Update README.md
Browse files
README.md
CHANGED
|
@@ -16,6 +16,19 @@ there was originally going to be a better logo but i couldnt get any image model
|
|
| 16 |
|
| 17 |
---
|
| 18 |
|
|
|
|
|
|
|
| 19 |
Lune Mamba 3B is a Claude-OSS series model based on Granite 4.0 H(ybrid) Micro.
|
| 20 |
|
| 21 |
-
Claude-OSS is a (non-affiliated with Anthropic!) attempt to replicate the style of Anthropic's Claude model on top of open source bases.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
|
| 17 |
---
|
| 18 |
|
| 19 |
+
#### Info
|
| 20 |
+
|
| 21 |
Lune Mamba 3B is a Claude-OSS series model based on Granite 4.0 H(ybrid) Micro.
|
| 22 |
|
| 23 |
+
Claude-OSS is a (non-affiliated with Anthropic!) attempt to replicate the style of Anthropic's Claude model on top of open source bases.
|
| 24 |
+
|
| 25 |
+
*Benchmarks* | Granite 4.0 H Micro | Lune Mamba 3B | Lune Mamba 3B GRPO_IF
|
| 26 |
+
-|-|-|-
|
| 27 |
+
MMLU|63.7860|*64.2338*|**64.3443**
|
| 28 |
+
IFEval*|**80.2218**|75.0462|*77.4492*
|
| 29 |
+
<small>* IFEval numbers calculated from prompt loose accuracy </small>
|
| 30 |
+
|
| 31 |
+
#### Artifacts
|
| 32 |
+
- SFT checkpoint: [allura-forge/claumba-micro-sft](/allura-forge/claumba-micro-sft)
|
| 33 |
+
- KTO checkpoint: You are here!
|
| 34 |
+
- GRPO (on IFeval) checkpoint: [allura-org/Lune-Mamba-3b-v1-GRPO_IF](/allura-org/Lune-Mamba-3b-v1-GRPO_IF)
|