embedding

.zeno_cache/EMBEDDINGhuman-with-embeddings.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e2ddbd958723a349787a13b792e697688b04c1b1c057137db818af26c1936c9
+size 3477209

.zeno_cache/EMBEDDINGhuman.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca384c44b4e2b31f7912cadc6ebfcaf1b10c8f49f12bdf23b4d1412cac3686eb
+size 76193865

.zeno_cache/OUTPUThuman-with-embeddings.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f92db696d6c39e571601552125d0b4dd2a6382071394ce0693f71fafbdab5da
+size 280865

.zeno_cache/OUTPUThuman.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d4cdb204864c66d497e210a2349247ca45b498539088d27642b2d1c94f9b4b02
-size 549553
+oid sha256:ca48e73a22a1d00d38d1cea8115828aedf8907b22cc27e956005a83e028a94ea
+size 2965059

.zeno_cache/POSTDISTILLbert_scorehuman-with-embeddings.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:374317448ac7ec37fa499adfa775c4370a7cf3feca2854bc861810f075a398d5
+size 25744

.zeno_cache/POSTDISTILLbert_scorehuman.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4066a8d5da3f7e5cb0a259bc843a7ee88666ce876becdd684a68ec69c422362b
-size 50680
+oid sha256:b3874299934f57785c20ef632025c3caceb8cffc9b9029a3c026c7366be7fac4
+size 275525

.zeno_cache/PREDISTILLlength.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:24d45a20c33742deb9f8ed76718f01a9d2b5d35b7f4f371b55b1ed6e26f70b1c
-size 37806
+oid sha256:5a32550e54669e452e59e07de2437b1e7af9a6aa544d731eda77fccb6fdd45b2
+size 204736

config.toml

@@ -4,6 +4,6 @@ models = ["human"]
 metadata = "wmt20-de-en.tsv"
 data_column = "text"
 label_column = "label"
-batch_size = 36
+batch_size = 2000
 port = 7860 
 host = "0.0.0.0"

model.py

@@ -1,18 +1,18 @@
 from zeno import distill, model, metric, ZenoOptions
 from inspiredco.critique import Critique
 import os
+from sentence_transformers import SentenceTransformer
 
-# from sentence_transformers import SentenceTransformer
-
-# sentence_embed = SentenceTransformer("paraphrase-multilingual-mpnet-base-v2")
-# client = Critique(api_key=os.environ["INSPIREDCO_API_KEY"])
+client = Critique(api_key=os.environ["INSPIREDCO_API_KEY"])
 
 
 @model
 def pred_fns(name):
+    sentence_embed = SentenceTransformer("paraphrase-multilingual-mpnet-base-v2")
+
     def pred(df, ops):
-        return df["translation"]
-        # , sentence_embed.encode(df[ops.label_column].tolist())
+        embed = sentence_embed.encode(df[ops.data_column].tolist()).tolist()
+        return df["translation"], embed
 
     return pred
 
@@ -24,13 +24,11 @@ def bert_score(df, ops):
         d["references"] = [d.pop("label")]
         d["target"] = d.pop(ops.output_column)
 
-    return [0] * len(eval_dict)
-
-    # result = client.evaluate(
-    #     metric="bert_score", config={"model": "bert-base-uncased"}, dataset=eval_dict
-    # )
+    result = client.evaluate(
+        metric="bert_score", config={"model": "bert-base-uncased"}, dataset=eval_dict
+    )
 
-    # return [round(r["value"], 6) for r in result["examples"]]
+    return [round(r["value"], 6) for r in result["examples"]]
 
 
 @metric

requirements.txt

@@ -1,2 +1,3 @@
-zenoml>=0.3.12
-inspiredco
+zenoml>=0.3.14
+inspiredco
+sentence_transformers