search

MainRo / docker-deepspeech-server

A dockerfile to run deepspeech-server
Mozilla Public License 2.0
29 stars 13 forks source link

Issue with versions? #4

Open ppfearn opened 5 years ago

ppfearn commented 5 years ago

I've been trying to get this working but I receive a strange error on startup: Top line(s): deepspeech-server | Exception in callback None() deepspeech-server | handle: <Handle cancelled> deepspeech-server | Traceback (most recent call last): deepspeech-server | File "/usr/local/lib/python3.7/asyncio/events.py", line 88, in _run

Final line: deepspeech-server | AttributeError: 'x' object has no attribute 'lm_alpha'

I noticed that it's been a while since the last update and I can see that the latest python is being pulled. Is there a possible issue with versions?

MainRo commented 5 years ago

if you re-generated the docker image recently, then it uses the latest release of deepspeech-server, where some fields of the configuration file have been renamed to match the ones used in the deepspeech API. typically lm_alpha replaced lm_weight. Do you use a configuration file from deepspeech-server v1.0.0 or below ?

MainRo commented 5 years ago

btw to avoid such issues, I should version this repo based on a fixed version of deepspeech and deepspeech-server.

ppfearn commented 5 years ago

if you re-generated the docker image recently, then it uses the latest release of deepspeech-server, where some fields of the configuration file have been renamed to match the ones used in the deepspeech API. typically lm_alpha replaced lm_weight. Do you use a configuration file from deepspeech-server v1.0.0 or below ?

I haven't gotten that far yet as it errors before any config issue. I have managed to run the deepspeech server manually (a while ago) by installing the whole thing myself, so I was just going to use that config (after figuring out how to put it into docker-compose.yaml). That config looks like the following:

{ "deepspeech": { "model" :"models/output_graph.pb", "alphabet": "models/alphabet.txt", "lm": "models/lm.binary", "trie": "models/trie", "features": { "n_features": 26, "n_context": 9, "beam_width": 500, "lm_weight": 1.5, "vwc_weight": 2.25 } }, "server": { "http": { "host": "0.0.0.0", "port": 8080, "request_max_size": 1048576 } }, "log": { "level": [ { "logger": "deepspeech_server", "level": "DEBUG"} ] } }

My current docker compose yaml will get the latest version it can. It looks like the following:

version: "2" services: deepspeech_server: container_name: deepspeech-server build: ./deepspeech_server ports:

(sorry for the terrible indenting, I couldn't get it to be code and on separate lines)

btw to avoid such issues, I should version this repo based on a fixed version of deepspeech and deepspeech-server.

Which versions should I use in your opinion? This whole process is about trying to get something that's easier to handle as deepspeech does appear to move at a relatively quick pace. If I can get it handled in docker, I'm hoping that I can check versions easily, but always know that I can use a fallback set of versions if it doesn't work.

Thanks. All help is appreciated with this