cosmo3769 commited on
Commit
92ebaf8
·
verified ·
1 Parent(s): 31f4e5f

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +55 -0
README.md ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # starcoderbase-3b-GPTQ
2
+
3
+ Quantized starcoderbase-3b model to GPTQ format (4-bit precision).
4
+
5
+ ## Benchmark
6
+
7
+ [Benchmarking script](https://github.com/cosmo3769/Quantized-LLMs/blob/main/notebooks/llmbenchmark-starcodebase-3b-lm-eval-harness.ipynb)
8
+
9
+ ### Baseline starcoderbase-3b model (non-quantized)
10
+
11
+ | Tasks |Version|Filter|n-shot| Metric |Value | |Stderr|
12
+ |-----------------------|-------|------|------|---------------|-----:|---|-----:|
13
+ |codexglue_code2text |N/A |none |None |smoothed_bleu_4|1.3519|± |0.3067|
14
+ | - code2text_go | 1|none |None |smoothed_bleu_4|1.5781|± |0.3734|
15
+ | - code2text_java | 1|none |None |smoothed_bleu_4|1.2778|± |0.1991|
16
+ | - code2text_javascript| 1|none |None |smoothed_bleu_4|1.1443|± |0.1181|
17
+ | - code2text_php | 1|none |None |smoothed_bleu_4|0.5171|± |0.5171|
18
+ | - code2text_python | 1|none |None |smoothed_bleu_4|2.8338|± |1.5323|
19
+ | - code2text_ruby | 3|none |None |smoothed_bleu_4|0.7601|± |0.7601|
20
+
21
+ | Groups |Version|Filter|n-shot| Metric |Value | |Stderr|
22
+ |-------------------|-------|------|------|---------------|-----:|---|-----:|
23
+ |codexglue_code2text|N/A |none |None |smoothed_bleu_4|1.3519|± |0.3067|
24
+
25
+ | Tasks |Version|Filter|n-shot| Metric |Value| |Stderr|
26
+ |---------------------------------------------|------:|------|------|-----------|----:|---|-----:|
27
+ |bigbench_code_line_description_generate_until| 1|none |None |exact_match| 0|± | 0|
28
+
29
+ | Tasks |Version|Filter|n-shot|Metric|Value| |Stderr|
30
+ |----------------------------------------------|------:|------|------|------|----:|---|-----:|
31
+ |bigbench_code_line_description_multiple_choice| 0|none |None |acc | 0.25|± |0.0564|
32
+
33
+ ### Quantized starcoderbase-3b model to GPTQ format
34
+
35
+ | Tasks |Version|Filter|n-shot| Metric |Value | |Stderr|
36
+ |-----------------------|-------|------|------|---------------|-----:|---|-----:|
37
+ |codexglue_code2text |N/A |none |None |smoothed_bleu_4|0.9254|± |0.2109|
38
+ | - code2text_go | 1|none |None |smoothed_bleu_4|1.4702|± |0.4813|
39
+ | - code2text_java | 1|none |None |smoothed_bleu_4|0.6907|± |0.6907|
40
+ | - code2text_javascript| 1|none |None |smoothed_bleu_4|0.9469|± |0.0339|
41
+ | - code2text_php | 1|none |None |smoothed_bleu_4|0.5171|± |0.5171|
42
+ | - code2text_python | 1|none |None |smoothed_bleu_4|1.1676|± |0.2156|
43
+ | - code2text_ruby | 3|none |None |smoothed_bleu_4|0.7601|± |0.7601|
44
+
45
+ | Groups |Version|Filter|n-shot| Metric |Value | |Stderr|
46
+ |-------------------|-------|------|------|---------------|-----:|---|-----:|
47
+ |codexglue_code2text|N/A |none |None |smoothed_bleu_4|0.9254|± |0.2109|
48
+
49
+ | Tasks |Version|Filter|n-shot| Metric |Value| |Stderr|
50
+ |---------------------------------------------|------:|------|------|-----------|----:|---|-----:|
51
+ |bigbench_code_line_description_generate_until| 1|none |None |exact_match| 0|± | 0|
52
+
53
+ | Tasks |Version|Filter|n-shot|Metric|Value| |Stderr|
54
+ |----------------------------------------------|------:|------|------|------|----:|---|-----:|
55
+ |bigbench_code_line_description_multiple_choice| 0|none |None |acc | 0.1|± | 0.1|