sathiskumarjothi0 commited on
Commit
bc1fea7
·
verified ·
1 Parent(s): 3d481d5

Model card, licence, charts

Browse files
Files changed (1) hide show
  1. README.md +3 -3
README.md CHANGED
@@ -75,7 +75,7 @@ questions; it returns typed answers (yes/no, a choice between options you descri
75
  with a calibrated probability for every option, in one pass. It never generates text, so there is nothing to parse and
76
  nothing to hallucinate. It runs on your own machines: your documents never leave them.
77
 
78
- About 26B parameters: Google Gemma 4 26B-A4B (mixture of experts, about 4B active per token) with Bev's decision adapter. Both are in this repository, so nothing else is downloaded. English text, up to 4,096 input tokens; needs one GPU with about 60 GB of memory (bf16).
79
 
80
  | Model | Size | Runs on | Best for | Status |
81
  |---|---|---|---|---|
@@ -275,7 +275,7 @@ A few answers that are close to a tie can change with batch size on a mixture-of
275
  | File | Contents |
276
  |---|---|
277
  | `model-*.safetensors`, `config.json`, tokenizer files | Gemma 4 26B-A4B, unchanged (bf16) |
278
- | `adapter/` | Bev's decision adapter, applied when the model loads |
279
  | `bev_config.json` | Bev settings and calibration temperatures |
280
  | `NOTICE` | Gemma attribution (Apache-2.0) |
281
  | `assets/` | the charts on this page |
@@ -291,5 +291,5 @@ If Bev is useful to you, you can support its development: [buymeacoffee.com/sjot
291
 
292
  ## Licence and credit
293
 
294
- Apache-2.0, © 2026 Dr. Sathiskumar JOTHI. Built on [Gemma 4 26B-A4B](https://huggingface.co/google/gemma-4-26B-A4B-it) (Apache-2.0) by Google, included unchanged with Bev's decision adapter, see `NOTICE`. Bev is named after Sir William Beveridge (1879–1963), British
295
  economist, whose 1942 Beveridge Report laid the foundations of the welfare state and the NHS.
 
75
  with a calibrated probability for every option, in one pass. It never generates text, so there is nothing to parse and
76
  nothing to hallucinate. It runs on your own machines: your documents never leave them.
77
 
78
+ About 26B parameters, built on Google Gemma 4 26B-A4B (mixture of experts, about 4B active per token). Everything it needs is in this repository, so nothing else is downloaded. English text, up to 4,096 input tokens; needs one GPU with about 60 GB of memory (bf16).
79
 
80
  | Model | Size | Runs on | Best for | Status |
81
  |---|---|---|---|---|
 
275
  | File | Contents |
276
  |---|---|
277
  | `model-*.safetensors`, `config.json`, tokenizer files | Gemma 4 26B-A4B, unchanged (bf16) |
278
+ | `adapter/` | Bev's decision weights, loaded together with Gemma's |
279
  | `bev_config.json` | Bev settings and calibration temperatures |
280
  | `NOTICE` | Gemma attribution (Apache-2.0) |
281
  | `assets/` | the charts on this page |
 
291
 
292
  ## Licence and credit
293
 
294
+ Apache-2.0, © 2026 Dr. Sathiskumar JOTHI. Built on [Gemma 4 26B-A4B](https://huggingface.co/google/gemma-4-26B-A4B-it) (Apache-2.0) by Google. Bev is named after Sir William Beveridge (1879–1963), British
295
  economist, whose 1942 Beveridge Report laid the foundations of the welfare state and the NHS.