tssst commited on
Commit
6f32206
·
verified ·
1 Parent(s): 9e1bbfa

Upload 17 files

Browse files
.gitattributes CHANGED
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ Nemesia-Gemma3-12B-v2.0-F16.gguf filter=lfs diff=lfs merge=lfs -text
37
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
Nemesia-Gemma3-12B-v2.0-F16.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a47475dfb6837d51e8fbc960b33769aeaea73f49be0ea99113a5cd71468eae41
3
+ size 23540152096
README.md CHANGED
@@ -1,3 +1,64 @@
1
  ---
2
- license: gemma
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ base_model:
3
+ - mlabonne/gemma-3-12b-it-abliterated
4
+ - google/gemma-3-12b-it
5
+ - qiuxi337/gemma-3-12b-it-grpo
6
+ library_name: transformers
7
+ tags:
8
+ - mergekit
9
+ - merge
10
+
11
  ---
12
+ # temp
13
+
14
+ This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
15
+
16
+ ## Merge Details
17
+ ### Merge Method
18
+
19
+ This model was merged using the NuSLERP merge method using [mlabonne/gemma-3-12b-it-abliterated](https://huggingface.co/mlabonne/gemma-3-12b-it-abliterated) as a base.
20
+
21
+ Well, it's been a while, hasn't it? It's only been a short 5 months since I posted the last Nemesia version, saying the famous last words
22
+
23
+ > I will try swapping out a model or two in the merge and trying again to upload as a v2.0.
24
+
25
+ Then, exams and university stuff crushed me whole which wasn't great, and I didn't have any time or computational resources to get back to merging on a reasonable timeframe until now.
26
+
27
+ Honestly, in the time since, Qwen2.5-7B really has not turned out to be all that interesting, Mistral Nemo stealing most of its thunder. So, I decided to update my Nemesia mergeset with a
28
+ new base: **Gemma3-12B**! In my experience, it has been alright, so I'm excited to tinker with it.
29
+
30
+ In my testing, this thing is alright- I couldn't run any GGUFs of it since I don't even know what on earth has happened to my `.venv` in the 5 months past, I'm pretty sure my Transformers installation
31
+ is corrupted, and I'm re-cloning llama.cpp as I write this. I ran the FP16 and it seemed coherent enough, but I leave that to you to decide.
32
+
33
+ Uses all of the special NuSLERP options because they're there, and is based on the abliterated version of Gemma3 instead of the base because I <3 mlabonne.
34
+
35
+ ### Models Merged
36
+
37
+ The following models were included in the merge:
38
+ * [google/gemma-3-12b-it](https://huggingface.co/google/gemma-3-12b-it)
39
+ * [qiuxi337/gemma-3-12b-it-grpo](https://huggingface.co/qiuxi337/gemma-3-12b-it-grpo)
40
+
41
+ ### Configuration
42
+
43
+ The following YAML configuration was used to produce this model:
44
+
45
+ ```yaml
46
+ models:
47
+ - model: mlabonne/gemma-3-12b-it-abliterated
48
+ parameters:
49
+ weight: 1.0
50
+ - model: qiuxi337/gemma-3-12b-it-grpo
51
+ parameters:
52
+ weight: 0.5
53
+ - model: google/gemma-3-12b-it
54
+ parameters:
55
+ weight: 0.2
56
+ merge_method: nuslerp
57
+ base_model: mlabonne/gemma-3-12b-it-abliterated
58
+ parameters:
59
+ normalize: true
60
+ int8_mask: true
61
+ nuslerp_flatten: false
62
+ nuslerp_row_wise: true
63
+ dtype: float16
64
+ ```
added_tokens.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "<image_soft_token>": 262144
3
+ }
config.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Gemma3ForConditionalGeneration"
4
+ ],
5
+ "boi_token_index": 255999,
6
+ "eoi_token_index": 256000,
7
+ "eos_token_id": [
8
+ 1,
9
+ 106
10
+ ],
11
+ "image_token_index": 262144,
12
+ "initializer_range": 0.02,
13
+ "mm_tokens_per_image": 256,
14
+ "model_type": "gemma3",
15
+ "text_config": {
16
+ "attention_bias": false,
17
+ "attention_dropout": 0.0,
18
+ "attn_logit_softcapping": null,
19
+ "cache_implementation": "hybrid",
20
+ "final_logit_softcapping": null,
21
+ "head_dim": 256,
22
+ "hidden_activation": "gelu_pytorch_tanh",
23
+ "hidden_size": 3840,
24
+ "initializer_range": 0.02,
25
+ "intermediate_size": 15360,
26
+ "max_position_embeddings": 131072,
27
+ "model_type": "gemma3_text",
28
+ "num_attention_heads": 16,
29
+ "num_hidden_layers": 48,
30
+ "num_key_value_heads": 8,
31
+ "query_pre_attn_scalar": 256,
32
+ "rms_norm_eps": 1e-06,
33
+ "rope_local_base_freq": 10000.0,
34
+ "rope_scaling": {
35
+ "factor": 8.0,
36
+ "rope_type": "linear"
37
+ },
38
+ "rope_theta": 1000000.0,
39
+ "sliding_window": 1024,
40
+ "sliding_window_pattern": 6,
41
+ "torch_dtype": "bfloat16",
42
+ "use_cache": true,
43
+ "vocab_size": 262208
44
+ },
45
+ "torch_dtype": "float16",
46
+ "transformers_version": "4.51.3",
47
+ "vision_config": {
48
+ "attention_dropout": 0.0,
49
+ "hidden_act": "gelu_pytorch_tanh",
50
+ "hidden_size": 1152,
51
+ "image_size": 896,
52
+ "intermediate_size": 4304,
53
+ "layer_norm_eps": 1e-06,
54
+ "model_type": "siglip_vision_model",
55
+ "num_attention_heads": 16,
56
+ "num_channels": 3,
57
+ "num_hidden_layers": 27,
58
+ "patch_size": 14,
59
+ "torch_dtype": "bfloat16",
60
+ "vision_use_head": false
61
+ }
62
+ }
mergekit_config.yml ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ models:
2
+ - model: mlabonne/gemma-3-12b-it-abliterated
3
+ parameters:
4
+ weight: 1.0
5
+ - model: qiuxi337/gemma-3-12b-it-grpo
6
+ parameters:
7
+ weight: 0.5
8
+ - model: google/gemma-3-12b-it
9
+ parameters:
10
+ weight: 0.2
11
+ merge_method: nuslerp
12
+ base_model: mlabonne/gemma-3-12b-it-abliterated
13
+ parameters:
14
+ normalize: true
15
+ int8_mask: true
16
+ nuslerp_flatten: false
17
+ nuslerp_row_wise: true
18
+ dtype: float16
model-00001-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d4e87bfc31218de4b01d23a8095552e2169f3e719ead934baec60949732d171
3
+ size 4939493352
model-00002-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8b1c2f0c53647c22cdc719a7ef4a94a2d70d3333f451b5190f29cff55814567
3
+ size 4931296512
model-00003-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3778ec3c69fa20cf3a3bb0a7fc7074b613ce945ebcf4dc31742ba3d0fbb1bcd3
3
+ size 4931296512
model-00004-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c00aa65941098a39998dace46d3bdbd70d577cee56e50efcb34cbe113207a622
3
+ size 4931296512
model-00005-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:976d86a5b281df8a9a8ced0ffb935540ada0d438bab2c4e839b3e54704c53465
3
+ size 4641409128
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_pan_and_scan": null,
5
+ "do_rescale": true,
6
+ "do_resize": true,
7
+ "image_mean": [
8
+ 0.5,
9
+ 0.5,
10
+ 0.5
11
+ ],
12
+ "image_processor_type": "Gemma3ImageProcessor",
13
+ "image_seq_length": 256,
14
+ "image_std": [
15
+ 0.5,
16
+ 0.5,
17
+ 0.5
18
+ ],
19
+ "pan_and_scan_max_num_crops": null,
20
+ "pan_and_scan_min_crop_size": null,
21
+ "pan_and_scan_min_ratio_to_activate": null,
22
+ "processor_class": "Gemma3Processor",
23
+ "resample": 2,
24
+ "rescale_factor": 0.00392156862745098,
25
+ "size": {
26
+ "height": 896,
27
+ "width": 896
28
+ }
29
+ }
processor_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "image_seq_length": 256,
3
+ "processor_class": "Gemma3Processor"
4
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "boi_token": "<start_of_image>",
3
+ "bos_token": {
4
+ "content": "<bos>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false
9
+ },
10
+ "eoi_token": "<end_of_image>",
11
+ "eos_token": {
12
+ "content": "<eos>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false
17
+ },
18
+ "image_token": "<image_soft_token>",
19
+ "pad_token": {
20
+ "content": "<pad>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false
25
+ },
26
+ "unk_token": {
27
+ "content": "<unk>",
28
+ "lstrip": false,
29
+ "normalized": false,
30
+ "rstrip": false,
31
+ "single_word": false
32
+ }
33
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4667f2089529e8e7657cfb6d1c19910ae71ff5f28aa7ab2ff2763330affad795
3
+ size 33384568
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
3
+ size 4689074
tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff