Jonathan Shen
Jonathan Shen
Use the same instance of iSize to fix soumith#155 After fixing this for SpatialConvolution, I went ahead and made the the same changes to other modules. Are there any tests...
I'm not very familiar with how cudnn.SpatialConvolution works, but it seems when an input of a different size comes in, output gets a new reference or something, and so copies...
### What would you like to happen? Currently transforms that return a TimestampedValue need to be typed as plain "TimestampedValue" rather than generic "TimestampedValue[T]" so all underlying information about what...
Otherwise there is an error ``` File "/private/tmp/beam/sdks/python/apache_beam/transforms/userstate.py", line 304, in validate_stateful_dofn if (timer_spec._attached_callback != getattr(dofn, method_name, AttributeError: 'NoneType' object has no attribute '__func__' ``` R: @charlesccychen - [x] [**Choose...
I'm trying to finetune llama with some expanded tokens using `resize_token_embeddings()` and passing `modules_to_save=['embed_tokens', 'lm_head']`, but it seems there is some misconfiguration ``` Traceback (most recent call last): File "/home/jonathanasdf/.local/lib/python3.10/site-packages/transformers/trainer.py",...
``` import guidance import transformers class Transformer(guidance.llms.Transformers): @staticmethod def role_start(role): return role + ':' @staticmethod def role_end(role): return '' def test_healing_prefix(): model_str = 'TheBloke/vicuna-7B-1.1-HF' model = transformers.AutoModelForCausalLM.from_pretrained(model_str, load_in_8bit=True, device_map='auto') tokenizer...
Current examples require hardcoding the number of elements in a list, i.e. ``` [{{#geneach 'items' num_iterations=3}}"{{gen 'this'}}",{{/geneach}}] ``` We may want to let the model decide when to stop generating....
AWS service accounts can assume an identity using AWS_WEB_IDENTITY_TOKEN_FILE and AWS_ROLE_ARN. https://aws.amazon.com/blogs/opensource/introducing-fine-grained-iam-roles-service-accounts A call to `aws sts get-caller-identity` returns the correct assumed role, but `aws.py` doesn't seem to support this...
Runs fine on EC2, but running on Fargate gives ``` _gs.client = storage.Client() File "/usr/local/lib/python3.9/site-packages/google/cloud/storage/client.py", line 161, in __init__ super(Client, self).__init__( File "/usr/local/lib/python3.9/site-packages/google/cloud/client.py", line 319, in __init__ Client.__init__( File "/usr/local/lib/python3.9/site-packages/google/cloud/client.py",...
``` RuntimeError: Failed to import transformers.models.llama.modeling_llama because of the following error (look up to see its traceback): /opt/venv/lib/python3.11/site-packages/flash_attn_2_cuda.cpython-311-x86_64-linux-gnu.so: undefined symbol: _ZN2at4_ops9_pad_enum4callERKNS_6TensorEN3c108ArrayRefINS5_6SymIntEEElNS5_8optionalIdEE ``` I'm using cuda 12.1, pytorch nightly 2.2.0+cu121, flash...