init model

config.json

@@ -0,0 +1,44 @@
+{
+  "_name_or_path": "/mnt/output/pretrained_models/phi-1_5_dev",
+  "architectures": [
+    "PhiForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "embd_pdrop": 0.0,
+  "eos_token_id": 2,
+  "freeze_mm_mlp_adapter": false,
+  "hidden_act": "gelu_new",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_grid_pinpoints": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 8192,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 2048,
+  "mm_hidden_size": 256,
+  "mm_input_embeds": 1024,
+  "mm_projector_type": "swin_conv",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_tower": "/path/to/swinbase/model.pkl",
+  "model_type": "llava_phi",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 24,
+  "partial_rotary_factor": 0.5,
+  "pretraining_tp": 1,
+  "projector_outdim": 2048,
+  "qk_layernorm": false,
+  "resid_pdrop": 0.0,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "swin_type": "base",
+  "tie_word_embeddings": false,
+  "transformers_version": "4.36.2",
+  "tune_mm_mlp_adapter": true,
+  "use_cache": true,
+  "use_mm_proj": true,
+  "vocab_size": 50296,
+  "with_layernorm": false,
+  "with_norm": true
+}

mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6edaacf5e1173ed1450798e1a00f3c5e3448697ec2b30dbe33b18397b2370d10
+size 125861573