Update to 2.5.6a1 (#60)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

Dockerfile

@@ -0,0 +1,25 @@
+FROM python:3.9 as base
+
+ARG PACKAGE_NAME="llama-lang"
+ARG LLAMA_ENVIRONMENT
+
+# Install Ubuntu libraries
+RUN apt-get -yq update
+
+# Install python packages
+WORKDIR /app/${PACKAGE_NAME}
+COPY ./requirements.txt /app/${PACKAGE_NAME}/requirements.txt
+RUN --mount=type=cache,target=/root/.cache/pip \
+    pip install -r requirements.txt
+# Copy all files to the container
+COPY ./scripts /app/${PACKAGE_NAME}/scripts
+COPY ./test /app/${PACKAGE_NAME}/test
+
+WORKDIR /app/${PACKAGE_NAME}
+
+RUN chmod a+x /app/${PACKAGE_NAME}/scripts/start.sh
+
+ENV PACKAGE_NAME=$PACKAGE_NAME
+ENV LLAMA_ENVIRONMENT=$LLAMA_ENVIRONMENT
+
+ENTRYPOINT /app/${PACKAGE_NAME}/scripts/start.sh -e ${LLAMA_ENVIRONMENT}

deploy/build.sh

@@ -0,0 +1,57 @@
+#!/bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# x all executed commands are printed to the terminal
+# u unset variables are errors
+# a export all variables to the environment
+# E any trap on ERR is inherited by shell functions
+# -o pipefail | produces a failure code if any stage fails
+set -Eeuoxa pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+function get_version {
+    local version=$(cat ../pyproject.toml | grep version | sed 's/version[^"]*//g' | sed 's/"//g')
+    echo $version
+}
+
+function get_previous_revision {
+    local revision=$(curl "https://pypi.org/pypi/lamini/json" | python3 -c "import sys, json, re; project = json.load(sys.stdin); versions = list(sorted(project['releases'], key=lambda x : tuple(int(component) for component in x.split('.') if all([x.isnumeric() for x in component]) ))); latest_version = versions[-1]; latest_release = project['releases'][latest_version]; filenames = [re.findall('-\d+-|$', release['filename'])[0].strip('-') for release in latest_release] ; print(max([0] + [int(version) for version in filenames if len(version) > 0]))")
+    echo $revision
+}
+
+function get_old_name {
+    local version="$(get_version)"
+    local old_name="lamini-$version-py3-none-any.whl"
+    echo $old_name
+}
+
+function get_new_name {
+    local previous_revision="$(get_previous_revision)"
+    local next_revision=$((previous_revision + 1))
+    local version="$(get_version)"
+    local new_name="lamini-$version-$next_revision-py3-none-any.whl"
+    echo $new_name
+}
+
+# build
+cd $LOCAL_DIRECTORY/..
+mkdir -p lamini
+touch lamini/__init__.py
+pip3 install wheel build
+python3 -m build --wheel
+cd $LOCAL_DIRECTORY
+
+old_name="$(get_old_name)"
+new_name="$(get_new_name)"
+
+echo "old version $old_name"
+echo "new version $new_name"
+
+# mv the build to the new minor version
+mv $LOCAL_DIRECTORY/../dist/$old_name $LOCAL_DIRECTORY/../dist/$new_name
+
+# upload it
+

deploy/release.sh

@@ -0,0 +1,23 @@
+#!/bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# x all executed commands are printed to the terminal
+# u unset variables are errors
+# a export all variables to the environment
+# E any trap on ERR is inherited by shell functions
+# -o pipefail | produces a failure code if any stage fails
+set -Eeuoa pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+# Get the version number in pyproject.toml
+{
+  git add .
+  git commit -m "v$1"
+  git push
+} || {
+  echo "No Changes to Push"
+}
+gh release create v$1 --title "v$1" --notes "checkpoint"

lamini/__init__.py

@@ -8,7 +8,7 @@
 from lamini.runners.basic_model_runner import BasicModelRunner
 from lamini.runners.mistral_runner import MistralRunner
 from lamini.api.lamini import Lamini
-from lamini.classify.lamini_classifier import LaminiClassifier, BinaryLaminiClassifier
+from lamini.classify.lamini_classifier import LaminiClassifier
 from lamini.api.classifier import Classifier
 from lamini.api.embedding import Embedding
 from lamini.generation.generation_node import GenerationNode
@@ -29,4 +29,3 @@
 
 max_workers = os.environ.get("LAMINI_MAX_WORKERS", 10)
 batch_size = os.environ.get("LAMINI_BATCH_SIZE", 5)
-retry = os.environ.get("LAMINI_RETRY", False)

lamini/api/lamini.py

@@ -8,13 +8,13 @@
 import jsonlines
 import pandas as pd
 from lamini.api.lamini_config import get_config
+from lamini.api.rest_requests import get_version
 from lamini.api.synchronize import sync
 from lamini.api.train import Train
 from lamini.api.utils.async_inference_queue import AsyncInferenceQueue
 from lamini.api.utils.completion import Completion
 from lamini.api.utils.upload_client import get_dataset_name, upload_to_blob
 from lamini.generation.token_optimizer import TokenOptimizer
-from lamini.api.rest_requests import get_version
 
 logger = logging.getLogger(__name__)
 
@@ -67,14 +67,13 @@ def generate(
         metadata: Optional[List] = None,
     ):
         if isinstance(prompt, str) or (isinstance(prompt, list) and len(prompt) == 1):
-            req_data = self.make_llm_req_map(
+            result = self.completion.generate(
                 prompt=prompt,
                 model_name=model_name or self.model_name,
                 output_type=output_type,
                 max_tokens=max_tokens,
                 max_new_tokens=max_new_tokens,
             )
-            result = self.completion.generate(req_data)
             if output_type is None:
                 if isinstance(prompt, list) and len(prompt) == 1:
                     result = [single_result["output"] for single_result in result]
@@ -105,7 +104,7 @@ async def async_generate(
         callback: Optional[Callable] = None,
         metadata: Optional[List] = None,
     ):
-        req_data = self.make_llm_req_map(
+        req_data = self.completion.make_llm_req_map(
             prompt=prompt,
             model_name=model_name or self.model_name,
             output_type=output_type,
@@ -361,22 +360,3 @@ def get_jobs(self):
 
     def evaluate(self, job_id=None):
         return self.trainer.evaluate(job_id)
-
-    def make_llm_req_map(
-        self,
-        model_name,
-        prompt,
-        output_type,
-        max_tokens,
-        max_new_tokens,
-    ):
-        req_data = {}
-        req_data["model_name"] = model_name
-        req_data["prompt"] = prompt
-        req_data["output_type"] = output_type
-        req_data["max_tokens"] = max_tokens
-        if max_new_tokens is not None:
-            req_data["max_new_tokens"] = max_new_tokens
-        if self.model_config:
-            req_data["model_config"] = self.model_config.as_dict()
-        return req_data

lamini/api/lamini_config.py

@@ -69,7 +69,7 @@ def get_configured_url(config):
     if environment == "LOCAL":
         url = config.get("local.url", "http://localhost:5001")
     elif environment == "STAGING":
-        url = config.get("staging.url", "https://api.staging.powerml.co")
+        url = config.get("staging.url", "https://staging.lamini.ai")
     else:
         url = config.get("production.url", "https://api.lamini.ai")
     return url

lamini/api/rest_requests.py

@@ -34,21 +34,6 @@ def check_version(resp):
             print(resp.headers["X-Warning"])
 
 
-def retry_once(func):
-    async def wrapped(*args, **kwargs):
-        try:
-            result = await func(*args, **kwargs)
-        except Exception as e:
-            if lamini.retry:
-                result = await func(*args, **kwargs)
-            else:
-                raise e
-        return result
-
-    return wrapped
-
-
-@retry_once
 async def make_async_web_request(client, key, url, http_method, json=None):
     try:
         headers = {