gpqa_pass@1:1_samples": 0.494949494949495,
|
35 |
+
"gpqa_pass@1:1_samples_stderr": 0.035621707606254015,
|
36 |
+
"gpqa_pass@1:4_samples": 0.494949494949495,
|
37 |
+
"gpqa_pass@1:4_samples_stderr": 0.028918511422630695,
|
38 |
+
"gpqa_pass@1:8_samples": 0.5050505050505051,
|
39 |
+
"gpqa_pass@1:8_samples_stderr": 0.027279058659552692
|
40 |
+
},
|
41 |
+
"all": {
|
42 |
+
"gpqa_pass@1:1_samples": 0.494949494949495,
|
43 |
+
"gpqa_pass@1:1_samples_stderr": 0.035621707606254015,
|
44 |
+
"gpqa_pass@1:4_samples": 0.494949494949495,
|
45 |
+
"gpqa_pass@1:4_samples_stderr": 0.028918511422630695,
|
46 |
+
"gpqa_pass@1:8_samples": 0.5050505050505051,
|
47 |
+
"gpqa_pass@1:8_samples_stderr": 0.027279058659552692
|
48 |
+
}
|
49 |
+
},
|
50 |
+
"versions": {
|
51 |
+
"lighteval|gpqa:diamond|0": 1
|
52 |
+
},
|
53 |
+
"config_tasks": {
|
54 |
+
"lighteval|gpqa:diamond": {
|
55 |
+
"name": "gpqa:diamond",
|
56 |
+
"prompt_function": "gpqa_instruct",
|
57 |
+
"hf_repo": "Idavidrein/gpqa",
|
58 |
+
"hf_subset": "gpqa_diamond",
|
59 |
+
"metric": [
|
60 |
+
{
|
61 |
+
"metric_name": "gpqa_pass@1:1_samples",
|
62 |
+
"higher_is_better": true,
|
63 |
+
"category": "5",
|
64 |
+
"use_case": "6",
|
65 |
+
"sample_level_fn": "compute",
|
66 |
+
"corpus_level_fn": "mean"
|
67 |
+
},
|
68 |
+
{
|
69 |
+
"metric_name": "gpqa_pass@1:4_samples",
|
70 |
+
"higher_is_better": true,
|
71 |
+
"category": "5",
|
72 |
+
"use_case": "6",
|
73 |
+
"sample_level_fn": "compute",
|
74 |
+
"corpus_level_fn": "mean"
|
75 |
+
},
|
76 |
+
{
|
77 |
+
"metric_name": "gpqa_pass@1:8_samples",
|
78 |
+
"higher_is_better": true,
|
79 |
+
"category": "5",
|
80 |
+
"use_case": "6",
|
81 |
+
"sample_level_fn": "compute",
|
82 |
+
"corpus_level_fn": "mean"
|
83 |
+
}
|
84 |
+
],
|
85 |
+
"hf_revision": null,
|
86 |
+
"hf_filter": null,
|
87 |
+
"hf_avail_splits": [
|
88 |
+
"train"
|
89 |
+
],
|
90 |
+
"trust_dataset": true,
|
91 |
+
"evaluation_splits": [
|
92 |
+
"train"
|
93 |
+
],
|
94 |
+
"few_shots_split": null,
|
95 |
+
"few_shots_select": null,
|
96 |
+
"generation_size": 32768,
|
97 |
+
"generation_grammar": null,
|
98 |
+
"stop_sequence": [],
|
99 |
+
"num_samples": null,
|
100 |
+
"suite": [
|
101 |
+
"lighteval"
|
102 |
+
],
|
103 |
+
"original_num_docs": 198,
|
104 |
+
"effective_num_docs": 198,
|
105 |
+
"must_remove_duplicate_docs": false,
|
106 |
+
"version": 1
|
107 |
+
}
|
108 |
+
},
|
109 |
+
"summary_tasks": {
|
110 |
+
"lighteval|gpqa:diamond|0": {
|
111 |
+
"hashes": {
|
112 |
+
"hash_examples": "50ecb6f5d091bd95",
|
113 |
+
"hash_full_prompts": "f27d3d556ab1c377",
|
114 |
+
"hash_input_tokens": "2feb29bf3801b58a",
|
115 |
+
"hash_cont_tokens": "e41f97da9d2685b2"
|
116 |
+
},
|
117 |
+
"truncated": 0,
|
118 |
+
"non_truncated": 198,
|
119 |
+
"padded": 0,
|
120 |
+
"non_padded": 198,
|
121 |
+
"effective_few_shots": 0.0,
|
122 |
+
"num_truncated_few_shots": 0
|
123 |
+
}
|
124 |
+
},
|
125 |
+
"summary_general": {
|
126 |
+
"hashes": {
|
127 |
+
"hash_examples": "a9318dbdd867770b",
|
128 |
+
"hash_full_prompts": "122c405ea50f3de4",
|
129 |
+
"hash_input_tokens": "a263b8e88b1dea74",
|
130 |
+
"hash_cont_tokens": "bb93b1c97d7cd5b0"
|
131 |
+
},
|
132 |
+
"truncated": 0,
|
133 |
+
"non_truncated": 198,
|
134 |
+
"padded": 0,
|
135 |
+
"non_padded": 198,
|
136 |
+
"num_truncated_few_shots": 0
|
137 |
+
}
|
138 |
+
}
|