Merge pull request #30 from YerevaNN/minor_changes_for_run

minor changes for run

chemlactica/config/config_yamls/galactica_125m_pretrain_config.yaml

@@ -13,7 +13,7 @@ train_config:
   bf16_full_eval: true
   fp16: false
   tf32: true
-  evaluation_strategy: "steps"
+  evaluation_strategy: "no"
   save_total_limit: 8
   grad_accumulation_scheduler: false
   dynamic_grad_accumulation: false
@@ -27,4 +27,5 @@ model_config:
   block_size: 2048
   vocab_size: 50000
   separator_token: </s>
+  separator_token_id: 2
   tokenizer_path: "./chemlactica/tokenizer/ChemLacticaTokenizer66"

chemlactica/utils/dataset_utils.py

@@ -73,14 +73,15 @@ def process_str(str, random_number_generator, model_config):
     # it's wierd workaround but works for now
     try:
         compound = load_jsonl_line(str["text"])
+        compound = delete_empty_tags(compound)
+        str["text"] = generate_formatted_string(
+            compound, random_number_generator, model_config
+        )
+        string = str
     except Exception as e:
         print(e)
-        return ""
-    compound = delete_empty_tags(compound)
-    str["text"] = generate_formatted_string(
-        compound, random_number_generator, model_config
-    )
-    return str
+        string = ""
+    return string
 
 
 def group_texts(examples, model_config, eos_token_id):

chemlactica/utils/text_format_utils.py

@@ -105,41 +105,36 @@ def generate_formatted_string(compound_json, rng, model_config):
 
 
 def format_key_value(key, value, rng):
-    try:
-        if key == "CID":
-            return ""
-        formatted_string = ""
-        if key == "related":
-            if len(value) > 10:
-                # value = random.sample(value, 5)
-                value = rng.choice(value, size=10, replace=False, shuffle=False)
-            for pair in value:
-                rounded_sim = "{:.2f}".format(float(pair["similarity"]))
-                formatted_string += f"{SPECIAL_TAGS['similarity']['start']}{pair['SMILES']} {rounded_sim}{SPECIAL_TAGS['similarity']['end']}"  # noqa
-        elif key == "experimental":
-            for pair in value:
-                formatted_string += f"[PROPERTY]{pair['PROPERTY_NAME']} {pair['PROPERTY_VALUE']}[/PROPERTY]"  # noqa
-        elif key == "synonyms":
-            for val in value:
-                formatted_string += f"{SPECIAL_TAGS['synonym']['start']}{val['name']}{SPECIAL_TAGS['synonym']['end']}"  # noqa
-        else:
-            try:
-                if SPECIAL_TAGS[key].get("type") is float:
-                    value = "{:.2f}".format(float(value))
-                    assert len(value.split(".")[-1]) == 2
-                start = SPECIAL_TAGS[key]["start"]
-                end = SPECIAL_TAGS[key]["end"]
-            except Exception as e:
-                print(e)
-                print("Failed to parse: ", key, value)
-                start = value = end = ""
-            return f"{start}{value}{end}"
-
-        return formatted_string
-    except Exception as e:
-        print(e)
-        print("Failed to parse: ", key, value)
+    if key == "CID":
         return ""
+    formatted_string = ""
+    if key == "related":
+        if len(value) > 10:
+            # value = random.sample(value, 5)
+            value = rng.choice(value, size=10, replace=False, shuffle=False)
+        for pair in value:
+            rounded_sim = "{:.2f}".format(float(pair["similarity"]))
+            formatted_string += f"{SPECIAL_TAGS['similarity']['start']}{pair['SMILES']} {rounded_sim}{SPECIAL_TAGS['similarity']['end']}"  # noqa
+    elif key == "experimental":
+        for pair in value:
+            formatted_string += f"[PROPERTY]{pair['PROPERTY_NAME']} {pair['PROPERTY_VALUE']}[/PROPERTY]"  # noqa
+    elif key == "synonyms":
+        for val in value:
+            formatted_string += f"{SPECIAL_TAGS['synonym']['start']}{val['name']}{SPECIAL_TAGS['synonym']['end']}"  # noqa
+    else:
+        try:
+            if SPECIAL_TAGS[key].get("type") is float:
+                value = "{:.2f}".format(float(value))
+                assert len(value.split(".")[-1]) == 2
+            start = SPECIAL_TAGS[key]["start"]
+            end = SPECIAL_TAGS[key]["end"]
+        except Exception as e:
+            print(e)
+            print("Failed to parse: ", key, value)
+            start = value = end = ""
+        return f"{start}{value}{end}"
+
+    return formatted_string
 
 
 def main():

submit_run.py

@@ -13,15 +13,15 @@
 model_size = "125m"
 train_type = "pretrain"
 train_name = "_".join([model_name, model_size, train_type])
-job_name = "gal_relform2"
+job_name = "gal_relform3"
 
 slurm_params = {
     "slurm_job_name": job_name,
     "timeout_min": 60 * 24 * 2,
     "nodes": 1,
     "tasks_per_node": 1,
     "gpus_per_node": num_gpus,
-    "cpus_per_task": num_gpus * 20,
+    "cpus_per_task": num_gpus * 10,
     "mem_gb": num_gpus * 40.0 + 20.0,
     "stderr_to_stdout": True,
 }
@@ -41,10 +41,10 @@
     "dir_data_types": "computed",
     "training_data_dirs": "/nfs/ap/mnt/sxtn/rdkit_computed_rel+form/train_rdkit_computed_rel+form",
     "valid_data_dir": "/nfs/ap/mnt/sxtn/rdkit_computed_rel+form/valid_rdkit_computed_rel+form",
-    "max_steps": 20000,
+    "max_steps": 19000,
     # "num_train_epochs": 24,
-    "eval_steps": 2000,
-    "save_steps": 2000,
+    "eval_steps": 0,
+    "save_steps": 6300,
     "train_batch_size": 16,
     "valid_batch_size": 16,
     "dataloader_num_workers": 1,

test_status.yaml

@@ -1 +1 @@
-33f0567f336bd041b6b687cb3258a855d948b6b8: PASS
+210ba73adff95aacc5f6c215dc6218923d85f767: FAIL