Upload Gemma2ForCausalLM
Browse files- config.json +5 -2
- generation_config.json +1 -2
- model.safetensors +1 -1
config.json
CHANGED
|
@@ -6,7 +6,6 @@
|
|
| 6 |
"attention_dropout": 0.0,
|
| 7 |
"attn_logit_softcapping": 50.0,
|
| 8 |
"bos_token_id": 2,
|
| 9 |
-
"cache_implementation": "hybrid",
|
| 10 |
"eos_token_id": 1,
|
| 11 |
"final_logit_softcapping": 30.0,
|
| 12 |
"head_dim": 256,
|
|
@@ -14,6 +13,10 @@
|
|
| 14 |
"hidden_size": 8,
|
| 15 |
"initializer_range": 0.02,
|
| 16 |
"intermediate_size": 32,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
"max_position_embeddings": 8192,
|
| 18 |
"model_type": "gemma2",
|
| 19 |
"num_attention_heads": 4,
|
|
@@ -25,7 +28,7 @@
|
|
| 25 |
"rope_theta": 10000.0,
|
| 26 |
"sliding_window": 4096,
|
| 27 |
"torch_dtype": "float32",
|
| 28 |
-
"transformers_version": "4.
|
| 29 |
"use_cache": true,
|
| 30 |
"vocab_size": 256249
|
| 31 |
}
|
|
|
|
| 6 |
"attention_dropout": 0.0,
|
| 7 |
"attn_logit_softcapping": 50.0,
|
| 8 |
"bos_token_id": 2,
|
|
|
|
| 9 |
"eos_token_id": 1,
|
| 10 |
"final_logit_softcapping": 30.0,
|
| 11 |
"head_dim": 256,
|
|
|
|
| 13 |
"hidden_size": 8,
|
| 14 |
"initializer_range": 0.02,
|
| 15 |
"intermediate_size": 32,
|
| 16 |
+
"layer_types": [
|
| 17 |
+
"sliding_attention",
|
| 18 |
+
"full_attention"
|
| 19 |
+
],
|
| 20 |
"max_position_embeddings": 8192,
|
| 21 |
"model_type": "gemma2",
|
| 22 |
"num_attention_heads": 4,
|
|
|
|
| 28 |
"rope_theta": 10000.0,
|
| 29 |
"sliding_window": 4096,
|
| 30 |
"torch_dtype": "float32",
|
| 31 |
+
"transformers_version": "4.55.0.dev0",
|
| 32 |
"use_cache": true,
|
| 33 |
"vocab_size": 256249
|
| 34 |
}
|
generation_config.json
CHANGED
|
@@ -1,8 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"_from_model_config": true,
|
| 3 |
"bos_token_id": 2,
|
| 4 |
-
"cache_implementation": "hybrid",
|
| 5 |
"eos_token_id": 1,
|
| 6 |
"pad_token_id": 0,
|
| 7 |
-
"transformers_version": "4.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"_from_model_config": true,
|
| 3 |
"bos_token_id": 2,
|
|
|
|
| 4 |
"eos_token_id": 1,
|
| 5 |
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.55.0.dev0"
|
| 7 |
}
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 8405552
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3c553714d961ce15585363bf35e318a9a32ad55182f2b70a6ddf55cd4866a656
|
| 3 |
size 8405552
|