Update README.md
Browse files
README.md
CHANGED
@@ -18,40 +18,53 @@ pipeline_tag: text-generation
|
|
18 |
|
19 |
- [HF model](https://huggingface.co/Vikhrmodels/Vikhr-Gemma-2B-instruct)
|
20 |
|
21 |
-
## Perplexity
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
|
27 |
-
|
|
28 |
-
|
|
29 |
-
|
|
30 |
-
|
|
31 |
-
|
|
32 |
-
|
|
33 |
-
|
|
34 |
-
|
|
35 |
-
|
|
36 |
-
|
|
37 |
-
|
|
38 |
-
|
|
39 |
-
|
|
40 |
-
|
|
41 |
-
|
|
42 |
-
|
|
43 |
-
|
|
44 |
-
|
|
45 |
-
|
|
46 |
-
|
|
47 |
-
| Vikhr-Gemma-2B-instruct-Q4_K_S | 10.4924 +/- 0.07386 |
|
48 |
-
| Vikhr-Gemma-2B-instruct-Q5_K_S | 10.2098 +/- 0.07198 |
|
49 |
-
| Vikhr-Gemma-2B-instruct-IQ3_XS | 11.1461 +/- 0.07907 |
|
50 |
-
| Vikhr-Gemma-2B-instruct-Q3_K | 10.7416 +/- 0.07606 |
|
51 |
-
| Vikhr-Gemma-2B-instruct-Q3_K_M | 10.7416 +/- 0.07606 |
|
52 |
-
| Vikhr-Gemma-2B-instruct-IQ3_M | 10.7035 +/- 0.07573 |
|
53 |
-
| Vikhr-Gemma-2B-instruct-Q3_K_L | 10.6242 +/- 0.07506 |
|
54 |
-
| Vikhr-Gemma-2B-instruct-IQ2_XS | 15.8229 +/- 0.11909 |
|
55 |
-
| Vikhr-Gemma-2B-instruct-IQ1_M | 29.5443 +/- 0.24168 |
|
56 |
-
| Vikhr-Gemma-2B-instruct-IQ3_XXS | 11.6931 +/- 0.08282 |
|
57 |
-
| Vikhr-Gemma-2B-instruct-IQ2_XXS | 18.8375 +/- 0.14581 |
|
|
|
18 |
|
19 |
- [HF model](https://huggingface.co/Vikhrmodels/Vikhr-Gemma-2B-instruct)
|
20 |
|
21 |
+
## Perplexity
|
22 |
+
Veles Results:
|
23 |
+
| Model | Perplexity |
|
24 |
+
|:--------|:-------------------|
|
25 |
+
| Q4_K | 4.7254 +/- 0.03867 |
|
26 |
+
| Q4_0 | 4.8067 +/- 0.03922 |
|
27 |
+
| Q8_0 | 4.6042 +/- 0.03751 |
|
28 |
+
| Q4_1 | 4.7798 +/- 0.03933 |
|
29 |
+
| F32 | 4.6013 +/- 0.03749 |
|
30 |
+
| Q6_K | 4.6244 +/- 0.03760 |
|
31 |
+
| BF16 | 4.6015 +/- 0.03749 |
|
32 |
+
| Q2_K | 5.6819 +/- 0.04737 |
|
33 |
+
| Q5_0 | 4.6876 +/- 0.03855 |
|
34 |
+
| Q5_K | 4.6428 +/- 0.03789 |
|
35 |
+
| Q3_K_S | 5.1485 +/- 0.04257 |
|
36 |
+
| Q2_K_S | 6.3124 +/- 0.05359 |
|
37 |
+
| F16 | 4.6013 +/- 0.03749 |
|
38 |
+
| Q4_K_M | 4.7254 +/- 0.03867 |
|
39 |
+
| Q5_K_M | 4.6428 +/- 0.03789 |
|
40 |
+
| Q5_1 | 4.6518 +/- 0.03794 |
|
41 |
+
| Q4_K_S | 4.7631 +/- 0.03916 |
|
42 |
+
| Q5_K_S | 4.6509 +/- 0.03803 |
|
43 |
+
| Q3_K | 4.8339 +/- 0.03965 |
|
44 |
+
| Q3_K_M | 4.8339 +/- 0.03965 |
|
45 |
+
| Q3_K_L | 4.7981 +/- 0.03934 |
|
46 |
|
47 |
+
Wikitext-2 Results:
|
48 |
+
| Model | Perplexity |
|
49 |
+
|:--------|:--------------------|
|
50 |
+
| Q4_K | 10.4374 +/- 0.07339 |
|
51 |
+
| Q4_0 | 10.6480 +/- 0.07452 |
|
52 |
+
| Q8_0 | 10.1209 +/- 0.07105 |
|
53 |
+
| Q4_1 | 10.5574 +/- 0.07476 |
|
54 |
+
| F32 | 10.1191 +/- 0.07099 |
|
55 |
+
| Q6_K | 10.1503 +/- 0.07117 |
|
56 |
+
| BF16 | 10.1189 +/- 0.07098 |
|
57 |
+
| Q2_K | 12.8851 +/- 0.09332 |
|
58 |
+
| Q5_0 | 10.2551 +/- 0.07251 |
|
59 |
+
| Q5_K | 10.1975 +/- 0.07184 |
|
60 |
+
| Q3_K_S | 11.6028 +/- 0.08333 |
|
61 |
+
| Q2_K_S | 14.7951 +/- 0.10960 |
|
62 |
+
| F16 | 10.1191 +/- 0.07099 |
|
63 |
+
| Q4_K_M | 10.4374 +/- 0.07339 |
|
64 |
+
| Q5_K_M | 10.1975 +/- 0.07184 |
|
65 |
+
| Q5_1 | 10.2348 +/- 0.07208 |
|
66 |
+
| Q4_K_S | 10.4924 +/- 0.07386 |
|
67 |
+
| Q5_K_S | 10.2098 +/- 0.07198 |
|
68 |
+
| Q3_K | 10.7416 +/- 0.07606 |
|
69 |
+
| Q3_K_M | 10.7416 +/- 0.07606 |
|
70 |
+
| Q3_K_L | 10.6242 +/- 0.07506 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|