Switch back to gpt2 512 model. Fixes gibberish gpt2 output

[djsamseng]

Switches back to the GPT2 512 model that was used for the demo on the README. While slower, the results match the quality of the README's demo.

Current (distilgpt2_64_6)	These changes (gpt2_512)

[realcarlos]

Hi , I want to try:

private let model = gpt2_512()
public let tokenizer = GPT2Tokenizer()
public let seqLen = 512

but got errors, could you help to tell the what I should do?
/Users/xxx/swift-coreml-transformers/Sources/GPT2.swift:29:25 Cannot find 'gpt2_512' in scope

[djsamseng]

Hi , I want to try:

private let model = gpt2_512()
public let tokenizer = GPT2Tokenizer()
public let seqLen = 512

but got errors, could you help to tell the what I should do? /Users/xxx/swift-coreml-transformers/Sources/GPT2.swift:29:25 Cannot find 'gpt2_512' in scope

Hello!
Did you add the model to the project/build? https://github.com/huggingface/swift-coreml-transformers/pull/29/files#diff-2d85bd611302dc0e5de877b114c903d4047fccb7312ce3f9125b52bfc3144194R71

[realcarlos]

Hi , I want to try:

private let model = gpt2_512()
public let tokenizer = GPT2Tokenizer()
public let seqLen = 512

but got errors, could you help to tell the what I should do? /Users/xxx/swift-coreml-transformers/Sources/GPT2.swift:29:25 Cannot find 'gpt2_512' in scope

Hello! Did you add the model to the project/build? https://github.com/huggingface/swift-coreml-transformers/pull/29/files#diff-2d85bd611302dc0e5de877b114c903d4047fccb7312ce3f9125b52bfc3144194R71

yes, I have the model , it seems I missed the file which having gpt2_512() , could you tell me which file I should add?

[djsamseng]

Should be this one https://github.com/huggingface/swift-coreml-transformers/blob/master/Resources/gpt2-512.mlmodel

[realcarlos]

Should be this one https://github.com/huggingface/swift-coreml-transformers/blob/master/Resources/gpt2-512.mlmodel

I add this file to Compile Sources, but still failed:

2023-05-18 21:51:38.157876+0800 CoreMLGPT2[38603:9307067] Metal API Validation Enabled
2023-05-18 21:51:51.294500+0800 CoreMLGPT2[38603:9307067] [espresso] [Espresso::handle_ex_plan] exception=ANECF error: failed to load ANE model. Error=ANECCompile(/var/mobile/Library/Caches/com.apple.aned/tmp/ai.katia.Katia.CoreMLGPT2/6F284D8730500443BBF24C75BC12B6C3C41CC90940F082B78438ACA25771C66B/1D4509F6B6CDC93B38ECBC87E6282B6193403F9DA75BD53D25DD3A58D2AC3AE1/) FAILED: err=(
CompilationFailure
)
2023-05-18 21:51:51.296062+0800 CoreMLGPT2[38603:9307067] [coreml] Error plan build: -1.
2023-05-18 21:51:51.308255+0800 CoreMLGPT2[38603:9307067] [client] doUnloadModel:options:qos:error:: nil _ANEModel
2023-05-18 21:51:51.308282+0800 CoreMLGPT2[38603:9307067] [espresso] ANECF error:

[kpacholak]

+1 from me