Update to v2

ml4bio · Apr 20, 2024 · a9c8c39 · a9c8c39
1 parent d466905
commit a9c8c39
Show file tree

Hide file tree

Showing 20 changed files with 643 additions and 980 deletions.
diff --git a/.gitignore b/.gitignore
@@ -1,2 +1,6 @@
-
+*.pt
 *.ckpt
+__pycache__/
+*.py[cod]
+*$py.class
+*.DS_Store
diff --git a/FastMsaApp/__init__.py b/FastMsaApp/__init__.py
diff --git a/FastMsaApp/asgi.py b/FastMsaApp/asgi.py
diff --git a/FastMsaApp/fm_app.py b/FastMsaApp/fm_app.py
diff --git a/FastMsaApp/settings.py b/FastMsaApp/settings.py
diff --git a/FastMsaApp/thread_util.py b/FastMsaApp/thread_util.py
diff --git a/FastMsaApp/urls.py b/FastMsaApp/urls.py
diff --git a/FastMsaApp/wsgi.py b/FastMsaApp/wsgi.py
diff --git a/README.md b/README.md
@@ -41,42 +41,12 @@ optional arguments:
 * output_path: output dir -- seq/db/res, seq subdir contain all queries, db contain retrieved db, res contain all results
 * database_path: directory containing database in DataFrame and embedding saved in faiss index. All results produced in Offline Embedding section.
 
-
-
-## Server Mode Usage (Advanced Usage)
-
-For the server administor, Django is required to deploy the server like this:
-
-```
-nohup python3 manage.py runserver 0.0.0.0:7077 >> nohup_xxx.out 2>&1 &
-```
-
-Manage the backend's max workers
-
-Find the max_worker config(FastMsaApp/fm_app.py: line 32), change the value.
-The number of max_workers is recommend is `32`.
-
-
-
-Send a request by http content: 
-```
-# Use curl
-curl --location --request POST 'http://0.0.0.0:7077/fastmsa' \
---header 'Content-Type: application/x-www-form-urlencoded' \
---data-urlencode 'input=/user/linmingzhi/project/fastmsa/example/1pazA.fasta' \
---data-urlencode 'output=/user/linmingzhi/output/fastmsa/test026' \
---data-urlencode 'tarnum=320000'
+## Structure prediction (Optional)
+Install ColabFold
 ```
-
-## Server Test Mode (Advanced Usage)
-```
-# Run a server instance hang on
-python manage.py runserver 0.0.0.0:7077
+pip install -q --no-warn-conflicts 'colabfold[alphafold-minus-jax] @ git+https://github.com/sokrypton/ColabFold
 ```
-
-## One-line Submission Command (Advanced Usage)
-
-For fastMSA only, try the following one-line command:
+Run batch prediction
 ```
-./fastMSA.sh -i example/1pazA.fasta
+colabfold_batch $MSA_DIR $PREDICTION_RES
 ```
diff --git a/conf/scale_conf.yaml b/conf/scale_conf.yaml
@@ -0,0 +1,30 @@
+trainer:
+  master_port: "39533"
+  devices: "0"
+  gpus: [0]
+  seed: 1234
+  acc_step: 4
+  accelerator: "cuda"
+  strategy: ddp_find_unused_parameters_true
+  precision: 16
+  batch_size: 600 
+  max_epochs: 100
+  ur90_path: /mnt/data/aim/hongliang/uniref30_202103/uniref30_2103.tsv 
+
+model:
+  resume: False
+  ckpt_path: ./ckpt 
+
+callback:
+  monitor: val_acc
+  dirpath: ./model_uni/
+  filename: pl_biencoder-{epoch:03d}-{val_acc:.4f}
+  mode: max
+  save_top_k: 5
+  save_on_train_epoch_end: True
+
+logger:
+  log_every_n_steps: 10
+  use_wandb: False
+  project: uniclust_pl
+  log_model: False
diff --git a/db.sqlite3 b/db.sqlite3