truefoundry 0.11.9rc1__tar.gz → 0.11.11__tar.gz

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: truefoundry
-Version: 0.11.9rc1
+Version: 0.11.11
 Summary: TrueFoundry CLI
 Author-email: TrueFoundry Team <abhishek@truefoundry.com>
 Requires-Python: <3.14,>=3.8.1

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "truefoundry"
-version = "0.11.9rc1"  # This is a placeholder version, it will be replaced when releasing
+version = "0.11.11"  # This is a placeholder version, it will be replaced when releasing
 requires-python = ">=3.8.1,<3.14"
 description = "TrueFoundry CLI"
 authors = [{ name = "TrueFoundry Team", email = "abhishek@truefoundry.com" }]

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/truefoundry/common/storage_provider_utils.py

@@ -22,6 +22,23 @@ from truefoundry.pydantic_v1 import BaseModel
 logger = logging.getLogger("truefoundry")
 
 
+def truncate_path_for_progress(
+    path: str, max_length: int, relpath: bool = False
+) -> str:
+    if relpath:
+        path = os.path.relpath(path)
+    if len(path) <= max_length:
+        return path
+    parts = path.split(os.sep)
+    result = parts[-1]  # start with filename
+    i = len(parts) - 2
+    # keep prepending directories until adding more would exceed max_len - 3
+    while i >= 0 and len(result) + len(parts[i]) + 1 <= max_length - 3:
+        result = parts[i] + os.sep + result
+        i -= 1
+    return "..." + os.sep + result
+
+
 class MultiPartUploadStorageProvider(str, Enum):
     S3_COMPATIBLE = "S3_COMPATIBLE"
     AZURE_BLOB = "AZURE_BLOB"
@@ -182,7 +199,7 @@ def _file_part_upload(
                 return response
 
 
-def s3_compatible_multipart_upload(
+def s3_compatible_multipart_upload(  # noqa: C901
     multipart_upload: MultiPartUpload,
     local_file: str,
     multipart_info: _FileMultiPartInfo,
@@ -190,13 +207,15 @@ def s3_compatible_multipart_upload(
     progress_bar: Optional[Progress] = None,
     abort_event: Optional[Event] = None,
     exception_class=HttpRequestException,
-):
+) -> None:
     abort_event = abort_event or Event()
     parts = []
 
-    if progress_bar:
+    if progress_bar is not None:
         multi_part_upload_progress = progress_bar.add_task(
-            f"[green]Uploading {local_file}:", start=True
+            f"[green]⬆ {truncate_path_for_progress(local_file, 64, relpath=True)}",
+            start=True,
+            visible=True,
         )
 
     def upload(part_number: int, seek: int) -> None:
@@ -221,7 +240,7 @@ def s3_compatible_multipart_upload(
             multipart_info.num_parts,
             local_file,
         )
-        if progress_bar:
+        if progress_bar is not None:
             progress_bar.update(
                 multi_part_upload_progress,
                 advance=multipart_info.part_size,
@@ -254,10 +273,12 @@ def s3_compatible_multipart_upload(
         timeout=2 * 60,
     )
     response.raise_for_status()
+    if progress_bar is not None:
+        progress_bar.refresh()
     logger.debug("Multipart upload of %s completed", local_file)
 
 
-def azure_multi_part_upload(
+def azure_multi_part_upload(  # noqa: C901
     multipart_upload: MultiPartUpload,
     local_file: str,
     multipart_info: _FileMultiPartInfo,
@@ -265,12 +286,14 @@ def azure_multi_part_upload(
     progress_bar: Optional[Progress] = None,
     abort_event: Optional[Event] = None,
     exception_class=HttpRequestException,
-):
+) -> None:
     abort_event = abort_event or Event()
 
-    if progress_bar:
+    if progress_bar is not None:
         multi_part_upload_progress = progress_bar.add_task(
-            f"[green]Uploading {local_file}:", start=True
+            f"[green]⬆ {truncate_path_for_progress(local_file, 64, relpath=True)}",
+            start=True,
+            visible=True,
         )
 
     def upload(part_number: int, seek: int):
@@ -289,7 +312,7 @@ def azure_multi_part_upload(
             abort_event=abort_event,
             exception_class=exception_class,
         )
-        if progress_bar:
+        if progress_bar is not None:
             progress_bar.update(
                 multi_part_upload_progress,
                 advance=multipart_info.part_size,
@@ -328,4 +351,6 @@ def azure_multi_part_upload(
             timeout=2 * 60,
         )
         response.raise_for_status()
+    if progress_bar is not None:
+        progress_bar.refresh()
     logger.debug("Multipart upload of %s completed", local_file)

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/truefoundry/deploy/_autogen/models.py

@@ -1,6 +1,6 @@
 # generated by datamodel-codegen:
 #   filename:  application.json
-#   timestamp: 2025-08-06T19:58:42+00:00
+#   timestamp: 2025-09-09T12:09:08+00:00
 
 from __future__ import annotations
 
@@ -504,34 +504,49 @@ class Profile(str, Enum):
     """
     Name of the MIG profile to use. One of the following based on gpu type
     Please refer to https://docs.nvidia.com/datacenter/tesla/mig-user-guide/#supported-mig-profiles for more details
-    A100 40 GB - [1g.5gb, 1g.10gb, 2g.10gb, 3g.20gb, 4g.20gb]
-    A100 80 GB / H100 80 GB - [1g.10gb, 1g.20gb, 2g.20gb, 3g.40gb, 4g.40gb]
-    H100 94 GB - [1g.12gb, 1g.24gb, 2g.24gb, 3g.47gb, 4g.47gb]
-    H100 96 GB - [1g.12gb, 1g.24gb, 2g.24gb, 3g.48gb, 4g.48gb]
+    A30 - [1g.6gb, 2g.12gb, 4g.24gb]
+    A100 40 GB - [1g.5gb, 1g.10gb, 2g.10gb, 3g.20gb, 4g.20gb, 7g.40gb]
+    A100 80 GB / H100 80 GB - [1g.10gb, 1g.20gb, 2g.20gb, 3g.40gb, 4g.40gb, 7g.80gb]
+    H100 94 GB - [1g.12gb, 1g.24gb, 2g.24gb, 3g.47gb, 4g.47gb, 7g.94gb]
+    H100 96 GB - [1g.12gb, 1g.24gb, 2g.24gb, 3g.48gb, 4g.48gb, 7g.96gb]
     H200 141 GB - [1g.18gb, 1g.35gb, 2g.35gb, 3g.71gb, 4g.71gb]
+    B200 180 GB - [1g.23gb, 1g.45gb, 2g.45gb, 3g.90gb, 4g.90gb, 7g.180gb]
     """
 
+    field_1g_6gb = "1g.6gb"
+    field_2g_12gb = "2g.12gb"
     field_1g_5gb = "1g.5gb"
     field_1g_10gb = "1g.10gb"
-    field_1g_12gb = "1g.12gb"
-    field_1g_18gb = "1g.18gb"
-    field_1g_20gb = "1g.20gb"
-    field_1g_24gb = "1g.24gb"
-    field_1g_35gb = "1g.35gb"
     field_2g_10gb = "2g.10gb"
-    field_2g_20gb = "2g.20gb"
-    field_2g_24gb = "2g.24gb"
-    field_2g_35gb = "2g.35gb"
     field_3g_20gb = "3g.20gb"
-    field_3g_40gb = "3g.40gb"
-    field_3g_47gb = "3g.47gb"
-    field_3g_48gb = "3g.48gb"
-    field_3g_71gb = "3g.71gb"
     field_4g_20gb = "4g.20gb"
+    field_1g_20gb = "1g.20gb"
+    field_2g_20gb = "2g.20gb"
+    field_3g_40gb = "3g.40gb"
     field_4g_40gb = "4g.40gb"
+    field_1g_12gb = "1g.12gb"
+    field_1g_24gb = "1g.24gb"
+    field_2g_24gb = "2g.24gb"
+    field_3g_47gb = "3g.47gb"
     field_4g_47gb = "4g.47gb"
+    field_3g_48gb = "3g.48gb"
     field_4g_48gb = "4g.48gb"
+    field_1g_18gb = "1g.18gb"
+    field_1g_35gb = "1g.35gb"
+    field_2g_35gb = "2g.35gb"
+    field_3g_71gb = "3g.71gb"
     field_4g_71gb = "4g.71gb"
+    field_1g_23gb = "1g.23gb"
+    field_1g_45gb = "1g.45gb"
+    field_2g_45gb = "2g.45gb"
+    field_3g_90gb = "3g.90gb"
+    field_4g_90gb = "4g.90gb"
+    field_4g_24gb = "4g.24gb"
+    field_7g_40gb = "7g.40gb"
+    field_7g_80gb = "7g.80gb"
+    field_7g_94gb = "7g.94gb"
+    field_7g_96gb = "7g.96gb"
+    field_7g_180gb = "7g.180gb"
 
 
 class NvidiaMIGGPU(BaseModel):
@@ -542,7 +557,7 @@ class NvidiaMIGGPU(BaseModel):
     )
     profile: Profile = Field(
         ...,
-        description="Name of the MIG profile to use. One of the following based on gpu type\nPlease refer to https://docs.nvidia.com/datacenter/tesla/mig-user-guide/#supported-mig-profiles for more details\nA100 40 GB - [1g.5gb, 1g.10gb, 2g.10gb, 3g.20gb, 4g.20gb]\nA100 80 GB / H100 80 GB - [1g.10gb, 1g.20gb, 2g.20gb, 3g.40gb, 4g.40gb]\nH100 94 GB - [1g.12gb, 1g.24gb, 2g.24gb, 3g.47gb, 4g.47gb]\nH100 96 GB - [1g.12gb, 1g.24gb, 2g.24gb, 3g.48gb, 4g.48gb]\nH200 141 GB - [1g.18gb, 1g.35gb, 2g.35gb, 3g.71gb, 4g.71gb]",
+        description="Name of the MIG profile to use. One of the following based on gpu type\nPlease refer to https://docs.nvidia.com/datacenter/tesla/mig-user-guide/#supported-mig-profiles for more details\nA30 - [1g.6gb, 2g.12gb, 4g.24gb]\nA100 40 GB - [1g.5gb, 1g.10gb, 2g.10gb, 3g.20gb, 4g.20gb, 7g.40gb]\nA100 80 GB / H100 80 GB - [1g.10gb, 1g.20gb, 2g.20gb, 3g.40gb, 4g.40gb, 7g.80gb]\nH100 94 GB - [1g.12gb, 1g.24gb, 2g.24gb, 3g.47gb, 4g.47gb, 7g.94gb]\nH100 96 GB - [1g.12gb, 1g.24gb, 2g.24gb, 3g.48gb, 4g.48gb, 7g.96gb]\nH200 141 GB - [1g.18gb, 1g.35gb, 2g.35gb, 3g.71gb, 4g.71gb]\nB200 180 GB - [1g.23gb, 1g.45gb, 2g.45gb, 3g.90gb, 4g.90gb, 7g.180gb]",
     )
 
 
@@ -552,7 +567,7 @@ class NvidiaTimeslicingGPU(BaseModel):
         None,
         description="Name of the Nvidia GPU. One of [P4, P100, V100, T4, A10G, A100_40GB, A100_80GB]\nThis field is required for Node Selector and can be ignored in Nodepool Selector.\nOne instance of the card contains the following amount of memory -\nP4: 8 GB, P100: 16 GB, V100: 16 GB, T4: 16 GB, A10G: 24 GB, A100_40GB: 40GB, A100_80GB: 80 GB",
     )
-    gpu_memory: conint(ge=1, le=200000) = Field(
+    gpu_memory: conint(ge=1, le=400000) = Field(
         ...,
         description="Amount of GPU memory (in MB) to allocate. Please note, this limit is not being enforced today but will be in future. Applications are expected to operate in co-opertative mode",
     )
@@ -681,11 +696,11 @@ class Resources(BaseModel):
         0.5,
         description="CPU limit beyond which the usage cannot be exceeded. 1 CPU means 1 CPU core. Fractional CPU can be requested\nlike `0.5`. CPU limit should be >= cpu request.",
     )
-    memory_request: conint(ge=1, le=2000000) = Field(
+    memory_request: conint(ge=1, le=4000000) = Field(
         200,
         description="Requested memory which determines the minimum cost incurred. The unit of memory is in megabytes(MB).\nSo 1 means 1 MB and 2000 means 2GB.",
     )
-    memory_limit: conint(ge=1, le=2000000) = Field(
+    memory_limit: conint(ge=1, le=4000000) = Field(
         500,
         description="Memory limit after which the application will be killed with an OOM error. The unit of memory is\nin megabytes(MB). So 1 means 1 MB and 2000 means 2GB. MemoryLimit should be greater than memory request.",
     )
@@ -697,7 +712,7 @@ class Resources(BaseModel):
         2000,
         description="Disk storage limit. The unit of memory is in megabytes(MB). Exceeding this limit will result in eviction.\nIt should be greater than the request. This is ephemeral storage and will be wiped out on pod restarts or eviction",
     )
-    shared_memory_size: Optional[conint(ge=64, le=2000000)] = Field(
+    shared_memory_size: Optional[conint(ge=64, le=4000000)] = Field(
         None,
         description="Define the shared memory requirements for your workload. Machine learning libraries like Pytorch can use Shared Memory\nfor inter-process communication. If you use this, we will mount a `tmpfs` backed volume at the `/dev/shm` directory.\nAny usage will also count against the workload's memory limit (`resources.memory_limit`) along with your workload's memory usage.\nIf the overall usage goes above `resources.memory_limit` the user process may get killed.\nShared Memory Size cannot be more than the defined Memory Limit for the workload.",
     )
@@ -1236,7 +1251,7 @@ class ContainerTaskConfig(BaseModel):
         description="Configure environment variables to be injected in the task either as plain text or secrets. [Docs](https://docs.truefoundry.com/docs/env-variables)",
     )
     resources: Optional[Resources] = None
-    mounts: Optional[List[VolumeMount]] = Field(
+    mounts: Optional[List[Union[SecretMount, StringDataMount, VolumeMount]]] = Field(
         None, description="Configure data to be mounted to Workflow pod(s) as a volume."
     )
     service_account: Optional[str] = Field(None, description="")
@@ -1636,7 +1651,7 @@ class PySparkTaskConfig(BaseModel):
         None,
         description="Configure environment variables to be injected in the task either as plain text or secrets. [Docs](https://docs.truefoundry.com/docs/env-variables)",
     )
-    mounts: Optional[List[Union[SecretMount, StringDataMount, VolumeMount]]] = Field(
+    mounts: Optional[List[VolumeMount]] = Field(
         None, description="Configure data to be mounted to Workflow pod(s) as a volume."
     )
     service_account: Optional[str] = Field(None, description="")

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/truefoundry/deploy/builder/builders/tfy_python_buildpack/dockerfile_template.py

@@ -80,7 +80,9 @@ CUDA_VERSION_TO_IMAGE_TAG: Dict[str, str] = {
     # From 12.4+ onwards, the image tags drop the cudnn version
     CUDAVersion.CUDA_12_4_CUDNN9.value: "12.4.1-cudnn-runtime-ubuntu22.04",
     CUDAVersion.CUDA_12_5_CUDNN9.value: "12.5.1-cudnn-runtime-ubuntu22.04",
-    CUDAVersion.CUDA_12_6_CUDNN9.value: "12.6.1-cudnn-runtime-ubuntu22.04",
+    CUDAVersion.CUDA_12_6_CUDNN9.value: "12.6.3-cudnn-runtime-ubuntu22.04",
+    CUDAVersion.CUDA_12_8_CUDNN9.value: "12.8.1-cudnn-runtime-ubuntu22.04",
+    CUDAVersion.CUDA_12_9_CUDNN9.value: "12.9.1-cudnn-runtime-ubuntu22.04",
 }
 
 

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/truefoundry/deploy/cli/commands/patch_command.py

@@ -58,9 +58,19 @@ def patch_command(
     ]
     if output_file:
         with open(output_file, "w") as fd:
-            subprocess.run(yq_command, stdout=fd)
+            p = subprocess.run(
+                yq_command, stdout=fd, stderr=subprocess.PIPE, check=False
+            )
+        if p.returncode != 0:
+            stderr = p.stderr.decode("UTF-8")
+            raise Exception(f"Failed to patch yaml file: {stderr}")
     else:
-        p = subprocess.run(yq_command, stdout=subprocess.PIPE)
+        p = subprocess.run(
+            yq_command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, check=False
+        )
+        if p.returncode != 0:
+            stderr = p.stderr.decode("UTF-8")
+            raise Exception(f"Failed to patch yaml file: {stderr}")
         output = p.stdout.decode("UTF-8")
         callback = RichOutputCallBack()
         callback.print_line(output)

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/truefoundry/deploy/v2/lib/patched_models.py

@@ -91,6 +91,8 @@ class CUDAVersion(str, enum.Enum):
     CUDA_12_4_CUDNN9 = "12.4-cudnn9"
     CUDA_12_5_CUDNN9 = "12.5-cudnn9"
     CUDA_12_6_CUDNN9 = "12.6-cudnn9"
+    CUDA_12_8_CUDNN9 = "12.8-cudnn9"
+    CUDA_12_9_CUDNN9 = "12.9-cudnn9"
 
 
 class GPUType(str, enum.Enum):
@@ -99,12 +101,19 @@ class GPUType(str, enum.Enum):
     V100 = "V100"
     T4 = "T4"
     A10G = "A10G"
+    A10_4GB = "A10_4GB"
+    A10_8GB = "A10_8GB"
+    A10_12GB = "A10_12GB"
+    A10_24GB = "A10_24GB"
     A100_40GB = "A100_40GB"
     A100_80GB = "A100_80GB"
     L4 = "L4"
+    L40S = "L40S"
     H100_80GB = "H100_80GB"
     H100_94GB = "H100_94GB"
+    H100_96GB = "H100_96GB"
     H200 = "H200"
+    B200 = "B200"
 
 
 class TPUType(str, enum.Enum):

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/truefoundry/ml/artifact/truefoundry_artifact_repo.py

@@ -40,6 +40,7 @@ from truefoundry.common.storage_provider_utils import (
     azure_multi_part_upload,
     decide_file_parts,
     s3_compatible_multipart_upload,
+    truncate_path_for_progress,
 )
 from truefoundry.ml._autogen.client import (  # type: ignore[attr-defined]
     ApiClient,
@@ -125,14 +126,14 @@ def _signed_url_upload_file(
             augmented_raise_for_status(response, exception_class=MlFoundryException)  # type: ignore
         return
 
-    task_progress_bar = progress_bar.add_task(
-        f"[green]Uploading {local_file}:", start=True
+    task_id = progress_bar.add_task(
+        f"[green]⬆ {truncate_path_for_progress(local_file, 64, relpath=True)}",
+        start=True,
+        visible=True,
     )
 
     def callback(length):
-        progress_bar.update(
-            task_progress_bar, advance=length, total=os.stat(local_file).st_size
-        )
+        progress_bar.update(task_id, advance=length, total=os.stat(local_file).st_size)
         if abort_event and abort_event.is_set():
             raise Exception("aborting upload")
 
@@ -147,6 +148,9 @@ def _signed_url_upload_file(
         ) as response:
             augmented_raise_for_status(response, exception_class=MlFoundryException)  # type: ignore
 
+    if progress_bar is not None:
+        progress_bar.refresh()
+
 
 def _download_file_using_http_uri(
     http_uri,
@@ -167,11 +171,18 @@ def _download_file_using_http_uri(
         exception_class=MlFoundryException,  # type: ignore
     ) as response:
         augmented_raise_for_status(response, exception_class=MlFoundryException)  # type: ignore
-        file_size = int(response.headers.get("Content-Length", 0))
+        file_size = int(response.headers.get("Content-Length", -1))
+        if file_size == 0 and callback:
+            # special case for empty files
+            callback(1, 1)
+        file_size = file_size if file_size > 0 else 0
         with open(download_path, "wb") as output_file:
             for chunk in response.iter_content(chunk_size=chunk_size):
                 if callback:
                     callback(len(chunk), file_size)
+                    print(
+                        f"Sent callback for {download_path} {len(chunk)} / {file_size}"
+                    )
                 if not chunk:
                     break
                 output_file.write(chunk)
@@ -371,7 +382,7 @@ class MlFoundryArtifactsRepository:
             is MultiPartUploadStorageProvider.AZURE_BLOB
         ):
             azure_multi_part_upload(
-                multipart_upload=MultiPartUpload.parse_obj(multipart_upload.to_dict()),
+                multipart_upload=MultiPartUpload.parse_obj(multipart_upload.dict()),
                 local_file=local_file,
                 executor=executor,
                 multipart_info=multipart_info,
@@ -658,15 +669,18 @@ class MlFoundryArtifactsRepository:
             logger.info("Downloading %s to %s", remote_file_path, local_path)
 
         if progress_bar is not None:
-            download_progress_bar = progress_bar.add_task(
-                f"[green]Downloading to {remote_file_path}:", start=True
+            task_id = progress_bar.add_task(
+                f"[green]⬇ {truncate_path_for_progress(remote_file_path, 64)}",
+                start=True,
+                visible=True,
             )
 
-        def callback(chunk, total_file_size):
+        def callback(chunk_size: int, total_file_size: int):
+            nonlocal task_id
             if progress_bar is not None:
                 progress_bar.update(
-                    download_progress_bar,
-                    advance=chunk,
+                    task_id,
+                    advance=chunk_size,
                     total=total_file_size,
                 )
             if abort_event and abort_event.is_set():
@@ -677,6 +691,9 @@ class MlFoundryArtifactsRepository:
             download_path=local_path,
             callback=callback,
         )
+
+        if progress_bar is not None:
+            progress_bar.refresh()
         logger.debug("Downloaded %s to %s", remote_file_path, local_path)
 
     def _download_artifact(

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/truefoundry/ml/log_types/artifacts/artifact.py

@@ -347,10 +347,11 @@ class ArtifactVersion:
             artifact_version.download(path="<your-desired-download-path>")
             ```
         """
-        download_path = self._download(
+        download_info = self._download(
             path=path, overwrite=overwrite, progress=progress
         )
-        return download_path.download_dir
+        logger.info("Downloaded artifact contents to %s", download_info.download_dir)
+        return download_info.download_dir
 
     def delete(self) -> bool:
         """

{truefoundry-0.11.9rc1 → truefoundry-0.11.11}/truefoundry/ml/log_types/artifacts/model.py

@@ -440,6 +440,7 @@ class ModelVersion:
         download_info = self._download(
             path=path, overwrite=overwrite, progress=progress
         )
+        logger.info("Downloaded model contents to %s", download_info.model_dir)
         return download_info
 
     def delete(self) -> bool: