DeepSpeed
Hybrid Engine Refactor and Llama Inference Support
#3425
Merged
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Overview
Commits
44
Changes
View On
GitHub
Hybrid Engine Refactor and Llama Inference Support
#3425
jeffra
merged 44 commits into
master
from
cholmes/llama-inference-v2
Rebase changes onto correct git history
21b0540e
Further generalize rotate half rotary position embeddings
bd74c32a
Upgrade local clang-format to match CI
56e1de9b
Restore GeGLU behavior and template for SiLU. Add unit test.
264d49b3
Restore experimental qkv reset
14c6a9b7
Switch to named constant to improve readability
b01e7ea7
Name refactor to align with functionality rather than implementation
0215867e
Finish hybrid engine integration
0f371d84
Fix for MLP dimensions
48db17b4
Refactor changes out of base container (#506)
ce19b9c8
Update explanations
88a821df
Complete merge, fix BF16 integration
e42859bd
Merge master
6bad46d1
BF16_AVAILABLE should derive solely from the op_builder
eebfcdf3
Merge remote-tracking branch 'public/master' into cholmes/llama-infer…
6ad5f0f7
Refactor on top of additional model support
63fe26ff
cmikeh2
requested a review
from
jeffra
3 years ago
cmikeh2
requested a review
from
tjruwase
3 years ago
cmikeh2
requested a review
from
RezaYazdaniAminabadi
3 years ago
cmikeh2
requested a review
from
mrwyattii
3 years ago
cmikeh2
requested a review
from
awan-10
3 years ago
cmikeh2
requested a review
from
arashb
3 years ago
Guard is_bf16_supported check
e9137d18
Even stronger guards
838b6f4b
Remove deprecated policy members
a496dd19
Merge branch 'master' into cholmes/llama-inference-v2
123e195a
Another guard
c6703440
Merge branch 'cholmes/llama-inference-v2' of github.com:microsoft/Dee…
6830f5a5
Bad check for BF16 support
708fb456
Merge fix
ace19679
Call correct parent func
439f26f3
Revert API change for TP
81c8e629
Reorder inheritance for Hybrid containers, provide justification
80a2c67b
Add missing policy kwargs
4bd0c6e8
Merge remote-tracking branch 'public/master' into cholmes/llama-infer…
a974832a
Remove debug code
775282a8
awan-10
commented on 2023-05-02
awan-10
commented on 2023-05-02
awan-10
commented on 2023-05-02
awan-10
commented on 2023-05-02
awan-10
commented on 2023-05-02
awan-10
approved these changes on 2023-05-02
BF16 model inference support
9725f09e
Remove debug code
e0e70fe9
Review feedback
9d645153
jeffra
approved these changes on 2023-05-02
Merge branch 'master' into cholmes/llama-inference-v2
c94c67bc
lekurile
approved these changes on 2023-05-02
Fix inheritance
79ad7d0a
jeffra
enabled auto-merge (squash)
3 years ago
Merge branch 'cholmes/llama-inference-v2' of github.com:microsoft/Dee…
a044359b
disabled auto-merge
3 years ago
Manually disabled by user
don't use cache dir for torch installs
5245e0a4
jeffra
requested a review
from
loadams
3 years ago
Merge branch 'master' into cholmes/llama-inference-v2
8c10a296
loadams
approved these changes on 2023-05-03
Align APIs
cd1b6176
Merge branch 'cholmes/llama-inference-v2' of github.com:microsoft/Dee…
149e1937
add HE unit test for OPT
8adc056d
fix typo, missing policy ref to client module
995c3ae2
Merge branch 'master' into cholmes/llama-inference-v2
1596640a
Merge branch 'master' into cholmes/llama-inference-v2
d124817a
jeffra
merged
0a61d5d6
into master
3 years ago
jeffra
deleted the cholmes/llama-inference-v2 branch
3 years ago
Login to write a write a comment.
Login via GitHub
Reviewers
loadams
jeffra
lekurile
awan-10
tjruwase
RezaYazdaniAminabadi
mrwyattii
arashb
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub