v2.4.0 #250

p-ferreira · 2024-06-05T13:50:15Z

What's Changed

Minor vllm optimization, code style fixes by @dbobrenko in Minor vllm optimization, code style fixes #236
Update miner requirements, increase storage by @dbobrenko in Update miner requirements, increase storage #237
Fix repeated tasks on CUDA OOM by @dbobrenko in Fix repeated tasks on CUDA OOM #243
Fix Llama3 miner 4bit flag, update requirements, etc by @dbobrenko in Fix Llama3 miner 4bit flag, update requirements, etc #235
Update documentation for base miners by @bkb2135 in Features/discourage base miners #244
Bittensor 7.0.0, vllm 0.4.2 upgrade by @dbobrenko in Bittensor 7.0.0, vllm 0.4.2 upgrade #246
Update forward.py by @steffencruz in Update forward.py #251
Increase miner sampling for each step by @bkb2135 in Increase miner sampling for each step #249
Features/streaming penalty by @p-ferreira in Features/streaming penalty #234

Remerge main into pre-staging

Minor vllm optimization, code style fixes

Update miner requirements, increase storage

Fix repeated tasks on CUDA OOM

Fix Llama3 miner 4bit flag, update requirements, etc

…ners Features/discourage base miners

Bittensor 7.0.0, vllm 0.4.2 upgrade

Reduces max conversation turns from 3 to 2.

…turns Update forward.py

…ampling Sample All Miners Each Step

Features/streaming penalty

…grade update bittensor requirements

Hotfix/tokenizer issue

p-ferreira and others added 30 commits May 16, 2024 19:14

clean up legacy prompting synapse

9b33454

initial adjustments to protocol + chunks logging

5d3e1b9

fix chunk processing + mock

b6b7e22

adjust dendrite response

f3e5466

refactors base pipeline apply function

36344f0

adds tokenizer to flow + dendrite refactor

b269365

implements streaming reward function

7227814

Merge pull request #225 from opentensor/main

dfe2de7

Remerge main into pre-staging

adds tokens_per_chunk in forward flow + adds global penalties

a47bdf1

Increase default max memory to 80 GB

c5d9d06

Change example run to 1 GPU

02cd837

Fix validator hotkeys IndexError

107a4d6

Revert vllm changes

807f162

Remove redundant validator code

77c913e

Fix comment style

8c2cfce

Restore validator changes

a908392

Restore config.py

6c9eed9

Remove clean function

44ec400

Update comments

07ff31a

Update miner requirements, increase minumum storage

154a5ef

drops synapse final data out of accumulated chunks

919780d

Set maximum GPU mem to 65

28fe4bc

Set maximum GPU mem to 62

43862f6

Decrease min memory to 62

5337dde

Merge pull request #236 from dbobrenko/hotfix/vllm-optimize

dcb891a

Minor vllm optimization, code style fixes

Merge pull request #237 from dbobrenko/feature/llama-requirements

afdf322

Update miner requirements, increase storage

Merge branch 'main' into features/stream-adjustments

a737825

adds unit tests for streaming reward model

4b862e7

Properly update removed or replaced hotkeys

da4a066

Set scores device

4d292f5

p-ferreira and others added 10 commits June 4, 2024 14:48

fix stream synpase import

ff0de98

fix mock dendrite call

1d3e6bc

Sample all available uids

e88e694

Do not exclude uids that were just queried

ee72240

Merge pull request #243 from macrocosm-os/hotfix/oom-repeated-tasks

db3ef35

Fix repeated tasks on CUDA OOM

Merge pull request #235 from dbobrenko/hotfix/hotkeys-index-error

1be9848

Fix Llama3 miner 4bit flag, update requirements, etc

Merge branch 'staging' into features/discourage-base-miners

7925dcd

Merge branch 'staging' into feature/bittensor-7.0.0

51e11dd

Merge pull request #244 from macrocosm-os/features/discourage-base-mi…

e77eeef

…ners Features/discourage base miners

Merge branch 'staging' into features/stream-adjustments

dfb1c76

p-ferreira added the 2.4.0 label Jun 5, 2024

Merge pull request #246 from macrocosm-os/feature/bittensor-7.0.0

4e05207

Bittensor 7.0.0, vllm 0.4.2 upgrade

p-ferreira changed the title ~~v2.3.2~~ v2.4.0 Jun 5, 2024

p-ferreira and others added 7 commits June 5, 2024 09:55

updates versioning

1961f6c

Merge branch 'staging' into features/stream-adjustments

41d1c67

Update config.py

54932c7

Update forward.py

ec2b764

Reduces max conversation turns from 3 to 2.

Merge pull request #251 from macrocosm-os/hotfix/reduce-conversation-…

b1cbef3

…turns Update forward.py

Merge pull request #249 from macrocosm-os/feature/remove-random-uid-s…

2aa77c8

…ampling Sample All Miners Each Step

Merge pull request #234 from macrocosm-os/features/stream-adjustments

834d70c

Features/streaming penalty

p-ferreira requested review from steffencruz, bkb2135 and dbobrenko June 5, 2024 18:01

p-ferreira marked this pull request as ready for review June 5, 2024 18:01

p-ferreira added 2 commits June 5, 2024 14:17

update bittensor requirements

8d38f12

Merge pull request #252 from macrocosm-os/hotfix/bittensor-package-up…

794d3f7

…grade update bittensor requirements

steffencruz approved these changes Jun 5, 2024

View reviewed changes

p-ferreira and others added 2 commits June 5, 2024 19:19

fix tokenizer issue, fix logging issue, adapt mock miner for unit test

ea0843d

Merge pull request #253 from macrocosm-os/hotfix/tokenizer-issue

29002ea

Hotfix/tokenizer issue

p-ferreira merged commit 6c98a6f into main Jun 5, 2024
2 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

v2.4.0 #250

v2.4.0 #250

p-ferreira commented Jun 5, 2024 •

edited

Loading

v2.4.0 #250

v2.4.0 #250

Conversation

p-ferreira commented Jun 5, 2024 • edited Loading

What's Changed

p-ferreira commented Jun 5, 2024 •

edited

Loading