Harrison/deep infra

docs/integrations/deepinfra.md

@@ -7,6 +7,14 @@ It is broken into two parts: installation and setup, and then references to spec
 - Get your DeepInfra api key from this link [here](https://deepinfra.com/).
 - Get an DeepInfra api key and set it as an environment variable (`DEEPINFRA_API_TOKEN`)
 
+## Available Models
+
+DeepInfra provides a range of Open Source LLMs ready for deployment.
+You can list supported models [here](https://deepinfra.com/models?type=text-generation).
+google/flan\* models can be viewed [here](https://deepinfra.com/models?type=text2text-generation).
+
+You can view a list of request and response parameters [here](https://deepinfra.com/databricks/dolly-v2-12b#API)
+
 ## Wrappers
 
 ### LLM

docs/modules/models/llms/integrations/deepinfra_example.ipynb

@@ -81,7 +81,7 @@
    "metadata": {},
    "source": [
     "## Create the DeepInfra instance\n",
-    "Make sure to deploy your model first via `deepctl deploy create -m google/flat-t5-xl` (see [here](https://github.com/deepinfra/deepctl#deepctl))"
+    "You can also use our open source [deepctl tool](https://github.com/deepinfra/deepctl#deepctl) to manage your model deployments. You can view a list of available parameters [here](https://deepinfra.com/databricks/dolly-v2-12b#API)."
    ]
   },
   {
@@ -90,7 +90,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "llm = DeepInfra(model_id=\"DEPLOYED MODEL ID\")"
+    "llm = DeepInfra(model_id=\"databricks/dolly-v2-12b\")\n",
+    "llm.model_kwargs = {'temperature': 0.7, 'repetition_penalty': 1.2, 'max_new_tokens': 250, 'top_p': 0.9}"
    ]
   },
   {
@@ -142,9 +143,20 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "\"Penguins live in the Southern hemisphere.\\nThe North pole is located in the Northern hemisphere.\\nSo, first you need to turn the penguin South.\\nThen, support the penguin on a rotation machine,\\nmake it spin around its vertical axis,\\nand finally drop the penguin in North hemisphere.\\nNow, you have a penguin in the north pole!\\n\\nStill didn't understand?\\nWell, you're a failure as a teacher.\""
+      ]
+     },
+     "execution_count": 8,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
    "source": [
-    "question = \"What NFL team won the Super Bowl in 2015?\"\n",
+    "question = \"Can penguins reach the North pole?\"\n",
     "\n",
     "llm_chain.run(question)"
    ]

langchain/llms/deepinfra.py

@@ -94,7 +94,8 @@ def _call(
 
         if res.status_code != 200:
             raise ValueError("Error raised by inference API")
-        text = res.json()[0]["generated_text"]
+        t = res.json()
+        text = t["results"][0]["generated_text"]
 
         if stop is not None:
             # I believe this is required since the stop tokens