Add Gemma 4 31B IT Jacobian-lens adapter

#4
README.md CHANGED
@@ -98,15 +98,21 @@ shape match alone does not make their lenses interchangeable.
98
 
99
  | Base model | Adapter file | Prompts averaged | Convergence | Contributor |
100
  |---|---|---:|---|---|
 
101
  | `Qwen/Qwen3-0.6B` | [`miru/Qwen/Qwen3-0.6B.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/Qwen/Qwen3-0.6B.safetensors) | 376 | Yes (default early-stopping criterion reached) | [Rodrigo Laneth](https://rlaneth.com) ([@rlaneth](https://huggingface.co/rlaneth)) |
102
  | `Qwen/Qwen3-4B` | [`miru/Qwen/Qwen3-4B.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/Qwen/Qwen3-4B.safetensors) | 479 | Yes (default early-stopping criterion reached) | [Rodrigo Laneth](https://rlaneth.com) ([@rlaneth](https://huggingface.co/rlaneth)) |
103
  | `Qwen/Qwen3-8B` | [`miru/Qwen/Qwen3-8B.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/Qwen/Qwen3-8B.safetensors) | 461 | Yes (default early-stopping criterion reached) | [Rodrigo Laneth](https://rlaneth.com) ([@rlaneth](https://huggingface.co/rlaneth)) |
104
  | `Qwen/Qwen3.6-27B` | [`miru/Qwen/Qwen3.6-27B.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/Qwen/Qwen3.6-27B.safetensors) | 744 | Yes (default early-stopping criterion reached) | [Lucas Teske](https://huggingface.co/racerxdl) at [Teske's Lab](https://huggingface.co/TeskesLab), with [Paulo Matias](https://huggingface.co/thotypous) |
 
105
 
106
  ### Notes
107
 
108
  The `Qwen/Qwen3.6-27B` adapter was fitted against model revision [`6a9e13bd6fc8f0983b9b99948120bc37f49c13e9`](https://huggingface.co/Qwen/Qwen3.6-27B/commit/6a9e13bd6fc8f0983b9b99948120bc37f49c13e9) using the **Ambiente Computacional Marie Curie (FINEP 01.22.181.00)** at UFSCar.
109
 
 
 
 
 
110
  More models and community contributors will be added over time.
111
 
112
  ## Using an adapter in Miru Tracer
 
98
 
99
  | Base model | Adapter file | Prompts averaged | Convergence | Contributor |
100
  |---|---|---:|---|---|
101
+ | `google/gemma-4-31B-it` | [`miru/google/gemma-4-31B-it.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/google/gemma-4-31B-it.safetensors) | 425 | Yes (default early-stopping criterion reached) | [Lucas Teske](https://huggingface.co/racerxdl) at [Teske's Lab](https://huggingface.co/TeskesLab) |
102
  | `Qwen/Qwen3-0.6B` | [`miru/Qwen/Qwen3-0.6B.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/Qwen/Qwen3-0.6B.safetensors) | 376 | Yes (default early-stopping criterion reached) | [Rodrigo Laneth](https://rlaneth.com) ([@rlaneth](https://huggingface.co/rlaneth)) |
103
  | `Qwen/Qwen3-4B` | [`miru/Qwen/Qwen3-4B.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/Qwen/Qwen3-4B.safetensors) | 479 | Yes (default early-stopping criterion reached) | [Rodrigo Laneth](https://rlaneth.com) ([@rlaneth](https://huggingface.co/rlaneth)) |
104
  | `Qwen/Qwen3-8B` | [`miru/Qwen/Qwen3-8B.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/Qwen/Qwen3-8B.safetensors) | 461 | Yes (default early-stopping criterion reached) | [Rodrigo Laneth](https://rlaneth.com) ([@rlaneth](https://huggingface.co/rlaneth)) |
105
  | `Qwen/Qwen3.6-27B` | [`miru/Qwen/Qwen3.6-27B.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/Qwen/Qwen3.6-27B.safetensors) | 744 | Yes (default early-stopping criterion reached) | [Lucas Teske](https://huggingface.co/racerxdl) at [Teske's Lab](https://huggingface.co/TeskesLab), with [Paulo Matias](https://huggingface.co/thotypous) |
106
+ | `orcarouter/Qwen3.8-27B-Uncensored` | [`miru/orcarouter/Qwen3.8-27B-Uncensored.safetensors`](https://huggingface.co/returnmoe/jlens-adapters/blob/main/miru/orcarouter/Qwen3.8-27B-Uncensored.safetensors) | 761 | Yes (default early-stopping criterion reached) | [Lucas Teske](https://huggingface.co/racerxdl) at [Teske's Lab](https://huggingface.co/TeskesLab) |
107
 
108
  ### Notes
109
 
110
  The `Qwen/Qwen3.6-27B` adapter was fitted against model revision [`6a9e13bd6fc8f0983b9b99948120bc37f49c13e9`](https://huggingface.co/Qwen/Qwen3.6-27B/commit/6a9e13bd6fc8f0983b9b99948120bc37f49c13e9) using the **Ambiente Computacional Marie Curie (FINEP 01.22.181.00)** at UFSCar.
111
 
112
+ The `orcarouter/Qwen3.8-27B-Uncensored` adapter was fitted against revision [`404ea47aaa5d8a8b00049c9e9750089aca011ab2`](https://huggingface.co/orcarouter/Qwen3.8-27B-Uncensored/commit/404ea47aaa5d8a8b00049c9e9750089aca011ab2) using Miru Tracer 0.3.2 on two H100 NVL GPUs at the **Ambiente Computacional Marie Curie (FINEP 01.22.181.00)** at UFSCar. It averaged 761 WikiText calibration prompts with no skips; the final rolling change was `0.0019094125973001387`, below `0.002`. It uses the same frozen prompt pool and fitting settings as the original Qwen3.8-27B adapter proposed in [PR #5](https://huggingface.co/returnmoe/jlens-adapters/discussions/5), but stopped at a different prompt count (761 versus 756). The adapters are not interchangeable, and no normal-versus-uncensored behavioral comparison is claimed.
113
+
114
+ The `google/gemma-4-31B-it` adapter was fitted against model revision [`842da3794eaa0b77d5f08bae87a17459d91ff475`](https://huggingface.co/google/gemma-4-31B-it/commit/842da3794eaa0b77d5f08bae87a17459d91ff475) and reached the default convergence criterion after 425 prompts using the same Marie Curie environment.
115
+
116
  More models and community contributors will be added over time.
117
 
118
  ## Using an adapter in Miru Tracer
miru/google/gemma-4-31B-it.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cedd4b62f49a7f79e07c1bb107ad1d48d39dc7ad2790611de7e9480b4543386a
3
+ size 3410446944
miru/orcarouter/Qwen3.8-27B-Uncensored.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b6e1d730944dd7be5af3a2ea4327771d85c5f7d9dd35dd76bd5fd997f5b6bfd
3
+ size 3303160072