Add model weights and configs

README.md

@@ -1,3 +1,66 @@
 ---
 license: cc-by-nc-sa-4.0
+tags:
+- Helical
+- RNA
+- Biology
+- Transformers
+- Genomics
+- Mamba2
+- Sequence
+library_name: transformers
 ---
+# Mamba2-mRNA
+Mamba2-mRNA is a state-space model built on the Mamba2 architecture, trained at single-nucleotide resolution. This innovative model offers several advantages, including faster processing speeds compared to traditional transformer models, efficient handling of long sequences, and reduced memory requirements. Its state-space approach enables better modeling of biological sequences by capturing both local and long-range dependencies in mRNA data. The single-nucleotide resolution allows for precise prediction and analysis of genetic elements.
+
+# Helical<a name="helical"></a>
+
+#### Install the package
+
+Run the following to install the [Helical](https://github.com/helicalAI/helical) package via pip:
+```console
+pip install --upgrade helical
+```
+
+#### Generate Embeddings
+```python
+from helical import Mamba2mRNA, Mamba2mRNAConfig
+import torch
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+input_sequences = ["ACU"*20, "AUG"*20, "AUG"*20, "ACU"*20, "AUU"*20]
+
+mamba2_mrna_config = Mamba2mRNAConfig(batch_size=5, device=device)
+mamba2_mrna = Mamba2mRNA(configurer=mamba2_mrna_config)
+
+# prepare data for input to the model
+processed_input_data = mamba2_mrna.process_data(input_sequences)
+
+# generate the embeddings for the input data
+embeddings = mamba2_mrna.get_embeddings(processed_input_data)
+```
+
+#### Fine-Tuning
+Classification fine-tuning example:
+```python
+from helical import Mamba2mRNAFineTuningModel, Mamba2mRNAConfig
+import torch
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+input_sequences = ["ACU"*20, "AUG"*20, "AUG"*20, "ACU"*20, "AUU"*20]
+labels = [0, 2, 2, 0, 1]
+
+mamba2_mrna_config = Mamba2mRNAConfig(batch_size=5, device=device, max_length=100)
+mamba2_mrna_fine_tune = Mamba2mRNAFineTuningModel(mamba2_mrna_config=mamba2_mrna_config, fine_tuning_head="classification", output_size=3)
+
+# prepare data for input to the model
+train_dataset = mamba2_mrna_fine_tune.process_data(input_sequences)
+
+# fine-tune the model with the relevant training labels
+mamba2_mrna_fine_tune.train(train_dataset=train_dataset, train_labels=labels)
+
+# get outputs from the fine-tuned model on a processed dataset
+outputs = mamba2_mrna_fine_tune.get_outputs(train_dataset)
+```

config.json

@@ -0,0 +1,39 @@
+{
+    "architectures": [
+      "Mamba2ForCausalLM"
+    ],
+    "bos_token_id": 0,
+    "chunk_size": 256,
+    "conv_kernel": 4,
+    "eos_token_id": 2,
+    "expand": 2,
+    "head_dim": 32,
+    "hidden_act": "silu",
+    "hidden_size": 512,
+    "initializer_range": 0.1,
+    "layer_norm_epsilon": 1e-05,
+    "model_type": "mamba2",
+    "n_groups": 1,
+    "num_heads": 32,
+    "num_hidden_layers": 16,
+    "pad_token_id": 1,
+    "rescale_prenorm_residual": false,
+    "residual_in_fp32": true,
+    "rms_norm": true,
+    "state_size": 128,
+    "tie_word_embeddings": false,
+    "time_step_floor": 0.0001,
+    "time_step_limit": [
+      0.0,
+      Infinity
+    ],
+    "time_step_max": 0.1,
+    "time_step_min": 0.001,
+    "time_step_rank": 32,
+    "torch_dtype": "float32",
+    "transformers_version": "4.45.1",
+    "use_bias": false,
+    "use_cache": true,
+    "use_conv_bias": true,
+    "vocab_size": 13
+  }

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e890c76d8e9775c00ccd7c12a988ea5aa90d8630d5cd6c52dce481acff98d6c0
+size 110684944

special_tokens_map.json

@@ -0,0 +1,9 @@
+{
+    "bos_token": "[BOS]",
+    "cls_token": "[CLS]",
+    "eos_token": "[SEP]",
+    "mask_token": "[MASK]",
+    "pad_token": "[PAD]",
+    "sep_token": "[SEP]",
+    "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,64 @@
+{
+    "add_prefix_space": false,
+    "added_tokens_decoder": {
+      "0": {
+        "content": "[CLS]",
+        "lstrip": false,
+        "normalized": false,
+        "rstrip": false,
+        "single_word": false,
+        "special": true
+      },
+      "1": {
+        "content": "[SEP]",
+        "lstrip": false,
+        "normalized": false,
+        "rstrip": false,
+        "single_word": false,
+        "special": true
+      },
+      "2": {
+        "content": "[BOS]",
+        "lstrip": false,
+        "normalized": false,
+        "rstrip": false,
+        "single_word": false,
+        "special": true
+      },
+      "3": {
+        "content": "[MASK]",
+        "lstrip": false,
+        "normalized": false,
+        "rstrip": false,
+        "single_word": false,
+        "special": true
+      },
+      "4": {
+        "content": "[PAD]",
+        "lstrip": false,
+        "normalized": false,
+        "rstrip": false,
+        "single_word": false,
+        "special": true
+      },
+      "6": {
+        "content": "[UNK]",
+        "lstrip": false,
+        "normalized": false,
+        "rstrip": false,
+        "single_word": false,
+        "special": true
+      }
+    },
+    "bos_token": "[BOS]",
+    "clean_up_tokenization_spaces": false,
+    "cls_token": "[CLS]",
+    "eos_token": "[SEP]",
+    "mask_token": "[MASK]",
+    "model_max_length": 8000,
+    "pad_token": "[PAD]",
+    "padding_side": "left",
+    "sep_token": "[SEP]",
+    "tokenizer_class": "CharTokenizer",
+    "unk_token": "[UNK]"
+  }