Skip to content

Gemma3 #402

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
May 14, 2025
Merged

Gemma3 #402

merged 5 commits into from
May 14, 2025

Conversation

mohiso22
Copy link
Contributor

No description provided.

quic-rishinr and others added 5 commits May 14, 2025 08:25
Signed-off-by: Rishin Raj <quic_rishinr@quicinc.com>
Signed-off-by: Mohit Soni <quic_mohisoni@quicinc.com>
This PR addresses an issue causing the qconfig dump to fail when QAIC is
not installed but QNN is present. With this update, the qconfig will be
successfully dumped in both QAIC and QNN environments.

---------

Signed-off-by: Abukhoyer Shaik <quic_abukhoye@quicinc.com>
Signed-off-by: Mohit Soni <quic_mohisoni@quicinc.com>
As part of Model-IP I/O encryption feature qaic-exec has added a new
command line option: `-io-encrypt` to enable support for
decryption/encryption of network I/O.
Added support in Qeff to pass this param to qaic-exec when a model needs
to be compiled with I/O encryption support.

CLI command: `python -m QEfficient.cloud.infer --model_name gpt2
--batch_size 1 --prompt_len 32 --ctx_len 128 --mxfp6 --num_cores 16
--device_group [0] --prompt "My name is" --mos 1
--aic_enable_depth_first --io-encrypt "chacha20"`

Note: If the flag is passed via infer CLI, CLI will exit after QPC
generation.

---------

Signed-off-by: Asmita Goswami <quic_asmigosw@quicinc.com>
Signed-off-by: Mohit Soni <quic_mohisoni@quicinc.com>
- Refactor the finetune main api
- Add support to override the PEFT config (yaml/json)
- Add support to validate the correctness of PEFT Config
- Some nit changes

```yaml
r: 16
lora_alpha: 64
target_modules:
  - q_proj
  - v_proj
  - k_proj
bias: none
task_type: CAUSAL_LM
lora_dropout: 0.1
```
Command:
```bash
python -m QEfficient.cloud.finetune \
    --model_name "meta-llama/Llama-3.2-1B" \
    --lr 5e-4 \
    --peft_config_file "lora_config.yaml"
```

#### Using Default LoRA Config:
```bash
python -m QEfficient.cloud.finetune \
    --model_name "meta-llama/Llama-3.2-1B" \
    --lr 5e-4
```

---------

Signed-off-by: vbaddi <quic_vbaddi@quicinc.com>
Signed-off-by: Meet Patel <quic_meetkuma@quicinc.com>
Co-authored-by: Meet Patel <quic_meetkuma@quicinc.com>
Signed-off-by: Mohit Soni <quic_mohisoni@quicinc.com>
Signed-off-by: Mohit Soni <quic_mohisoni@quicinc.com>
@mohiso22 mohiso22 marked this pull request as ready for review May 14, 2025 08:32
@mohiso22 mohiso22 merged commit 70ae12f into quic:add_gemma3 May 14, 2025
3 checks passed
quic-amitraj added a commit that referenced this pull request May 14, 2025
@quic-amitraj quic-amitraj mentioned this pull request May 14, 2025
vbaddi added a commit that referenced this pull request May 14, 2025
mohiso22 added a commit that referenced this pull request May 14, 2025
Signed-off-by: Rishin Raj <quic_rishinr@quicinc.com>
Signed-off-by: Mohit Soni <quic_mohisoni@quicinc.com>
Signed-off-by: Abukhoyer Shaik <quic_abukhoye@quicinc.com>
Signed-off-by: Asmita Goswami <quic_asmigosw@quicinc.com>
Signed-off-by: vbaddi <quic_vbaddi@quicinc.com>
Signed-off-by: Meet Patel <quic_meetkuma@quicinc.com>
Co-authored-by: Rishin Raj <quic_rishinr@quicinc.com>
Co-authored-by: Abukhoyer Shaik <quic_abukhoye@quicinc.com>
Co-authored-by: asmigosw <quic_asmigosw@quicinc.com>
Co-authored-by: Vinayak Baddi <68580231+vbaddi@users.noreply.github.com>
Co-authored-by: Meet Patel <quic_meetkuma@quicinc.com>
Signed-off-by: Mohit Soni <mohisoni@qti.qualcomm.com>
mohiso22 pushed a commit that referenced this pull request May 14, 2025
This reverts commit 70ae12f.

Signed-off-by: Mohit Soni <mohisoni@qti.qualcomm.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants