issues
search
hltcoe
/
sandle
Run a large language modeling SANDbox in your Local Environment
Other
7
stars
1
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Add optional presets to demo, like OpenAI
#57
ccmaymay
opened
2 years ago
0
Use more versatile defaults for demo
#56
ccmaymay
opened
2 years ago
0
Explain `authorized-users.txt` in README
#55
ccmaymay
closed
2 years ago
0
Add SANDLE_SINGLE_MODEL env var.
#54
ccmaymay
closed
2 years ago
0
Use `balanced_low_0` device map strategy
#53
ccmaymay
closed
2 years ago
2
Optionally use 8-bit inference with bits-and-bytes algorithm
#52
ccmaymay
closed
2 years ago
6
Streamline intro and setup section of README.
#51
ccmaymay
closed
2 years ago
0
Streamline README
#50
ccmaymay
closed
2 years ago
0
Add environment variable configuration for single-model deployment
#49
ccmaymay
closed
2 years ago
0
Add Alpa multi-node backend for OPT 175B
#48
ccmaymay
opened
2 years ago
11
Add HF Accelerate multi-node backend for Bloom 176B
#47
ccmaymay
closed
2 years ago
5
Plan for extra-large model support (OPT 175B, Bloom 176B)
#46
ccmaymay
closed
2 years ago
2
Add instructions/extended help to demo
#45
ccmaymay
opened
2 years ago
1
Remove (disabled) from temperature, top p and allow temperature to be zero (greedy)
#44
ccmaymay
closed
2 years ago
0
Update screenshot on README
#43
ccmaymay
closed
2 years ago
0
Standardize display names of models (standardize to opt format)
#42
ccmaymay
closed
2 years ago
0
Keep only one model in GPU memory at a time
#41
ccmaymay
closed
2 years ago
0
Support providing auth tokens in file
#40
ccmaymay
closed
2 years ago
0
Ensure sentry generates an error event on hf backend timeout or error
#39
ccmaymay
closed
2 years ago
0
Finish renaming project to sandle.
#38
ccmaymay
closed
2 years ago
0
Add help text to demo
#37
ccmaymay
closed
2 years ago
0
Allow plaintext (pre-64) passwords in demo
#36
ccmaymay
closed
2 years ago
1
Add support for multiple users
#35
ccmaymay
closed
2 years ago
0
Show status of backend in demo
#34
ccmaymay
closed
2 years ago
4
Fix failures from fuzz test
#33
ccmaymay
closed
2 years ago
1
Deploy on brtx
#32
ccmaymay
closed
1 year ago
3
Add fuzz testing
#31
ccmaymay
closed
2 years ago
1
Make GitHub banner optional
#30
ccmaymay
closed
2 years ago
0
Add support for `n` option (how many completions to generate)
#29
ccmaymay
closed
2 years ago
0
Add warning in demo when prompt ends with space.
#28
ccmaymay
closed
2 years ago
0
Develop stub backend or other workaround for silly people developing on Apple silicon
#27
ccmaymay
closed
2 years ago
0
Rename opt service
#26
ccmaymay
closed
2 years ago
0
Add more informative error message when backend is down
#25
ccmaymay
closed
2 years ago
1
Add HTTP 500 (404? others?) handlers
#24
ccmaymay
closed
2 years ago
0
Handle timeouts in web interface
#23
ccmaymay
opened
2 years ago
2
Simplify: Combine openai-adapter and demo services
#22
ccmaymay
closed
2 years ago
3
Allow non-trivial batching in streaming mode
#21
ccmaymay
closed
2 years ago
0
Fix usage on GPUs without enough memory for the model
#20
ccmaymay
opened
2 years ago
2
Add regression tests against public OpenAI API
#19
ccmaymay
closed
1 year ago
0
Allow disabling sampling (using greedy decoding) in interface
#18
ccmaymay
closed
2 years ago
0
Add button to re-run previous prompt in interface
#17
ccmaymay
closed
2 years ago
0
Check if enter key (shift-enter? ctrl-enter?) submits input
#16
ccmaymay
closed
2 years ago
1
Simplify demo build
#15
ccmaymay
closed
2 years ago
0
Allow user to change top-p, temperature parameters.
#14
ccmaymay
closed
2 years ago
0
Fix streaming performance so we can stream one token at a time, like OpenAI
#13
ccmaymay
opened
2 years ago
0
in API spec, /v1/models return type should be object, not list
#12
ccmaymay
closed
2 years ago
0
/v1/models/<model> does not work
#11
ccmaymay
closed
2 years ago
0
Do not include prompt in non-streaming generated text
#10
ccmaymay
closed
2 years ago
0
Support bloom
#9
ccmaymay
closed
2 years ago
0
Check for stop sequence bugs in streaming implementation
#8
ccmaymay
closed
2 years ago
2
Previous
Next