Hugging Face's logo

pytorch
/
gemma-3-27b-it-INT4

Transformers
PyTorch
Safetensors
English
torchao
gemma3
Model card Files
xet
 Community
appy1234 commited on
Commit
158ed2b
·
verified ·
1 Parent(s): 70205b1

Upload Gemma3ForConditionalGeneration

Browse files
Files changed (3) hide show
  1. config.json +17 -20
  2. generation_config.json +1 -1
  3. model.safetensors +3 -0
config.json CHANGED
@@ -23,20 +23,7 @@
23
  "group_size": 128,
24
  "int4_choose_qparams_algorithm": "hqq",
25
  "int4_packing_format": "tile_packed_to_4d",
26
- "layout": {
27
- "_data": {
28
- "inner_k_tiles": 8
29
- },
30
- "_type": "TensorCoreTiledLayout",
31
- "_version": 1
32
- },
33
- "preserve_zero": null,
34
- "set_inductor_config": true,
35
- "use_hqq": false,
36
- "zero_point_domain": {
37
- "_data": "NONE",
38
- "_type": "ZeroPointDomain"
39
- }
40
  },
41
  "_type": "Int4WeightOnlyConfig",
42
  "_version": 2
@@ -50,7 +37,9 @@
50
  "attention_bias": false,
51
  "attention_dropout": 0.0,
52
  "attn_logit_softcapping": null,
 
53
  "dtype": "bfloat16",
 
54
  "final_logit_softcapping": null,
55
  "head_dim": 128,
56
  "hidden_activation": "gelu_pytorch_tanh",
@@ -126,20 +115,28 @@
126
  "num_attention_heads": 32,
127
  "num_hidden_layers": 62,
128
  "num_key_value_heads": 16,
 
129
  "query_pre_attn_scalar": 168,
130
  "rms_norm_eps": 1e-06,
131
- "rope_local_base_freq": 10000.0,
132
- "rope_scaling": {
133
- "factor": 8.0,
134
- "rope_type": "linear"
 
 
 
 
 
 
135
  },
136
- "rope_theta": 1000000.0,
137
  "sliding_window": 1024,
 
138
  "use_bidirectional_attention": false,
139
  "use_cache": true,
140
  "vocab_size": 262208
141
  },
142
- "transformers_version": "4.57.6",
 
143
  "vision_config": {
144
  "attention_dropout": 0.0,
145
  "dtype": "bfloat16",
 
23
  "group_size": 128,
24
  "int4_choose_qparams_algorithm": "hqq",
25
  "int4_packing_format": "tile_packed_to_4d",
26
+ "set_inductor_config": true
 
 
 
 
 
 
 
 
 
 
 
 
 
27
  },
28
  "_type": "Int4WeightOnlyConfig",
29
  "_version": 2
 
37
  "attention_bias": false,
38
  "attention_dropout": 0.0,
39
  "attn_logit_softcapping": null,
40
+ "bos_token_id": 2,
41
  "dtype": "bfloat16",
42
+ "eos_token_id": 1,
43
  "final_logit_softcapping": null,
44
  "head_dim": 128,
45
  "hidden_activation": "gelu_pytorch_tanh",
 
115
  "num_attention_heads": 32,
116
  "num_hidden_layers": 62,
117
  "num_key_value_heads": 16,
118
+ "pad_token_id": 0,
119
  "query_pre_attn_scalar": 168,
120
  "rms_norm_eps": 1e-06,
121
+ "rope_parameters": {
122
+ "full_attention": {
123
+ "factor": 8.0,
124
+ "rope_theta": 1000000.0,
125
+ "rope_type": "linear"
126
+ },
127
+ "sliding_attention": {
128
+ "rope_theta": 10000.0,
129
+ "rope_type": "default"
130
+ }
131
  },
 
132
  "sliding_window": 1024,
133
+ "tie_word_embeddings": true,
134
  "use_bidirectional_attention": false,
135
  "use_cache": true,
136
  "vocab_size": 262208
137
  },
138
+ "tie_word_embeddings": true,
139
+ "transformers_version": "5.3.0.dev0",
140
  "vision_config": {
141
  "attention_dropout": 0.0,
142
  "dtype": "bfloat16",
generation_config.json CHANGED
@@ -9,5 +9,5 @@
9
  "pad_token_id": 0,
10
  "top_k": 64,
11
  "top_p": 0.95,
12
- "transformers_version": "4.57.6"
13
  }
 
9
  "pad_token_id": 0,
10
  "top_k": 64,
11
  "top_p": 0.95,
12
+ "transformers_version": "5.3.0.dev0"
13
  }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3eaeb78a03ce2f09090ef1114f65e1113ff57b5b7030ecb4aff3cfd424db21f
3
+ size 18270308104