Initial publish via teradata-opus-translate

README.md

@@ -67,6 +67,10 @@ Generation parameters are configurable at SQL time via the
 Requires Teradata 17.20+ with **BYOM 7.0.0.4** or newer (the conversion
 targets ONNX IR version 8, which BYOM 7.0.x requires).
 
+> **Note on schema name:** the SQL example below uses `mldb.ONNXSeq2Seq`.
+> On modern Teradata deployments BYOM is installed in the `td_mldb` database —
+> adjust the schema prefix in the SQL accordingly.
+
 ```python
 import getpass
 import teradataml as tdml
@@ -76,7 +80,7 @@ repo_id  = "Teradata/opus-mt_tiny_eng-deu"
 model_id = "opus-mt_tiny_eng-deu"          # used as BYOM model_id
 
 # 1. Download artifacts from this repo
-hf_hub_download(repo_id=repo_id, filename="onnx/model-int8.onnx", local_dir="./")
+hf_hub_download(repo_id=repo_id, filename="onnx/model-fp32.onnx", local_dir="./")
 hf_hub_download(repo_id=repo_id, filename="tokenizer.json",       local_dir="./")
 
 # 2. Connect to Teradata
@@ -87,7 +91,7 @@ tdml.create_context(
 )
 
 # 3. Load model + tokenizer into BYOM tables
-tdml.save_byom(model_id=model_id, model_file="onnx/model-int8.onnx",
+tdml.save_byom(model_id=model_id, model_file="onnx/model-fp32.onnx",
                table_name="translation_models")
 tdml.save_byom(model_id=model_id, model_file="tokenizer.json",
                table_name="translation_tokenizers")
@@ -98,15 +102,15 @@ SELECT id, sequences
 FROM mldb.ONNXSeq2Seq(
     ON (SELECT id, txt FROM your_input_table) AS InputTable
     ON (SELECT model_id, model FROM translation_models
-         WHERE model_id = '{{model_id}}') AS ModelTable DIMENSION
+         WHERE model_id = '{model_id}') AS ModelTable DIMENSION
     ON (SELECT model AS tokenizer FROM translation_tokenizers
-         WHERE model_id = '{{model_id}}') AS TokenizerTable DIMENSION
+         WHERE model_id = '{model_id}') AS TokenizerTable DIMENSION
     USING
         Accumulate('id')
         ModelOutputTensor('sequences')
         SkipSpecialTokens('true')
         OutputLength(512)
-        OverwriteCachedModel('{{model_id}}')
+        OverwriteCachedModel('true')
         Const_min_length(1)
         Const_max_length(64)
         Const_num_beams(4)
@@ -117,7 +121,7 @@ FROM mldb.ONNXSeq2Seq(
 print(tdml.DataFrame.from_query(query))
 ```
 
-Use `model-int8.onnx` unless you have a measured accuracy reason to ship `fp32`.
+An int8-quantized variant is also published as `onnx/model-int8.onnx`. The int8 variant does not accept `num_beams` (configured internally).
 
 ## How this model was converted
 
@@ -137,8 +141,7 @@ small sample set.
 >
 > convert_model(
 >     "Helsinki-NLP/<your-model>",
->     output_path="model-int8.onnx",
->     precision="int8",
+>     output_path="model-fp32.onnx",
 > )
 > convert_tokenizer(
 >     "Helsinki-NLP/<your-model>",
@@ -146,7 +149,7 @@ small sample set.
 > )
 > ```
 >
-> The resulting `model-int8.onnx` and `tokenizer.json` are ready to deploy
+> The resulting `model-fp32.onnx` and `tokenizer.json` are ready to deploy
 > with the Quickstart flow above.
 
 ## Disclaimer

onnx/model-int8.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:82246af52158c0692c87263c545f491747f60517fd4e9199f732b3ce5a776fe2
-size 94483810
+oid sha256:0edaff9b6d878d8fb3c2226135185419603bc2ee78ef92f192a3e52b77bafe45
+size 94960801