transformers: correctly load automodels (#1643)

* backends(transformers): use AutoModel with LLM types

* examples: animagine-xl

* Add codellama examples
This commit is contained in:
Ettore Di Giacinto 2024-01-26 00:13:21 +01:00 committed by GitHub
parent 3733250b3c
commit cb7512734d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
27 changed files with 1144 additions and 569 deletions

View file

@ -0,0 +1,14 @@
name: codellama-7b
backend: transformers
parameters:
model: codellama/CodeLlama-7b-hf
temperature: 0.2
top_k: 40
seed: -1
top_p: 0.95
usage: |
curl http://localhost:8080/v1/completions -H "Content-Type: application/json" -d '{
"model": "codellama-7b",
"prompt": "import socket\n\ndef ping_exponential_backoff(host: str):"
}'