Upload MM_LLMs

config.json

@@ -1,9 +1,8 @@
 {
-  "_name_or_path": "multimodal-tinyllama-whisper-small-siglip-v1/checkpoint-16700",
+  "_name_or_path": "multimodal-tinyllama-whisper-small-siglip/checkpoint-400",
   "architectures": [
     "MM_LLMs"
   ],
-  "attention_heads": 8,
   "audio_config": {
     "_name_or_path": "mesolitica/malaysian-whisper-small",
     "activation_dropout": 0.0,
@@ -204,11 +203,9 @@
     "use_weighted_layer_sum": false,
     "vocab_size": 51865
   },
-  "audio_conv_kernel": 240,
-  "audio_conv_stride": 220,
   "hidden_size": 2048,
   "image_config": {
-    "_name_or_path": "google/siglip-base-patch16-224",
+    "_name_or_path": "google/siglip-base-patch16-384",
     "add_cross_attention": false,
     "architectures": [
       "SiglipModel"
@@ -370,7 +367,7 @@
         "0": "LABEL_0",
         "1": "LABEL_1"
       },
-      "image_size": 224,
+      "image_size": 384,
       "intermediate_size": 3072,
       "is_decoder": false,
       "is_encoder_decoder": false,
@@ -418,8 +415,6 @@
       "use_bfloat16": false
     }
   },
-  "image_conv_kernel": 48,
-  "image_conv_stride": 36,
   "llm_config": {
     "_name_or_path": "mesolitica/malaysian-tinyllama-1.1b-16k-instructions-v3",
     "add_cross_attention": false,
@@ -502,7 +497,6 @@
     "vocab_size": 32004
   },
   "model_type": "mm_llms",
-  "n_frames": 6,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.37.1"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c00b609cae60891f2a3c1dbbd123dc28ad4cc0f1f799f7fcb7da6666e7e7a58
-size 3504434068
+oid sha256:d4805b652684d9bb425f952ec1e95324e7ddafa2ad5e3d7266fca43c3652ec60
+size 3236402604