Instructions to use trymirai/Flint-S with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Trellis
How to use trymirai/Flint-S with Trellis:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Release package
Browse files
README.md
CHANGED
|
@@ -30,13 +30,13 @@ Both builds vs the bf16 model in Mirai's in-house agent circuits (SGLang, Flint'
|
|
| 30 |
| Hard-problem battery | 45/55 | 50/55 | 49/55 |
|
| 31 |
| AppWorld | 168/240 | 157/240 | 187/240 |
|
| 32 |
| Assistant week v2 (native tools) | 409/656 | 400/656 | 460/656 |
|
| 33 |
-
| Assistant week v1 (text markers) |
|
| 34 |
| Creative build, fully correct | 4/4 | 3/4 | 4/4 |
|
| 35 |
| Agentic build, declared done | 4/4 | 4/4 | 4/4 |
|
| 36 |
|
| 37 |
Mean answer length is 1,605 tokens for Flint S 11.8 (bf16: 1,623; 99.5% of answers finished within 8k tokens) and 1,710 for Flint S 12.4 (+5%; 98.5% finished).
|
| 38 |
|
| 39 |
-
The 11.8 GB build ships the weights after 12M tokens of QAT. Training it for 171M tokens lowered its KL to 0.1138 but lost on
|
| 40 |
|
| 41 |
## What is in this repository
|
| 42 |
**These are dense bf16 checkpoints (about 80 GB per build), not the packed 12.4 / 11.8 GB files.** Each folder holds the exact dequantized S weights in the original Flint layout, so they run with Flint's SGLang serving kit (`<build>/sglang/`, one H200 or B200/B300 GPU, as in the base model card) and reproduce the scores above. No runtime reads a packed Flint S package yet.
|
|
|
|
| 30 |
| Hard-problem battery | 45/55 | 50/55 | 49/55 |
|
| 31 |
| AppWorld | 168/240 | 157/240 | 187/240 |
|
| 32 |
| Assistant week v2 (native tools) | 409/656 | 400/656 | 460/656 |
|
| 33 |
+
| Assistant week v1 (text markers, 8 seeds) | 72/272 | 65/272 | 60/136 (4 seeds) |
|
| 34 |
| Creative build, fully correct | 4/4 | 3/4 | 4/4 |
|
| 35 |
| Agentic build, declared done | 4/4 | 4/4 | 4/4 |
|
| 36 |
|
| 37 |
Mean answer length is 1,605 tokens for Flint S 11.8 (bf16: 1,623; 99.5% of answers finished within 8k tokens) and 1,710 for Flint S 12.4 (+5%; 98.5% finished).
|
| 38 |
|
| 39 |
+
The 11.8 GB build ships the weights after 12M tokens of QAT. Training it for 171M tokens lowered its KL to 0.1138 but lost on assistant week v2 (paired, 16 seeds: −5.2 points, 95% interval −9.1 to −1.1; v1 tied at 8 seeds), so the circuits decide. The 12.4 GB build is the 171M-token checkpoint.
|
| 40 |
|
| 41 |
## What is in this repository
|
| 42 |
**These are dense bf16 checkpoints (about 80 GB per build), not the packed 12.4 / 11.8 GB files.** Each folder holds the exact dequantized S weights in the original Flint layout, so they run with Flint's SGLang serving kit (`<build>/sglang/`, one H200 or B200/B300 GPU, as in the base model card) and reproduce the scores above. No runtime reads a packed Flint S package yet.
|