fix: Fill in input Tensor

UbiquitousLearning · Mar 26, 2024 · e6fd6e0 · e6fd6e0
1 parent f9c9c3f
commit e6fd6e0
Show file tree

Hide file tree

Showing 3 changed files with 16 additions and 13 deletions.
diff --git a/README.md b/README.md
@@ -58,14 +58,14 @@ Wait.. why on-device multimodal LLM? - It's a key building block for [intelligen
 
 |                                                                             | FP32 | INT4 |
 |-----------------------------------------------------------------------------|-----|------|
-| [LLaMA-1/2 7B](https://github.com/facebookresearch/llama)                   | ✔️  | ✔️   |
-| [Alpaca 7B](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2)                | ✔️  | ✔️   |
-| [TinyLLaMA 1.1B](https://github.com/jzhang38/TinyLlama)                     | ✔️  | ✔️   |
-| [Fuyu 8B](https://www.adept.ai/blog/fuyu-8b)                                | ✔️  | ✔️   |
-| [Vision Transformer](https://github.com/google-research/vision_transformer) | ✔️  | ✔️   |
-| [CLIP](https://github.com/openai/CLIP)                                      | ✔️  | ✔️   |
-| [ImageBind](https://github.com/facebookresearch/ImageBind) (3 modalities)   | ✔️  | ✔️   |
-| [LLaVA 7B](https://github.com/haotian-liu/LLaVA)                                                                | ✔️  | ✔️   |
+| [LLaMA-1/2 7B](https://github.com/facebookresearch/llama)                   | [✔️](https://huggingface.co/mllmTeam/llama-2-7b-mllm/tree/main)  | [✔️](https://huggingface.co/mllmTeam/llama-2-7b-mllm/tree/main)   |
+| [Alpaca 7B](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2)                | [✔️](https://huggingface.co/mllmTeam/chinese-alpaca-7b-mllm/tree/main)  | [✔️](https://huggingface.co/mllmTeam/chinese-alpaca-7b-mllm/tree/main)   |
+| [TinyLLaMA 1.1B](https://github.com/jzhang38/TinyLlama)                     | [✔️](https://huggingface.co/mllmTeam/tinyllama-1.1b-mllm/tree/main)  | [✔️](https://huggingface.co/mllmTeam/tinyllama-1.1b-mllm/tree/main)   |
+| [Fuyu 8B](https://www.adept.ai/blog/fuyu-8b)                                | [✔️](https://huggingface.co/mllmTeam/fuyu-8b-mllm/tree/main)  | [✔️](https://huggingface.co/mllmTeam/fuyu-8b-mllm/tree/main)   |
+| [Vision Transformer](https://github.com/google-research/vision_transformer) | [✔️](https://huggingface.co/mllmTeam/vit-base-patch16-224-mllm/tree/main)  | [✔️](https://huggingface.co/mllmTeam/vit-base-patch16-224-mllm/tree/main)   |
+| [CLIP](https://github.com/openai/CLIP)                                      | [✔️](https://huggingface.co/mllmTeam/clip-vit-base-patch32-mllm/tree/main)  | [✔️](https://huggingface.co/mllmTeam/clip-vit-base-patch32-mllm/tree/main)   |
+| [ImageBind](https://github.com/facebookresearch/ImageBind) (3 modalities)   | [✔️](https://huggingface.co/mllmTeam/imagebind_huge-mllm/tree/main)  | [✔️](https://huggingface.co/mllmTeam/imagebind_huge-mllm/tree/main)   |
+| [LLaVA 7B](https://github.com/haotian-liu/LLaVA)                                                                | [✔️](https://huggingface.co/mllmTeam/llava-1.5-7b-mllm/tree/main)  | [✔️](https://huggingface.co/mllmTeam/llava-1.5-7b-mllm/tree/main)   |
 
 ## Quick Start
 

diff --git a/src/Layer.hpp b/src/Layer.hpp
@@ -178,11 +178,11 @@ class Layer {
             }
             switch (input0.status()) {
             case TENSOR_STATIC_INIT: {
-                if (Tensor::gph_.find(input0.name()) == Tensor::gph_.end()) {
+                if (Tensor::gph_.find(input0.name()) == Tensor::gph_.end()|| input0.count() != Tensor::gph_[input0.name()].count()) {
                     Tensor::gph_[input0.name()] = input0;
                     Tensor::gph_[input0.name()].setName(input0.name());
                 }
-                if (Tensor::gph_.find(input1.name()) == Tensor::gph_.end()) {
+                if (Tensor::gph_.find(input1.name()) == Tensor::gph_.end()|| input1.count() != Tensor::gph_[input1.name()].count()) {
                     Tensor::gph_[input1.name()] = input1;
                     Tensor::gph_[input1.name()].setName(input1.name());
                 }
@@ -248,15 +248,15 @@ class Layer {
             }
             switch (input0.status()) {
             case TENSOR_STATIC_INIT: {
-                if (Tensor::gph_.find(input0.name()) == Tensor::gph_.end()) {
+                if (Tensor::gph_.find(input0.name()) == Tensor::gph_.end() || input0.count() != Tensor::gph_[input0.name()].count()) {
                     Tensor::gph_[input0.name()] = input0;
                     Tensor::gph_[input0.name()].setName(input0.name());
                 }
-                if (Tensor::gph_.find(input1.name()) == Tensor::gph_.end()) {
+                if (Tensor::gph_.find(input1.name()) == Tensor::gph_.end()|| input1.count() != Tensor::gph_[input1.name()].count()) {
                     Tensor::gph_[input1.name()] = input1;
                     Tensor::gph_[input1.name()].setName(input1.name());
                 }
-                if (Tensor::gph_.find(input2.name()) == Tensor::gph_.end()) {
+                if (Tensor::gph_.find(input2.name()) == Tensor::gph_.end()|| input2.count() != Tensor::gph_[input0.name()].count()) {
                     Tensor::gph_[input2.name()] = input2;
                     Tensor::gph_[input2.name()].setName(input2.name());
                 }

diff --git a/src/Module.hpp b/src/Module.hpp
@@ -78,6 +78,9 @@ class Module {
             for (auto &input : inputs) {
                 input.setTtype(TensorType::NORMAL_TENSOR);
                 input.status() = TENSOR_STATIC_INIT;
+                if(input.batch() == 0){
+                    Tensor::gph_[input.name()] = input;
+                }
             }
             tensor_status = TENSOR_STATIC_INIT;