Update README.md
Browse files
README.md
CHANGED
@@ -6,8 +6,8 @@ inference:
|
|
6 |
parameters:
|
7 |
max_new_tokens: 64
|
8 |
do_sample: true
|
9 |
-
temperature: 0.
|
10 |
-
repetition_penalty:
|
11 |
no_repeat_ngram_size: 4
|
12 |
eta_cutoff: 0.0006
|
13 |
renormalize_logits: true
|
@@ -77,15 +77,25 @@ No alignment has been done yet.
|
|
77 |
|
78 |
## [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
79 |
|
80 |
-
| Metric | kenhktsui/nano-phi-115M-v0.1|[kenhktsui/nano-phi-115M-control-v0.1](https://huggingface.co/kenhktsui/nano-phi-115M-control-v0.1)|
|
81 |
-
|
82 |
-
|
|
83 |
-
|
|
84 |
-
|
|
85 |
-
|
|
86 |
-
|
|
87 |
-
|
|
88 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
|
90 |
Details:
|
91 |
|
|
|
6 |
parameters:
|
7 |
max_new_tokens: 64
|
8 |
do_sample: true
|
9 |
+
temperature: 0.1
|
10 |
+
repetition_penalty: 10
|
11 |
no_repeat_ngram_size: 4
|
12 |
eta_cutoff: 0.0006
|
13 |
renormalize_logits: true
|
|
|
77 |
|
78 |
## [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
79 |
|
80 |
+
| Metric | kenhktsui/nano-phi-115M-v0.1|[kenhktsui/nano-phi-115M-control-v0.1](https://huggingface.co/kenhktsui/nano-phi-115M-control-v0.1)|[microsoft/phi-2](https://huggingface.co/microsoft/phi-2)|
|
81 |
+
|-----------------------|---------------------------|---------------------------|---------------------------|
|
82 |
+
| Model Para | 115M |115M |2.7B |
|
83 |
+
| Dataset Size | 0.26B |0.6B |250B |
|
84 |
+
| Training Token | 0.26B |0.6B |1.4T |
|
85 |
+
| Context Length |1024 |1024 |2048|
|
86 |
+
| Device |1xA100-40G|1xA100-40G |96xA100-80G|
|
87 |
+
| Training Time |2d4h |2d4h |14d|
|
88 |
+
|
89 |
+
|
90 |
+
| Metric | kenhktsui/nano-phi-115M-v0.1|[kenhktsui/nano-phi-115M-control-v0.1](https://huggingface.co/kenhktsui/nano-phi-115M-control-v0.1)|[microsoft/phi-2](https://huggingface.co/microsoft/phi-2) (Reproduced)|
|
91 |
+
|-----------------------|---------------------------|---------------------------|---------------------------|
|
92 |
+
| Avg. | 28.68 |28.75 |61.53 |
|
93 |
+
| ARC (25-shot) | 21.93 |21.67 |61.52 |
|
94 |
+
| HellaSwag (10-shot) | 27.87 |26.89 |75.13 |
|
95 |
+
| MMLU (5-shot) | 25.30 |24.76 |58.23 |
|
96 |
+
| TruthfulQA (0-shot) | 46.01 |47.69 |44.46 |
|
97 |
+
| Winogrande (5-shot) | 50.99 |51.46 |74.51 |
|
98 |
+
| GSM8K (5-shot) | 0.0 |0.0 |55.34 |
|
99 |
|
100 |
Details:
|
101 |
|