initial commit

README.md

@@ -0,0 +1,6 @@
+# iSEEEK
+Generative pretraining from the rankings of top expressing genes. 
+It was trained on more than 20 million single-cell transcriptomes with a sequence length of 64.
+
+
+

config.json

@@ -0,0 +1,32 @@
+{
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 1,
+  "embd_pdrop": 0.1,
+  "eos_token_id": 2,
+  "gradient_checkpointing": false,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 1024,
+  "n_head": 16,
+  "n_inner": null,
+  "n_layer": 8,
+  "n_positions": 1024,
+  "pad_token_id": 0,
+  "resid_pdrop": 0.1,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.10.0",
+  "use_cache": true,
+  "vocab_size": 21150
+}

copy.sh

@@ -0,0 +1,3 @@
+cp /data/ai/gene_expression/gpt2/gpt2-1024-8-16/config.json ./
+cp /data/ai/gene_expression/gpt2/gpt2-1024-8-16/pytorch_model.bin ./
+cp /data/ai/gene_expression/gpt2/gpt2tokenizer/* ./

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b9373ec77bd186fd9d38f09eb5361e8f5c181655b0e0b868fb879a16553f178
+size 502345283

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "</b>", "eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"model_max_length": 1024, "pad_token": "<pad>", "eos_token": "</s>", "bos_token": "</b>", "unk_token": "<unk>", "tokenizer_class": "PreTrainedTokenizerFast"}