Update model with LoRA adapter
Browse files
adapter_config.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"has_adapter": true, "adapter_path": "speech-lora", "adapter_name": "speech"}
|
config.json
CHANGED
|
@@ -56,6 +56,13 @@
|
|
| 56 |
"initializer_range": 0.02,
|
| 57 |
"mm_tokens_per_image": 256,
|
| 58 |
"model_type": "gemma3",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
"text_config": {
|
| 60 |
"attention_bias": false,
|
| 61 |
"attention_dropout": 0.0,
|
|
@@ -88,7 +95,6 @@
|
|
| 88 |
},
|
| 89 |
"torch_dtype": "bfloat16",
|
| 90 |
"transformers_version": "4.50.0.dev0",
|
| 91 |
-
"use_cache": false,
|
| 92 |
"vision_config": {
|
| 93 |
"attention_dropout": 0.0,
|
| 94 |
"hidden_act": "gelu_pytorch_tanh",
|
|
|
|
| 56 |
"initializer_range": 0.02,
|
| 57 |
"mm_tokens_per_image": 256,
|
| 58 |
"model_type": "gemma3",
|
| 59 |
+
"speech_lora": {
|
| 60 |
+
"dp": 0.01,
|
| 61 |
+
"layer": "((layers.*self_attn\\.(q|k|v|o)_proj)|(layers.*mlp\\.(gate|up|down)_proj))",
|
| 62 |
+
"lora_alpha": 320,
|
| 63 |
+
"r": 320,
|
| 64 |
+
"use_rslora": true
|
| 65 |
+
},
|
| 66 |
"text_config": {
|
| 67 |
"attention_bias": false,
|
| 68 |
"attention_dropout": 0.0,
|
|
|
|
| 95 |
},
|
| 96 |
"torch_dtype": "bfloat16",
|
| 97 |
"transformers_version": "4.50.0.dev0",
|
|
|
|
| 98 |
"vision_config": {
|
| 99 |
"attention_dropout": 0.0,
|
| 100 |
"hidden_act": "gelu_pytorch_tanh",
|
model-00001-of-00002.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4991785040
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a1ce1cb9238b56dba8ac52c728d2e62854a3b0dc601e64bd5e5b047a69bd3630
|
| 3 |
size 4991785040
|
{speech-adapter → speech-lora}/README.md
RENAMED
|
File without changes
|
{speech-adapter → speech-lora/speech}/adapter_config.json
RENAMED
|
@@ -12,25 +12,17 @@
|
|
| 12 |
"layers_pattern": null,
|
| 13 |
"layers_to_transform": null,
|
| 14 |
"loftq_config": {},
|
| 15 |
-
"lora_alpha":
|
| 16 |
"lora_bias": false,
|
| 17 |
-
"lora_dropout": 0.
|
| 18 |
"megatron_config": null,
|
| 19 |
"megatron_core": "megatron.core",
|
| 20 |
"modules_to_save": null,
|
| 21 |
"peft_type": "LORA",
|
| 22 |
-
"r":
|
| 23 |
"rank_pattern": {},
|
| 24 |
"revision": null,
|
| 25 |
-
"target_modules":
|
| 26 |
-
"q_proj",
|
| 27 |
-
"down_proj",
|
| 28 |
-
"k_proj",
|
| 29 |
-
"gate_proj",
|
| 30 |
-
"o_proj",
|
| 31 |
-
"v_proj",
|
| 32 |
-
"up_proj"
|
| 33 |
-
],
|
| 34 |
"task_type": "CAUSAL_LM",
|
| 35 |
"use_dora": false,
|
| 36 |
"use_rslora": true
|
|
|
|
| 12 |
"layers_pattern": null,
|
| 13 |
"layers_to_transform": null,
|
| 14 |
"loftq_config": {},
|
| 15 |
+
"lora_alpha": 320,
|
| 16 |
"lora_bias": false,
|
| 17 |
+
"lora_dropout": 0.01,
|
| 18 |
"megatron_config": null,
|
| 19 |
"megatron_core": "megatron.core",
|
| 20 |
"modules_to_save": null,
|
| 21 |
"peft_type": "LORA",
|
| 22 |
+
"r": 320,
|
| 23 |
"rank_pattern": {},
|
| 24 |
"revision": null,
|
| 25 |
+
"target_modules": "((layers.*self_attn\\.(q|k|v|o)_proj)|(layers.*mlp\\.(gate|up|down)_proj))",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
"task_type": "CAUSAL_LM",
|
| 27 |
"use_dora": false,
|
| 28 |
"use_rslora": true
|
{speech-adapter → speech-lora/speech}/adapter_model.safetensors
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7aacfcd0281ad18acbaac412bb2a23469adeb909011b08b384d7374cd611d423
|
| 3 |
+
size 2384262280
|